PDA

View Full Version : New Revised Methodology DOT Ratings


Wuornos
11-29-2007, 03:55 AM
Revised DOT Ratings

This is the first post of the revised DOT Ratings. I will not go into the methodology behind these again as this is already contained within a thread at http://tt.tennis-warehouse.com/showthread.php?t=165420.

I have just finished revising some of the methodological procedures within the DOT ratings which should have increased accuracy markedly. As a result it is now possible to provide reasonably accurate data from a sample of a single year without compromising the results. I have therefore revised the systems to provide all data on this basis and like the official rankings DOT Ratings are now based upon a rolling horizon of 12 months with all results having equal weighting chronologically within this sample although differing weights are applied as calculated by the DOT Algorithm for various tournaments. This will of course make these results inconsistent with previous results posted but with an improved level of accuracy for all output.


The Current Top 10 Women Players as calculated by the DOT Ratings.

1. Justine Henin 2660
2. Serena Williams 2599
3. Venus Williams 2592
4. Svetlana Kuznetsova 2578
5. Ana Ivanović 2573
6. Maria Sharapova 2569
7. Anna Chakvetadze 2559
8. Nicole Vaidišová 2552
9. Jelena Janković 2551
10 Marion Bartoli 2547

As you can see these new DOT Ratings now appear almost as a hybrid between the old DOT Ratings and the official WTA World Rankings.

Maria Sharapova, who was rate at number 2 in the Old DOT ratings because of the necessary longer time period for ratings to achieve the required level of accuracy, has dropped down to number 6 which is more in line with her official rank.

Jelena Janković who previously was not in the top 10 has now arrived at number 9, although this is still lower than her official rank. I believe this is because DOT Ratings are less susceptible to inflation due to high activity.

Svetlana Kuznetsova is now at number 4 which is higher than the old DOT Rating methodology but lower than her official rank. Again at first glance this appears to be a hybrid but is again due to the shorter timescale feeding into the new DOT Ratings coupled with the volume effect of the official rankings not being present within the New DOT Ratings allowing the Williams Sisters to climb above Svetlana.

Finally we now see Anna Chakvetadze appearing in the new DOT list at #7 when previously she was outside the top 10.

By and large these are not so different from the official rankings but with the benefit that they are consistent across time for earlier periods of the open era and are adjusted for the level of opposition so are not fixed levels in the same way that the official rankings are. Instead they are adjusted to be consistent at a lower single point in the population as these levels remain more constant within a changing population profile.

On to how these stack up in relation to the peak ratings of the open era.

1. Martina Navratilova 2805 Year Ending 1984 US Open
2. Steffi Graf 2798 Year Ending 1989 Australian Open
2. Margaret Smith Court 2798 Year Ending 1971 Australian Open
4. Serena Williams 2789 Year Ending 2003 Australian Open
5. Monica Seles 2778 Year Ending 1993 Australian Open
6. Chris Evert 2759 Year Ending 1983 French Open
7. Martina Hingis 2740 Year Ending 1998 Australian Open
8. Justine Henin 2728 Year Ending 2004 Australian Open
9. Billie Jean King 2723 Year Ending 1972 US Open
10. Hana Mandlíková 2700 Year Ending 1981 Wimbledon


There is still very little to pick between the top three despite the improved levels of accuracy although the top player has changed from Margaret Court to Martina Navratilova. Margaret Court and Steffi Graff are now tied at 2nd equal.

Chris Evert has dropped from 4th to 6th. The new algorithm now preferring both Monica Seles and Serena Williams to the Ice Queen. Serena jumps to number 4, sh may not be popular, but boy was she producing some great results during her peak year. Monica Seles remains at 5th place.

Into the top 10 comes Justine Henin at the expense of Arantxa Sánchez Vicario who drops to 12th. As can be seen in the current list Henin is currently playing at 2660, 68 points behind her peak. I doubt she will be able to overhaul her current peak on the back of an Australian Open win in 2008 but much depends on how the DOT Rating system evaluates the tournament. If we see the three of the current top 4 in the semi finals and Henin emerges victorious, DOT will place far greater emphasis on it as an indicator of quality play and therefore it might be possible , but we will just have to wait and see.

Continued ....

Wuornos
11-29-2007, 03:57 AM
Moving on to the Mens’ Current DOT Ratings.

1. Roger Federer 2768
2. Rafael Nadal 2656
3. Novak Đoković 2617
4. Nikolay Davydenko 2584
5. Andy Roddick 2576
6. Fernando González 2575
7. Tommy Haas 2562
8. Richard Gasquet 2551
9. David Ferrer 2548
10. Carlos Moyŕ 2546

No real surprises in the top 4. I haven’t checked but I think Andy Roddick might be ranked 6th at the moment in the official rankings and he is 5th here. Hardly a startling difference.

I know many will query the lack of David Nalbandian. Like the official rankings the DOT Ratings look for results over a period of 1 year and while David’s results have been very good recently this level would need to be maintained for a longer period before he appears in the DOT top 10.

No onto the Peak Open Era DOT Ratings for men. I know one of these outputs is going to be very controversial but overall I think the quality has improved. Several players have now fallen out of the DOT Rating top 10 as a result of the improved methodology and subsequent accuracy while other have entered. By and large these changes are in keeping with popular opinion and with the one exception should provide a list which while never pleasing everyone will be consistent with more opinions.

1. Rod Laver 2781 Achieved year ending US Open 1969
2. Roger Federer 2768 Achieved year ending US Open 2007
3. Mats Wilander 2722 Achieved year ending US Open 1988
4. Pete Sampras 2715 Achieved year ending Wimbledon 1994
5. John McEnroe 2709 Achieved year ending US Open 1984
6. Andre Agassi 2708 Achieved year ending Australian Open 2000
7. Ivan Lendl 2707 Achieved year ending Australian Open 1988
8. Jim Courier 2687 Achieved year ending French Open 1992
9. Jimmy Connors 2683 Achieved year ending US Open 1974
10. Björn Borg 2682 Achieved year ending Wimbledon Open 1979

Ok, I know you’re all going to hate the position of Wilander. But the same criteria has been applied to all of the above players. i.e Dominance x Opposition x Tournament Status

Lets look at some of the differences between this list and that generated by the old DOT methodology.

GOAT has changed Rod Laver now having the # 1 position but with Federer within a hairs breadth of this standard.

Wilander aside Sampras now ranks behind only Federer and Laver. Previously DOT ranked him 6th, so I feel this climb in the rankings is likely to meet with majority approval.

McEnroe climbs to number 5 based on his outstanding 1984 year, previously he was ranked at #7 by the old DOT Ratin methodology. Again I feel this will meet with popular approval.

The new methodology and resulting accuracy has now identified why many fans rate Agassi highly. He is now at # 6 in the peak ratings while previously he was languishing at 15th. Again I feel this will meet with majority approval.

Ivan Lendl has dropped from 3rd to 7th in the revised methodology. I have to say I am a fan of Lendl and I am disappointed in this result but again I feel this is more in keeping with popular opinion.

In 8th place is Jim Courier. This position is similar to the old ratings and has already been debated extensively with the thread ‘The underrating past greats’ so I don’t propose to go into it again here. I know it’s an unpopular outcome but DOT still rates Jim Courier up there based on his domination against whom and within what tournaments.

Like Agassi the revisions to methodology have now enabled the DOT Ratings to identify more the fans assessment of Jimmy Connors. He climbs from 14th to 9th as a result of the improved DOT accuracy again I feel this is more in keeping with popular opinion.

Björn Borg has suffered as a result of the revisions to the algorithm. He is now ranked 10th as opposed to his previously high 4th place. In some ways I feel this is more intuitive for what we would expect to see. DOT is looking for peak standard and McEnroe defeated Borg at his peak in two consecutive majors. The previous DOT Rankings which put Borg above McEnroe because of the longer timescales involved were I feel little out of kilter with the main premise of dominance being one of the three primary criteria in the evaluation of a player. However I believe this will not be the popular view which will tend to favour Borg.

Out of the top 10 all together go Ken Rosewall (now 14th) and Boris Becker (now 12th) in making way for Agassi and Connors. I think this swap will be well received by people, but please let me have your views.

There it is then. Dominance x Opposition x Tournament Status for all the top players of the open era. Improvements in methodology seem to move the results more in line with popular opinion but in adopting these criteria as a measurement of current playing standard and therefore deriving peak playing standards, in the name of consistency and fairness DOT must rate Wilander and Courier higher than popular opinion might dictate.

Finally I would just like to say. Even if you had a magic box that could mystically produce absolute figures regarding peak playing standards for every player and the box was never wrong, I think we would still get some surprises with some players being more highly though of than they deserve and vice versa. I’m not saying DOT is right, I am just saying that no matter how good you make this process will always be many players for whom a few will say ‘No Way’ and a few players for whom the majority will say ‘What! Really?’

One last question. I will continue answering threads and will continue to use DOT to support my arguments, because to be honest, I actually value the DOT output now as probably being superior to my own opinion. However I was just wondering whether users of this forum find the above type of analysis interesting in it’s own right and therefore whether people would want me to continue posting further developments and analysis as threads in their own right.

If you’ve managed to read this far, many thanks and finally please feel free to let me have your opinion on the above outputs, methodologies or validity of any aspect of the above.

Regards

Tim

pabloJD
11-29-2007, 09:01 AM
Hi, Tim. Very interesting stuff, as usual.
I guess that with the new methodology peak rating equals best non-calendar year. If it is so then it's not such a scandal that Bjorn Borg, to give an example, has the 9th best year for a player. His greatest accomplishment was winning a lot of Wimbledons and French Opens, which he did during the course of several years. The reverse example would be Serena Williams who won half her GS tournaments in 12 months.

Wuornos
11-29-2007, 09:54 AM
Hi, Tim. Very interesting stuff, as usual.
I guess that with the new methodology peak rating equals best non-calendar year. If it is so then it's not such a scandal that Bjorn Borg, to give an example, has the 9th best year for a player. His greatest accomplishment was winning a lot of Wimbledons and French Opens, which he did during the course of several years. The reverse example would be Serena Williams who won half her GS tournaments in 12 months.

Yes Pablo. That's absolutely correct. Peak rating equals best non calendar year adjusted for Opposition. The two exapmles you chose are good examples of this. Sometimes, however, the quality of opposition at the top of the sport is sufficient for a less dominant year to be considered better than a more dominant one. E.g Andy Roddick's performance in the year ending with Wimbledon of 2004 was probably less dominant than LLeyton Hewitt's performance in the year ending with Wimbledon 2002. This is due to the presence of Federer in Roddick peak year while Hewitt only had Marat Safin and Albert Costa to contend with. Admittedly Pete Sampras and Andre Agassi were still competing during this year but they neither were the force they had been. Consequently we see Andy Roddick's peak year rated at 2626 compared with LLeyton Hewitt's 2619. This is an example of an opposition effect.

Finally I would just like to spell out that the system is adjusted for the relative value of tournaments and uses it's own algorithm to do this. Borg would have benefitted from the Australia Open being much devalued during his time, while the other majors would have been inflated. The actual weightings show that the Australian Open was worth 18% of the years major rating points in 1979 compared with 25% today. French and Wimbledon in 1979 were worth 63% between them. So as you can see the weightings are adjusted for the poor status of Australian Open in Borg's era. Borg's problems in the 'eyes of DOT were

1. That he needed to win the US Open

and

2. When the challenge was posed from McEnroe which would have enabled him to benefit from the opposition effect as described above he was unable to answer the call.

Thanks for your interest and kind comments and keep on posting.

Regards

Tim

akv89
11-29-2007, 12:11 PM
Thanks for the effort Tim. I can't argue with your rankings for peak performance, but I don't think that peak performance over a single year correlates with overall greatness in tennis history. I would be very interested to see how the ratings change if the time interval for the DOT rating was changed from the best year to the best five years (consecutive or not) that a tennis player had, which I feel not only takes into account their peak performance, but also how well they were able to maintain that performance.

Steve132
11-29-2007, 02:55 PM
Tim: Some fascinating results.

I see these ratings as complementing rather than replacing the old DOT ratings. The new ratings show how great a player was at a particular point in time, while the old ones showed how good the player was at sustaining his or her peak achievement over a period. My view is that we need BOTH in order to assess how great a player is.

Wuornos
11-30-2007, 01:51 AM
Thanks for the effort Tim. I can't argue with your rankings for peak performance, but I don't think that peak performance over a single year correlates with overall greatness in tennis history. I would be very interested to see how the ratings change if the time interval for the DOT rating was changed from the best year to the best five years (consecutive or not) that a tennis player had, which I feel not only takes into account their peak performance, but also how well they were able to maintain that performance.

Yes, you're not the first person to say that.

I'm curently working on the systems I use to produce DOT to enable them to produce ratings for Single Year, Five Year and Career. Meaning each player will in effect have three ratings.

I am also considering a fourth category which is 'single event'. This would look to see which event resulted in the most DOT Rating Points for specific players. These events would tend to be the majors where all the top players were present and made the final stages, meaning we got to see how the best at that time performed with the highest level of pressure against the best competition.

I hope to post on this soon.

Thanks for the interest.

Regards

Tim

Wuornos
11-30-2007, 01:56 AM
Tim: Some fascinating results.

I see these ratings as complementing rather than replacing the old DOT ratings. The new ratings show how great a player was at a particular point in time, while the old ones showed how good the player was at sustaining his or her peak achievement over a period. My view is that we need BOTH in order to assess how great a player is.

Many thanks Steve

Many people are saying that this is what is required and I am working on it at the moment.

See my reply to AKV89 above.

Thanks for your interest and kind words.

Regards

Tim

FiveO
11-30-2007, 07:35 AM
1. Rod Laver 2781 Achieved year ending US Open 1969
2. Roger Federer 2768 Achieved year ending US Open 2007
3. Mats Wilander 2722 Achieved year ending US Open 1988
4. Pete Sampras 2715 Achieved year ending Wimbledon 1994
5. John McEnroe 2709 Achieved year ending US Open 1984
6. Andre Agassi 2708 Achieved year ending Australian Open 2000
7. Ivan Lendl 2707 Achieved year ending Australian Open 1988
8. Jim Courier 2687 Achieved year ending French Open 1992
9. Jimmy Connors 2683 Achieved year ending US Open 1974
10. Björn Borg 2682 Achieved year ending Wimbledon Open 1979

Ok, I know you’re all going to hate the position of Wilander. But the same criteria has been applied to all of the above players. i.e Dominance x Opposition x Tournament Status

IMO, No problem with Wilander scoring where he did. I doubt whether it was THE reason but subjectively Wilander's 3/4 Majors in one calendar year should trump Sampras's 3/4 being spread over two.

I would suspect however that if Sampras had won the '93 YEC v. Stich it may have tipped the balance in his favor?

What does surprise me is that Federer's 2006-2007 period reflected trumps his own calendar 2006 in that the time frame emcompasses much earlier exits from MS events at Indian Wells, Miami and Rome in 2007 v. 2006. But I would suspect that a pure calendar year 2006 would fall a negligible amount of points behind the period reflected.

Wuornos
11-30-2007, 04:01 PM
IMO, No problem with Wilander scoring where he did. I doubt whether it was THE reason but subjectively Wilander's 3/4 Majors in one calendar year should trump Sampras's 3/4 being spread over two.

I would suspect however that if Sampras had won the '93 YEC v. Stich it may have tipped the balance in his favor?

What does surprise me is that Federer's 2006-2007 period reflected trumps his own calendar 2006 in that the time frame emcompasses much earlier exits from MS events at Indian Wells, Miami and Rome in 2007 v. 2006. But I would suspect that a pure calendar year 2006 would fall a negligible amount of points behind the period reflected.

Great Post FiveO. You know your stuff don't you.

I think the difference here is that DOT adjusts rating points according to the strength of the event. If we consider the majors which tend to have most points we can see the following weighting for the men's majors of the past two years.

Australian Open 2006 Weighting = 70
French Open 2006 Weighting = 92
Wimbledon 2006 Weighting = 102
US Open 2006 Weighting = 101
Australian Open 2007 Weighting = 111
French Open 2007 Weighting = 102
Wimbledon 2007 Weighting = 110
US Open 2007 Weighting = 101

As I'm surre you are aware the DOT Ratings are adjusted for player strength (i.e domination) and this tends to be reflected in the calculation for Tournament Strength. This enables DOT to keep player ratings constant when a new dominant player enters the scene. Likewise if someone like Federer is dominating and other strong players come along and dominate the remaining field, this provides extra evidence as to Federer's strength and it is this effect we are seeing impact on Federer's 2007 rating. His dominance may be lower but the opposition adjuster is higher.

The last eight in the Australian Open of 2006 were Roger Federer, David Nalbandian, Marcos Baghdatis, Nikolay Davydenko, Sébastien Grosjean, Nicolas Kiefer, Ivan Ljubičić and Fabrice Santoro.

The same event in 2007 saw the last eight as Roger Federer, Rafael Nadal, Andy Roddick, Fernando González, Nikolay Davydenko, Tommy Haas, Tommy Robredo and Mardy Fish.

DOT is simply saying the evidence provided by winning the 2007 Australian Open was significantly better than that of winning the 2006 Open.

Had both tournaments had the same weighting then your point would have caused me huge concern and heartache in trying to work out what had gone wrong with the system.

It's great to see posts like this though, as it keeps me on my toes and provides fresh eyes to look at the output data and identify new problems.

I am still working on the request to provide DOT Rating data for single event, single year, five year and career. To do this I will need to release the output in the none standardised form. This means the exact amount of points scored by each player for each period will be as calculated and output series will have a geometric progression rather than an arithmetic progression.

Having said that things are looking good at the moment as I am able to weight each of the four categories mentioned above equally by calculating how far above or below the population mean each output is in terms of Standrad Deviation and simply adding these calculations together.

This means we get an overall rating which is able to compare players with long careers and lower levels of dominance, E.g. Chris Evert, with players of high dominance and shorter careers, like Seles, while keeping ouputs statistically sound.

More as a fun point it means we will be able to compare the careers of men with women players in the overall rating as the greater domination levels in the womens game are balanced by an increased level of standard deviation.

I should be in a position to release these data by the middle of next week.

All just good clean fun and thanks for your interest.

Take care

Regards

Tim

CyBorg
12-01-2007, 12:41 PM
For the very early periods of the open era it is necessary to use a standard statistical tool to provide a cautious estimate of the domination score had a full 4 years data been available. To do this we take the total weighting which is normally applied to players, i.e. 8*4events + 4*4evnts + 2*4 events +1*4events = 60. Where we have just three events (i.e after the US open of 1968 ), i.e. 8* 3 = 24, we apply a weighting of SQR(60)/SQR(24) on top of the basic weighting. Sounds complicated but it really isn’t.

Can someone explain this to me? I don't understand where the numbers are coming from. It mentions events but doesn't specifiy which events.

Maybe if I understand this part I will be able to comprehend why Wilander is third on this list. Even if you go by majors alone Wilander won three in his peak in 1988 but was hardly dominant in the surrounding years. I must be missing something.

FiveO
12-01-2007, 01:18 PM
Wuornos,

Are you calculating based on all, select or just GS events for the 12 month snapshots and lengthier calculations?

Just wondering.

5

AndrewD
12-01-2007, 02:16 PM
1. Martina Navratilova 2805 Year Ending 1984 US Open
2. Steffi Graf 2798 Year Ending 1989 Australian Open
2. Margaret Smith Court 2798 Year Ending 1971 Australian Open
4. Serena Williams 2789 Year Ending 2003 Australian Open
5. Monica Seles 2778 Year Ending 1993 Australian Open
6. Chris Evert 2759 Year Ending 1983 French Open
7. Martina Hingis 2740 Year Ending 1998 Australian Open
8. Justine Henin 2728 Year Ending 2004 Australian Open
9. Billie Jean King 2723 Year Ending 1972 US Open
10. Hana Mandlíková 2700 Year Ending 1981 Wimbledon

There is still very little to pick between the top three despite the improved levels of accuracy although the top player has changed from Margaret Court to Martina Navratilova. Margaret Court and Steffi Graff are now tied at 2nd equal.

Wuornos,

Why did you choose to end Navratilova's 84 season with the US Open when the Aus Open was played in December?

How did you get Navratilova's 84 season as rating higher than Court's or Graf's best year? In 69/70 Court won 7 of 8 majors plus the first of the next year. Same thing with Graf. In 88/89 she won 7 of 8 majors plus the first of the next year. Navratilova, on the other hand, won 6 of 8 majors in 83/84, less than the other two managed during calendar seasons.

Wuornos
12-02-2007, 01:19 AM
Wuornos,

Are you calculating based on all, select or just GS events for the 12 month snapshots and lengthier calculations?

Just wondering.

5

Hi FiveO

I used to calculate on all but received some quite severe criticism of my work because many believe that the very top players place more emphasis on the majors and treat the lesser events as having secondary importance. People were particularly critical of Sampras in relation to this and felt the methodology of including anything outside the majors was deflating some players scores.

Since then I have calculated majors only. To be honest the difference has been negligible as the weightings DOT was awarding to the minor events and the apparent randomisation within these events due to all top players being present simply flattened the disribution somewhat rather than made any major difference to placings.

Despite this I never switched back to the greater number of tournament methodology as my career in statistcs has taught me to produce models using the simplest methodologies to provide results that are fit for purpose. Therefore without significant differences, I continued on the major only route with a somewhat (!) reduced database and a massively reduced time in providing a system run.

Regards

Tim

PS Nearing completion now on the Single Event, Single Year, Five Year and Career Point count system.

Thanks for all your interest Five0. I do appreciate it.

Wuornos
12-02-2007, 01:23 AM
Can someone explain this to me? I don't understand where the numbers are coming from. It mentions events but doesn't specifiy which events.

Maybe if I understand this part I will be able to comprehend why Wilander is third on this list. Even if you go by majors alone Wilander won three in his peak in 1988 but was hardly dominant in the surrounding years. I must be missing something.

No Cyborg you're not missing anything. You're far to sharp for that. I know having read your previous posts.

The data above is based on a single year. This is something DOT has been criticesed for previously. I am currently putting together a system using simililar methodology, although the outputs will be on a different scale, which will evaluate players performances on various time frames. Single Event, Single Year, Five Year and Career.

I hope to be posting the initial outputs in relation to this later today and hopefully this will provide a better way of judging players.

Remember all periods will be adjusted not only for results biy also for standard of opposition.

Regards

Tim

Wuornos
12-02-2007, 01:50 AM
Wuornos,

Why did you choose to end Navratilova's 84 season with the US Open when the Aus Open was played in December?

How did you get Navratilova's 84 season as rating higher than Court's or Graf's best year? In 69/70 Court won 7 of 8 majors plus the first of the next year. Same thing with Graf. In 88/89 she won 7 of 8 majors plus the first of the next year. Navratilova, on the other hand, won 6 of 8 majors in 83/84, less than the other two managed during calendar seasons.

Hi Andrew

Big questions that i will try to answer in turn.

The above data are calculated off a single rolling year. Having said that in order for consistency to be maintained all Australian Opens are considered as being plaid in January. The question you are therefore asking in relation to why Navratilova's best year was selected as ending in with the US Open of 1984 has to do with the value of her result in the Australian Open of Decemeber 1984 compared with December 1983. This effect is two fold. First of all the Earlier year was weaker and in fact DOT rated this event with a 75 rating. While the later event received a 113 weighting. So results in the later event were actually worth 51% more. Having said that Martina won the earlier title but only made it to the semi final of the latter. Consequently her DOT points receved were 600 points for the ealier weaker event in winning it, but only 226 for the semi final of the stronger event. DOT therefore considers the performance of Martina in the year including the Australian Open of 1983 to the US Open of 1984, superior to French Open 2004 to Australian Open 2004. I hope this helps explain the basics of how the system works. With Future DOT Ratings I am removing the standardisation and simply showing these points at peak periods throughout a players career.

I think the second questio also has to inputs which effect results. First of all the above ratings are based off a single calendar year at the moment. Only when I post the revised figures later today will different lengths of period be shown. Secondly DOT is so much more than simple results. DOT Ratings are results x strength of opposition. DOT achieves this by being a double itterative process. You are right of course in saying that Court and Graf had exceptionally good years but where Martina really gains is in the fact that she had to achieve her results in the face of a direct challenge from Chris Evert. While I am not saying that Court or Graf couldn't have done this, I am saying they didn't and Navratilova did and therefore Navratilova has the best evidence and consequently has achieved the best single year rating.

I hope this explains a little how the system works.

Thanks for your interest and I would appreciate feedback on the new multi period system results that I will be posting later today.

Regards

Tim

AndrewD
12-02-2007, 04:13 AM
Tim,

It doesn't actually explain how the system works, sensibly. What you have described is, in essence, 'fiddling the system'. If it is to have any credibility then it can only count results from a calendar year.

Graf and Court were good enough to win 4 majors in the one year. Not a 'rolling' year but an actual calendar year. That sort of achievement carries far more weight than winning 4 events in succession but not within the same year due to the significantly greater prestige and infinitely greater pressure. A valid system must take that into account when apportioning ratings to the events.

As to the greater difficulty Navratilova faced due to the presence of Evert, I would say you've got that completely wrong. Between the end of 1982 and the beginning of 1985 Navratilova beat Evert 13 times in a row with only four matches going to 3 sets. Navratilova dominated Evert from 1980 through until Evert's retirement in 1988 winning 32 of the 44 matches they played. In the time frame you're looking at, 83/84, Navratilova had far more trouble with Mandliova, losing one match to her and having four others go to three sets. Evert, on the other hand, managed to win two sets. In any language that is not what you call a 'direct challenge'.

In winning her Grand Slam of 1970, Margaret Court beat Goolagong, King, Casals, Richey in fields that included Durr, Melville, Wade and Jones - all of them Hall of Famers and all of them, bar Casals, winners of majors.

I'm not as familiar with Graf's opponents but, suffice to say, she won the Grand Slam and that easily trumps 3 out of 4.

At the end of the day, Court and Graf achieved the pinnacle in our sport, the Grand Slam. Any system which doesn't rate their accomplishment (and their year/s) higher than that of another player who, with equal opportunity, wasn't capable of doing the same thing is seriously flawed.

Wuornos
12-02-2007, 08:18 AM
Tim,

It doesn't actually explain how the system works, sensibly. What you have described is, in essence, 'fiddling the system'. If it is to have any credibility then it can only count results from a calendar year.

Graf and Court were good enough to win 4 majors in the one year. Not a 'rolling' year but an actual calendar year. That sort of achievement carries far more weight than winning 4 events in succession but not within the same year due to the significantly greater prestige and infinitely greater pressure. A valid system must take that into account when apportioning ratings to the events.

As to the greater difficulty Navratilova faced due to the presence of Evert, I would say you've got that completely wrong. Between the end of 1982 and the beginning of 1985 Navratilova beat Evert 13 times in a row with only four matches going to 3 sets. Navratilova dominated Evert from 1980 through until Evert's retirement in 1988 winning 32 of the 44 matches they played. In the time frame you're looking at, 83/84, Navratilova had far more trouble with Mandliova, losing one match to her and having four others go to three sets. Evert, on the other hand, managed to win two sets. In any language that is not what you call a 'direct challenge'.

In winning her Grand Slam of 1970, Margaret Court beat Goolagong, King, Casals, Richey in fields that included Durr, Melville, Wade and Jones - all of them Hall of Famers and all of them, bar Casals, winners of majors.

I'm not as familiar with Graf's opponents but, suffice to say, she won the Grand Slam and that easily trumps 3 out of 4.

At the end of the day, Court and Graf achieved the pinnacle in our sport, the Grand Slam. Any system which doesn't rate their accomplishment (and their year/s) higher than that of another player who, with equal opportunity, wasn't capable of doing the same thing is seriously flawed.

Thanks Andrew that's some great feedback. I try not to 'fiddle the system' and try to rely on reaonable analysis of the data rather than personal views. The one thing I do try to keep out of it is the human judgement, interference or opinion. I like the system to produce the outputs directly from the data without artificially included weightings.

In this respect the calendar year is a funny thing. With years like 1977 and 1986 having differing numbers of majors etc. I felt that to do justice to players the best thing was to have a rolling horizon of four majors, but this is something which would not reward the calendar year Grand Slam more than any other period in which a player holds all four majors. To achieve this I needed to introduce human opinion and break one of major rules in creating the system. I was not prepared to do this. Part of my problem with this type of thing is that the Grand Slam was not always held in such high esteem. Had this been the case players like Borg and McEnroe might have competed rather than largely eschewing competing within that major. this is where we hit the stumbling block. Not only would we need to include additional weighting to the Calendar year slam in direct proportion to the amount of pressure that brings but it would also need to alter over time. No this compromise of what was essentially a purely analytical system was to much. I felt what would be lost in independence would be far greater than the essentially artificially based weighting brough into the system.

Moving on to your point regarding Evert and Navratilova. I'm not great one for head to head counts meaning anything significant. Firstly because many of the matches are not played when poth players are at their peak, although this is not the problem with the data you have quoted. Secondly that to simplify matters there is a certain element of Rock,Paper Scissors with Tennis. i.e. Player A may have the upper hand on player B, Player B on C and Player C on A. I believe you need to look at the population as a whole and domination of the rest of the population. Essentially what I am saying is not that Evert had a fantastic record specifically against Navratilova, but she was clearly more dominant of the rest of the population than players like Evonne Goolagong. Chris Evert won 9 Majors between 1980 and her retirement, I believe this makes her a more formidable competitor and Navratilova receives credit for winning consistently against that type of quality. The adjutment in relation to quality of oppostion faced is purely mathematical and not 'fiddling' if this what you meant, I am unsure. It achieves this through a what can be referred to as a double iterative process.

Having said all this, largely in defence of the system I use, I must now say that I agree with you in putting Court and Graf above Navratilova, although for different reasons. Two moths ago I had completed a system which concluded just this which used a time frame of four years with greater weight being given to the later results. Since then I have been pursuaded by others to change the sytem and I believe now these chamges have done nothing to improve the output only to compromise it.

I am therefore reverting DOT back to the system of two months ago and will make no further changes to its methodology.

You are the second person to post today that has convinced me that my older alogorithms were better. So thanks for the input. It was great.

Just for the record. Here are the top 20 women of the open era as calculated by the old DOT.

1 Margaret Smith Court 2880.29
2 Steffi Graf 2880.23
3 Martina Navratilova 2878
4 Monica Seles 2837
5 Chris Evert 2823
6 Serena Williams 2797
7 Billie Jean King 2781
8 Arantxa Sánchez Vicario 2769
9 Martina Hingis 2762
10 Evonne Goolagong 2752
11 Venus Williams 2750
12 Justine Henin 2749
13 Hana Mandlíková 2747
14 Lindsay Davenport 2731
15 Gabriela Sabatini 2721
16 Jennifer Capriati 2713
17 Ann Jones 2705
18 Amélie Mauresmo 2701
19 Maria Sharapova 2696
20 Kim Clijsters 2689

Thanks again for the feedback

Regards

Tim

AndrewD
12-02-2007, 10:27 AM
Tim,

The Grand Slam
1) The Grand Slam has been held in the highest regard since the term was coined. Anyone suggesting otherwise either knows nothing about tennis' history or is trying to make some excuse for the failure of their favourite player to win the Grand Slam.

2) Since the introduction of 'Open' tennis, there have only been four times a Grand Slam was possible on the women's side. The only time a player decided not to attempt it was in 1972 when Billie Jean King was obliged to stay in the States and play the fledgling Virginia Slims circuit. She considers it one of her greatest regrets but a necessity if women's pro tennis was to succeed. On every other occasion the woman (Court, Navratilova and Graf) actively chased the Grand Slam.

3) On the men's side, since the introduction of 'Open' tennis, the only time the Grand Slam has been a possibility was in 1969. The best any other player has done is Borg's winning the French and Wimbledon back-to-back.

4) With the Grand Slam not being a possibility on the men's side since 69 there is no sensible way you can claim it was not always held in high esteem. That no-one chased it was merely due to no-one being able to win it, apart from Borg who failed at the US. On the women's side it is very obvious that it was held in high esteem. If it hadn't been then a) none of the women I mentioned would have chased it and b) there wouldn't have been a substantial prize offered to Navratilova if she'd been able to accomplish it.

5) Borg and McEnroe WERE trying to achieve it. Unfortunately for them, Borg continually stumbled at the US Open so there was no reason to play Australia and try for the GS. McEnroe couldn't win the French so the GS wasn't attainable. Making the suggestion that the Grand Slam didn't matter to those players is completely without merit - especially for someone like McEnroe who revered Rod Laver and all he had achieved.

Other

1) If you aren't going by calendar year then you are 'fiddling the system'.

2) Human interference is integral to your system and your 'outputs' are not derived organically from the data, for the simple reason that YOU choose not to follow a calendar year.

3) Open tennis has been played for long enough to discount any discrepancies such as 1977 and 1986. Neither one of those years (or the year prior to or after it) is going to be significantly impacted by the absence of a major or the presence of an extra one. Using that as the argument for a non-calendar year carries no weight at all.

4) You said that " Martina really gains is in the fact that she had to achieve her results in the face of a direct challenge from Chris Evert." Then you say "I'm not great one for head to head counts meaning anything significant". You can't have it both ways. If you say that Evert posed a direct challenge to Navratilova then you can't ignore the fact that, during the 83/84 seasons, she posed absolutely no threat to her.

5) I will stress my point again. While Evert might have been, over the course of her entire career, more successful in singles than Goolagong, King, Casals, Richey, Durr, Melville, Wade and Jones she was not a factor during 83/84 and, apart from Mandlikova, she was the only player of genuine class in any tournament Navratilova played. Resultantly, Court had stronger opposition. Graf, you could argue, had slightly stronger opposition than Navratilova because she had to face Navratilova herself at Wimbledon.

6) Either way, you can't give bonus points based on the reputation of a player, you can only give it on their performances. Or, you can discount it altogether. If you consider reputation (which, against Navratilova, is all Evert had at the time) then you'd have to say the 2004 US Open and Wimbledon had an incredibly strong field due to the presence of Navratilova. Doesn't matter that she wasn't going to win (she was 48 at the time) because you're basing your ratings on reputation, not current results.

Wuornos
12-03-2007, 05:02 AM
Tim,

The Grand Slam
1) The Grand Slam has been held in the highest regard since the term was coined. Anyone suggesting otherwise either knows nothing about tennis' history or is trying to make some excuse for the failure of their favourite player to win the Grand Slam.

2) Since the introduction of 'Open' tennis, there have only been four times a Grand Slam was possible on the women's side. The only time a player decided not to attempt it was in 1972 when Billie Jean King was obliged to stay in the States and play the fledgling Virginia Slims circuit. She considers it one of her greatest regrets but a necessity if women's pro tennis was to succeed. On every other occasion the woman (Court, Navratilova and Graf) actively chased the Grand Slam.

3) On the men's side, since the introduction of 'Open' tennis, the only time the Grand Slam has been a possibility was in 1969. The best any other player has done is Borg's winning the French and Wimbledon back-to-back.

4) With the Grand Slam not being a possibility on the men's side since 69 there is no sensible way you can claim it was not always held in high esteem. That no-one chased it was merely due to no-one being able to win it, apart from Borg who failed at the US. On the women's side it is very obvious that it was held in high esteem. If it hadn't been then a) none of the women I mentioned would have chased it and b) there wouldn't have been a substantial prize offered to Navratilova if she'd been able to accomplish it.

5) Borg and McEnroe WERE trying to achieve it. Unfortunately for them, Borg continually stumbled at the US Open so there was no reason to play Australia and try for the GS. McEnroe couldn't win the French so the GS wasn't attainable. Making the suggestion that the Grand Slam didn't matter to those players is completely without merit - especially for someone like McEnroe who revered Rod Laver and all he had achieved.

Other

1) If you aren't going by calendar year then you are 'fiddling the system'.

2) Human interference is integral to your system and your 'outputs' are not derived organically from the data, for the simple reason that YOU choose not to follow a calendar year.

3) Open tennis has been played for long enough to discount any discrepancies such as 1977 and 1986. Neither one of those years (or the year prior to or after it) is going to be significantly impacted by the absence of a major or the presence of an extra one. Using that as the argument for a non-calendar year carries no weight at all.

4) You said that " Martina really gains is in the fact that she had to achieve her results in the face of a direct challenge from Chris Evert." Then you say "I'm not great one for head to head counts meaning anything significant". You can't have it both ways. If you say that Evert posed a direct challenge to Navratilova then you can't ignore the fact that, during the 83/84 seasons, she posed absolutely no threat to her.

5) I will stress my point again. While Evert might have been, over the course of her entire career, more successful in singles than Goolagong, King, Casals, Richey, Durr, Melville, Wade and Jones she was not a factor during 83/84 and, apart from Mandlikova, she was the only player of genuine class in any tournament Navratilova played. Resultantly, Court had stronger opposition. Graf, you could argue, had slightly stronger opposition than Navratilova because she had to face Navratilova herself at Wimbledon.

6) Either way, you can't give bonus points based on the reputation of a player, you can only give it on their performances. Or, you can discount it altogether. If you consider reputation (which, against Navratilova, is all Evert had at the time) then you'd have to say the 2004 US Open and Wimbledon had an incredibly strong field due to the presence of Navratilova. Doesn't matter that she wasn't going to win (she was 48 at the time) because you're basing your ratings on reputation, not current results.

Oh dear. I'm sorry you feel like that Andrew because I really do value your opinion and it is a disappoinment to me that you feel like this.

To answer each point in turn

1. Yes the Grand Slam has been recognised for as long as I can remember. It was always valued but whether the value was always at the level it is now, is another matter. I feel if the value of the Grand Slam despite it being recognised, for a very long time, was reduced in some previous decades. I cannot believe that it held the same cvalue when many players in the late 1970s didn't even bother playing the Australian Open. In any even, had it been achieved during this period it's value must have been less as the quality of opposition at the AO was so much lower. What do other people think?

2. Yes but before the move in month players like Evert weren't competing for the AO therefore excluding the possibility of chasing a GS. I ssume you are defining chasing as having the first three majors. Many players were not competing in the AO even before the change of month. Therefore GS became less of an indicator of greatness during these periods, therefoe the GS was not always held in in the same regard.

3. Yes but with people like Borg not playing the Australian at the start of the year, I'm not sure what this proves.


4. We'll have to disagree on that. I think I have proved it above.

5. Sorry I though McEnroe said in his Autobiography that he regretted not playing the AO. Not that he didn't feel it necessary because of the GS. Ithink you're making the GS the reason people didn't play to justify your argument when really there were other reasons.

other

1. I feel by introducing the artificial horizon of a calendar year I would be fiddling the sytem.

2. I disagree. The outputs are produced without human opinion or adjustment. To indroduce the calendar year as an artificial time horizon would be doing just that.

3. How can you say that the calendar year GS is the most significant measure of talent and then argue that the number of majors in a year is an irrelevance.

4. OK have it your way, Evert was a nobody in the 83 and 84 seasons and Martina should have no credit for beating her and therefore Martina's rating should be deflated from its current level. You better tell that to all the people Chris beat in winning two majors and a further four major finals not bad out of eight majors eh! That means Evert was a good player during this period. Then Navratilova beat her. Yes. Does that mean Chrissie was weak or that Navratilova was very strong? If the latter then that is what DOT is saying.

5. Over the course of anybodies career is an irelevance. DOT operates of a limited time frame wit Dominance x Opposition x Tournament. Chrisie scored higher using these statistical criteria than did Goolagong, King, Casals, Richey, Durr, Melville, Wade and Jones. Navratilova's score was adjusted accordingly to the wins she had against Evert. Nothing more, nothing less, no bias and no opinion. What do other people think. Was Chrissie's play in 1983 and 1984 poorer than Goolagong, King, Casals, Richey, Durr, Melville, Wade and Jones, or was Navratilova at the absolute peak of her game.

6. I don't give bonus points on reputation of a player I give bonus points on the quality of a player faced, i.e opposition. Chris was clearly dominant of the rest of the population in 1983 and 1984 with the exception of Navratilova. This is calculated independantly of any opinion. The fact that Martina then dominated her results in Martina's domination adjuster being raised. I find this argument far more persuasive than Chris dominated everyone else during this time because the whole world had become weaker than during the Court Era and Chris alone was the sole player at the old level.

I agree with you there are problems with the above ratings but not with those issues you have selected. As I have stated in anther thread I am reverting to the one methodology used two months ago and I know from your posts there that you find these measures more accepatble to your views. Strangely enough though they use all the same criteria as above. All that has changed is the time frame over which DOT is calculated.

Take care and keep posting

Regards

Tim

noeledmonds
12-03-2007, 08:17 AM
1. Rod Laver 2781 Achieved year ending US Open 1969
2. Roger Federer 2768 Achieved year ending US Open 2007
3. Mats Wilander 2722 Achieved year ending US Open 1988
4. Pete Sampras 2715 Achieved year ending Wimbledon 1994
5. John McEnroe 2709 Achieved year ending US Open 1984
6. Andre Agassi 2708 Achieved year ending Australian Open 2000
7. Ivan Lendl 2707 Achieved year ending Australian Open 1988
8. Jim Courier 2687 Achieved year ending French Open 1992
9. Jimmy Connors 2683 Achieved year ending US Open 1974
10. Björn Borg 2682 Achieved year ending Wimbledon Open 1979

My subjective opnion does agree more with these DOT ratings than the previous ones. It is very difficult to decide which rating is "better" as any comparison needs a 3rd ranking which is either subjective or simply another statistical anlaysis of sorts.

In terms of non-calander year dominance in the open-era I think that Laver followed by Federer is an almost definitive choice.

Wilander will always perform better in statistics than the public rate him. This is due to the far less tangible factors involved with Wilander. Although Wilander has both the peak performance and the achivements of a great player he did not have the personality or game style to leave as strong an influence in tennis as other players with similar achivements in tennis such as McEnroe. Whether or not you consider these factors important when analysing the greatness of a player is largely subjective. However on a statistical basis of this kind I agree that Wilander should be around the position he is.

Courier over Connors is a suprise to me. Connors's 1974 was one of the greatest single calender years of the open-era and it suprises me that he falls as low as he does. Perhaps Connors is heavily penalised for not entering the French Open but statistics will not incoperate the politics surrounding this.

Borg's position is low but not totally unreasonable when we consider how his career panned out. Borg is a difficult one in my opnion because the DOT ratings deliberitly ignore longitivity and concetrate on peak performance. This easily explains how Lendl and Connors would fall down such a list but Borg is different. Borg does not perform well in terms of longitivity or peak performance over a calander year. However Borg does perform well by achivement criteria and normally features in the top 5 male open-era players of people's subjective lists.

Becker falling out of the top 10 seems reasonable to me, purely subjectively. Rosewall should not feature in the top 10 of an open-era criteria based analysis. Although Rosewall should undoutably finish in the top 10 of an anlysis that incoperated pre-open era achivemnts also.

Ultimately it is difficult to say how good this rankings is. If you are trying to corelate with public opnion then I believe you have succeeded pretty well here. However public opnion is always skewed towards more recent players. There is a statistical anlaysis that adjusts public opnion based on the recentness of a players career. I cannot recall how this is calulated but perhaps this would be more reflective of the validity of a ranking system rather than public opnion in isolation.

Alternatively "elite opinion" is a good measure of the success of a ranking system. Elite opinion relates to expert or informed opnion. However even expert opnion is often flawed, for example many experts were previously players who have a bias or vested interest towards players they competed against in their own generation over other players.

Very interesting analysis though. Please to keep displaying the results of your analyses as the results do interest me.

CyBorg
12-03-2007, 10:24 AM
The data above is based on a single year. This is something DOT has been criticesed for previously. I am currently putting together a system using simililar methodology, although the outputs will be on a different scale, which will evaluate players performances on various time frames. Single Event, Single Year, Five Year and Career.

I hope to be posting the initial outputs in relation to this later today and hopefully this will provide a better way of judging players.

Remember all periods will be adjusted not only for results biy also for standard of opposition.

Thanks for clarifying, Tim.

How do you measure the quality of an event like the French Open?

Now I know that the atp ranking cannot be representative, because looking at history we know that the 10th rated player in the world can be the top clay courter in the world, while the first rated player in the world may not be in the top 20 on clay. Similarly any formula which calculates a player's value based on condensed year-round results may provide a sort of a dilluted median value.

I find that the French gets the short end of the stick at times when people rate the quality of its draws. This is mostly because some tend to undervalue the participants. I think that the French should do like Wimbledon and rate the seeds itself, rather than depend on the skewed atp rankings.

I do sympathize with the decision to grant majors a value based on the standard of opposition, which allows us to escape from particular myths regarding 'prestige', which are strictly manufactured.

CyBorg
12-03-2007, 10:27 AM
Borg does not perform well in terms of longitivity or peak performance over a calander year.

Why would this be true for the latter? I find this hard to believe because we know that Borg had dominant years in 78, 79 and 80. Just looking at the results and won-loss records it seems strange to suggest that his peak isn't what its cracked up to be. Perhaps his opposition on clay is undervalued? There's also the case that Borg played a bit of a reduced schedule in 1980, skipping both Dallas and Philadelphia. However he was still in Monte Carlo, Vegas and Stockholm...

CyBorg
12-03-2007, 10:31 AM
If you aren't going by calendar year then you are 'fiddling the system'.

Why is this? The calendar seems arbitrary to me. A year is a year, regardless of whether we start counting in January or in June.

CyBorg
12-03-2007, 10:40 AM
Yes the Grand Slam has been recognised for as long as I can remember. It was always valued but whether the value was always at the level it is now, is another matter. I feel if the value of the Grand Slam despite it being recognised, for a very long time, was reduced in some previous decades. I cannot believe that it held the same cvalue when many players in the late 1970s didn't even bother playing the Australian Open. In any even, had it been achieved during this period it's value must have been less as the quality of opposition at the AO was so much lower. What do other people think?

I agree. If we go back to the early-to-mid 70s in particular we're looking at a large number of majors that were not top-four events. French Open 70-71 for example; Wimbledon 72-73; Aussie for much of the 70s. And until the mid-80s the Aussie was too much of an afterthought to merit as much attention as the other four. I'm sure that Borg would have played there in 1980 had he won at the US Open, but to entertain the notion that Borg's fortunes would be riding on a minor event like that would be silly. There really was no true grand slam from the early-70s to the mid-80s. It was not as commercial a notion.

Yes but before the move in month players like Evert weren't competing for the AO therefore excluding the possibility of chasing a GS. I ssume you are defining chasing as having the first three majors. Many players were not competing in the AO even before the change of month. Therefore GS became less of an indicator of greatness during these periods, therefoe the GS was not always held in in the same regard.

The word 'chasing' is a very good one here. Today players are truly chasing the grand slam, perhaps buying into what the media says - but the media itself is feeding upon the newfound organization that came upon the sport roughly around 1990 with the introduction of the masters series events. Because of this there are no real events to challenge the majors - the atp makes sure that no event has as many ranking points as a major, as great a field as a major nor offers up as much money as the major. Going back 20 years and beyond, events like Boca West, Rome and Philadelphia often challenged majors and frequently had better draws. It was not uncommon for someone like Jimmy Connors to avoid playing in a particular major (in his case the French) out of knowledge that he could pick up as much money and ranking points elsewhere. No one really was chasing the grand slam in his time. It's not something you'd hear in the media.
[/QUOTE]

Moose Malloy
12-03-2007, 11:07 AM
It baffles me why so many continue to criticize Tim as being 'biased' with his systems(some of you don't really think he knows beforehand what his statistics will show?)

I don't necessarily agree with his conclusions, but I appreciate the effort.
At least he's trying, some of you are too harsh with criticism, considering you aren't putting any work in to this subject & he is.

Wuornos
12-04-2007, 04:30 AM
It baffles me why so many continue to criticize Tim as being 'biased' with his systems(some of you don't really think he knows beforehand what his statistics will show?)

I don't necessarily agree with his conclusions, but I appreciate the effort.
At least he's trying, some of you are too harsh with criticism, considering you aren't putting any work in to this subject & he is.

Many thanks for the kind words, Moose. They are much appreciated.

Regards

Tim

protourOS
12-04-2007, 07:09 AM
I agree- I think AndrewD is a little off the mark here

Tim your statistical analysis always gives new perspective to any tennis history discussion.
If possible, please keep up the good work!

I for one find the revised DOT methodology an intriguing analysis of the strongest year period of domination by players - and to expand it by taking in account single, 5 and say 10 year periods will surely provide some noteworthy results.

Wuornos
12-04-2007, 07:57 AM
I agree- I think AndrewD is a little off the mark here

Tim your statistical analysis always gives new perspective to any tennis history discussion.
If possible, please keep up the good work!

I for one find the revised DOT methodology an intriguing analysis of the strongest year period of domination by players - and to expand it by taking in account single, 5 and say 10 year periods will surely provide some noteworthy results.

Thanks for that.

It's great to know people enjoy what I do.

Cheers

Tim

Nickognito
12-04-2007, 05:06 PM
I'm not here to criticize Wuornos' work (i'm in this forum only thanks to his work) but I have to say something.

His ranking does not pretend, I think, to describe an order of players in the open era, but only an order of dominations.

In this point of view, is a method extraordinarily well made. But maybe who criticize him would expect some more interesting and relevant kind of ranking, that can tell us who are the best players in open era.

To know that, we have to make a subjective decision: Domination is an important factor in determining the best players in open era? Numbers do not help us in this only subjective decision and prove nothing.

In my opinion, domination is only one (important one, sure) of many factors that determines all times rankings.

If you have to decide between these two careers:

1)you win nothing for 5 years than you win 3Australian Open, 3 French Open and 2 Us Open in 5 years, than you win nothing for other 5 years.

2)you win 3 Australian open, 3 French Open, 3 Wimbledon and 3 Us Open in 15 years. You are never able to win more than a slam in a single year.

Who has the best results? In my opinion:

- if I don't know anything more, 2) has the best career
- if 1) dominated just because in that period he had no opponents, then 2) has by far the best career.
- If 2) won a slam every year only because he was able to win the weakest tournament of every year, but he's not able to win a great slam tournament, then 1) ha the best career.

So, in other words, domination itself means nothing without other considerations.

Maybe Wuornos' work doesnt care about it. Maybe he doesn't care about who is the best player in open era, but only about who dominated more.

But I wish that a so well-made ranking could someday help us about the 'open.era ranking' question.

So, maybe some people are criticizing him for this reason, in order to know more.

;-)

Regards,

c.

CyBorg
12-04-2007, 08:58 PM
So, in other words, domination itself means nothing without other considerations.

If you bothered reading the thread you would know that there are other considerations, such as the standard of opposition.

Nickognito
12-04-2007, 09:47 PM
I bothered reading this thread, of course, cause it's very interesting.

This rating is all about dominance. Since it's a very well made rating it tell us that dominance dependes on other factors, like the strength of opponents and tournaments played.

But definitely in this rating, the player who has all his wins in few years has a great advantage on the player who has greater career wins but with many pauses between them.

I thinks this rating's invetor believe that dominance is the best indicator of players strength (but maybe i'm wrong) and for sure he has the right to believe it, of course, but he and you cannot argue with the fact that this claim is subjective.

In any case, i don't thing i'm misinterpretating this rating or underrating it. If i do, please Wuornos let me know.

Wuornos
12-05-2007, 04:43 AM
I bothered reading this thread, of course, cause it's very interesting.

This rating is all about dominance. Since it's a very well made rating it tell us that dominance dependes on other factors, like the strength of opponents and tournaments played.

But definitely in this rating, the player who has all his wins in few years has a great advantage on the player who has greater career wins but with many pauses between them.

I thinks this rating's invetor believe that dominance is the best indicator of players strength (but maybe i'm wrong) and for sure he has the right to believe it, of course, but he and you cannot argue with the fact that this claim is subjective.

In any case, i don't thing i'm misinterpretating this rating or underrating it. If i do, please Wuornos let me know.

Yes Nick in a way you are correct. The way DOT measures playing strength is via Dominance x Opposition Strength x Tournament Status. This provides an overall playing performance at a single point in time. This in itself forms the DOT Ratings and I believe this gives a pretty good figure. I do not accept tha DAT Ratings are opinion based in this form.

How you then go on to to interpret these playing standards is up to you. I tend to quote Peak DOT as i believe the peak playing standard is the best indicator of greatness and yes you're right in saying my selection of this as an an evaluation of player greatness is a personal one and is subjective.

However, I suspect you might choose mean rating over a 10 year period, 5 year period etc or whichever is the highest. This of course is an equally valid view but can still be achieved using DOT. I therefore believe DOT in itself is a 100% fair system. It is the interpretation of it that can introduce bias. I am reminded of a book I read as astudent called 'The Use and Abuse of Statistics' DOT would be desribed as an added value statistic in that takes basic data and converts it into another data form that may be more useful to the layman. However, it is what the layman then decides to do with such a data series that becomes the abuse of the statistic concerned.

The use of Domination in the initial calculation really means how much did the player achieve in the minimum time constraints necessary to ensure we have an accurate reflection of performance. Typically this is four years with greater weighting given to the more recent year. Of course a DOT Rating for a particular point in time has to be derived off a limited time period, but I think when you discuss domination, you aren't really meaning this primary calculation but my selection of PEAK DOT in showing when a player was most dominant adjusted for opposition and tournaments status.

I am currently putting together an article which shows the relationships between Peak DOT Ratings, DOT Rankings and the official Rankings. I'm quite pleased with the results being output within this analsis as it is showing how some players have been able to achieve great success with a limited peak. E.g Jimmy Connors was according to DOT probably the best player in the world for a none consectutive period of four years. This is not so far removed from the five years he held the official #1 status. DOT disagrees with some of the 1970s period in which Connors was rated # 1 above Borg, but that's beside the point. The point is that DOT is happy to concede that Connors should have reigned for a long time as #1 despite the fact his DOT Rating never once exceeded the 2706 point mark. It is precisely this kind of inconsistency in the value of results and duration of reign that DOT has been designed to identify and allocate an appropriate rating to.

However, I do not accept that DOT will automatically penalise a player who has his particular succeses spread over a long era, just because domination is used in the calculation to derive a current playing standard. Of course this is an issue with players who have wins spread over a long period. The question is why. If it's because of a drop off in form then DOT will identify their peaks when they achieved their successes and they will get the credit they deserve for a high standard when they were achieving their results. If however they are playing to their standard for a long time and require a drop in form of the top player etc then that in itself, in my opinion is an indicator of a lower quality of player. Even if there is no drop in standard and they can only manage half a major per year, I still see this as being a poorer level of achievement than the player who can score the same number of titles at three majors a year. When looking at the open era Laver jumps to mind due to his short career during this period. Many players have scored more majors but none at the same level of play.

In short then I am saying that DOT wouldn't necessarily discriminate against the player who scores majors over a longer period. It would depend on whether he was peaking at the time he was scoring his major wins and had a reduced standard in between, or whether the long period he took to collect wins was because of a reduced level of opposition. DOT just isn't that simple in its evaluations.

Your example at #30 is a very good one and it is exactly these issues that DOT Evaluates. DOT would come to exactly the same conclusions that you have posted at 1. and 2. If these are genuinely your views then DOT should be an ideal tool in providing an actual numerical evaluation of your views.

Remember DOT was originally developed as a project attempting to design a statistical model designed to encapsulate the experienced fans and official pundits thought processes into a statistical system that then provides a numerical evaluation. It would be failing if it did not meet the points you were making at post #30 as the very minimum of its output.

Clear as mud eh!

Sorry if it's a bit rushed but I'm trying to reply during my lunchtime at work.

Watch out for my post on the relationship between DOT and the official rankings tomorrow.

Regards

Tim

Wuornos
12-05-2007, 04:55 AM
I bothered reading this thread, of course, cause it's very interesting.

This rating is all about dominance. Since it's a very well made rating it tell us that dominance dependes on other factors, like the strength of opponents and tournaments played.

But definitely in this rating, the player who has all his wins in few years has a great advantage on the player who has greater career wins but with many pauses between them.

I thinks this rating's invetor believe that dominance is the best indicator of players strength (but maybe i'm wrong) and for sure he has the right to believe it, of course, but he and you cannot argue with the fact that this claim is subjective.

In any case, i don't thing i'm misinterpretating this rating or underrating it. If i do, please Wuornos let me know.

Yes Nick in a way you are correct. The way DOT measures playing strength is via Dominance x Opposition Strength x Tournament Status. This provides an overall playing performance at a single point in time. This in iteself forms the DOT Ratings and I believe this gives a pretty good figure.

How you then go on to to interpret these playing standards is up to you. I tend to quote Peak DOT as i believe the peak playing standard is the best indicator of greatness and yes you're right in saying my selection of this is my personal preference.

However, I suspect you might choose mean rating over a 10 year period, 5 year period etc or whichever is the highest. This of course is an equally valid view but can still be achieved using DOT. I therefore believe DOT in itself is 100% fair system. It is the interpretation of it that can introduce bias. I am reminded of a book I read as astudent called 'The Use and Abuse of Statistics' DOT would be desribed as an added value statistic in that takes basic data and converts it into another data form that may be more useful to the layman. However, it is hwat the laymean then decides to do with such data series that becomes the abose of the statistic concerned.

The use of Domination in the initial calculation really means how much did the player achieve in the minimum time constraints necessary to ensure we have an accurate reflection of performance. Typically this is four years with greater weighting given to the more recent year.

I am currently putting together an article which shows the relationships between Peak DOT Ratings, Dot Rankings and the official Rankings. I'm quite pleased with the results being output within this analsis as it is showing how some players have been able to achieve great success with a limited peak. E.g Jimmy Connors was according to DOT probably the best player in the world for a none consectutive period of four years. This is not so far removed from the five years he held the official #1 status. DOT disagrees with some of the 1970s period in which Connors was rated # 1 above Borg, but that's beside the point. The point is that DOT is happy to concede that Connors should have reigned for a long time as #1 despite the fact his DOT Rating never once exceeded the 2706 point mark. It is precisely this kind of inconsistency in the value of results and duration of reign that DOT has been designed to identify and allocate an appropriate rating to.

However, I do not accept that DOT will automatically penalise a player who has his particular succeses spread over a long era, just because domination is used in the calculation to derive a current playing standard. Of course this is an issue with players who have wins spread over a long period. The question is why. If it's because of a drop off in form then DOT will identify their peaks when they achieved their successes and they will get the credit they deserve for a high standard. If however they are playing to their standard for a long time and require a drop in form of the top player etc then that in itself, in my opinion is an indicator of a lower quality of player. Even if there is no drop in standard and they can only manage half a major per year, I still see this as being a poorer level of achievement than the player who can score the same number of title in a shorter time frame. When looking at the open era Laver jumps to mind due to his short career during this period. many players have scored more majors but none at the same level of play.

In short then I am saying that DOT wouldn't necessarily discriminate against the player who scores majors over a longer period. It would depend on whether he was peaking at the time he was scoring his major wins and had a reduced standard in between or not, or whether the long period he took to collect wins was because of a reduced level of opposition. DOT just isn't that simple.

Clear as mud eh!

Sorry if it's a bit rushed but I'm trying to reply during my lunchtime at work.

Watch out for my post on the relationship between DOT and the official rankings tomorrow.

Regards

Tim

Nickognito
12-05-2007, 06:46 AM
Thanks Wuornos.
I'm only repeating myself and say that in my opinion DOT rankings do not rank in the right order players careers, because they give domimance too much importance. And I would like to 'abuse' your ranking :). I.e. I think that your ranking is so well made that iit's wishable to be able to integrate it with other considerations in order to make a better overall ranking. Maybe it's not up to you, but i think it's possible to improve your method. I don't know how, at the moment, i'm sorry. And obviously others can 'abuse' your ranking saying that it's already perfect in describing an open era ranking ;-)

Regards

c.

noeledmonds
12-05-2007, 10:16 AM
Why would this be true for the latter? I find this hard to believe because we know that Borg had dominant years in 78, 79 and 80. Just looking at the results and won-loss records it seems strange to suggest that his peak isn't what its cracked up to be. Perhaps his opposition on clay is undervalued? There's also the case that Borg played a bit of a reduced schedule in 1980, skipping both Dallas and Philadelphia. However he was still in Monte Carlo, Vegas and Stockholm...

I use "well" as a relative term. Obviously Borg did do very well in these years compared to what most players have achieved in their career. However Borg's best year only reaches number 10 on Tim's ranking list, which relative to Borg's reputation and generally held status is not very high at all.

CyBorg
12-05-2007, 04:43 PM
I use "well" as a relative term. Obviously Borg did do very well in these years compared to what most players have achieved in their career. However Borg's best year only reaches number 10 on Tim's ranking list, which relative to Borg's reputation and generally held status is not very high at all.

Yes, I see that and I am trying to figure out why. I think that part of the problem is that underrates Borg's clay opponents.

Nickognito
12-05-2007, 05:07 PM
or maybe you tend to overrate players of '70s :)

CyBorg
12-05-2007, 11:27 PM
or maybe you tend to overrate players of '70s :)

Or maybe you just don't watch the old matches, which is pretty much obvious.

Nickognito
12-06-2007, 12:19 AM
so maybe wuornos' numbers didn't watch old matches :)

'70s matches aren't old. Maybe you didn't watch old Hoad matches...

But you can keep trying to figure out why.

c.

chaognosis
12-06-2007, 12:37 AM
It's been a while since I've followed up on this stuff... quite a few changes, I see. The only real question I have--which you may have already answered in another thread--is this: why the move to these DOT ratings over your earlier ELO system? As impressive as your DOT ratings are, it was ELO that really blew me away; I still think it is the most elegant and satisfying statistical method I have yet seen. Was the development of these new ratings a response to popular opinion, or did your own philosophy change? Thanks, and keep up the excellent work.

P.S. If you can come up with any sort of reasonably reliable system for dealing with the sparse data of the pre-open years, such that even a rudimentary all-time ranking could be developed, then I will be REALLY impressed!

noeledmonds
12-06-2007, 02:58 AM
Yes, I see that and I am trying to figure out why. I think that part of the problem is that underrates Borg's clay opponents.

I think that the problem actually relates to Borg's specific years of dominance. If we look for example over Borg's ratings over a 3 year period he would surely come out very high. Borg won 2 slams a year 3 years in a row. The only other open-era players to achieve that are Federer and Sampras. Borg also won at least 1 grand slam title 8 years in a row, an open-era feat only matched by Sampras.

While single year dominance clearly looks at a single year, longitivity in a career tends to look at a career of well over a decade long. However Borg's strengths come somewhere in between these 2.

AndrewD
12-06-2007, 07:00 AM
To answer each point in turn

1. Yes the Grand Slam has been recognised for as long as I can remember. It was always valued but whether the value was always at the level it is now, is another matter. I feel if the value of the Grand Slam despite it being recognised, for a very long time, was reduced in some previous decades. I cannot believe that it held the same cvalue when many players in the late 1970s didn't even bother playing the Australian Open. In any even, had it been achieved during this period it's value must have been less as the quality of opposition at the AO was so much lower. What do other people think?

2. Yes but before the move in month players like Evert weren't competing for the AO therefore excluding the possibility of chasing a GS. I ssume you are defining chasing as having the first three majors. Many players were not competing in the AO even before the change of month. Therefore GS became less of an indicator of greatness during these periods, therefoe the GS was not always held in in the same regard.

3. Yes but with people like Borg not playing the Australian at the start of the year, I'm not sure what this proves.

4. We'll have to disagree on that. I think I have proved it above.

5. Sorry I though McEnroe said in his Autobiography that he regretted not playing the AO. Not that he didn't feel it necessary because of the GS. Ithink you're making the GS the reason people didn't play to justify your argument when really there were other reasons.

other

1. I feel by introducing the artificial horizon of a calendar year I would be fiddling the sytem.

2. I disagree. The outputs are produced without human opinion or adjustment. To indroduce the calendar year as an artificial time horizon would be doing just that.

3. How can you say that the calendar year GS is the most significant measure of talent and then argue that the number of majors in a year is an irrelevance.

4. OK have it your way, Evert was a nobody in the 83 and 84 seasons and Martina should have no credit for beating her and therefore Martina's rating should be deflated from its current level. You better tell that to all the people Chris beat in winning two majors and a further four major finals not bad out of eight majors eh! That means Evert was a good player during this period. Then Navratilova beat her. Yes. Does that mean Chrissie was weak or that Navratilova was very strong? If the latter then that is what DOT is saying.

5. Over the course of anybodies career is an irelevance. DOT operates of a limited time frame wit Dominance x Opposition x Tournament. Chrisie scored higher using these statistical criteria than did Goolagong, King, Casals, Richey, Durr, Melville, Wade and Jones. Navratilova's score was adjusted accordingly to the wins she had against Evert. Nothing more, nothing less, no bias and no opinion. What do other people think. Was Chrissie's play in 1983 and 1984 poorer than Goolagong, King, Casals, Richey, Durr, Melville, Wade and Jones, or was Navratilova at the absolute peak of her game.

6. I don't give bonus points on reputation of a player I give bonus points on the quality of a player faced, i.e opposition. Chris was clearly dominant of the rest of the population in 1983 and 1984 with the exception of Navratilova. This is calculated independantly of any opinion. The fact that Martina then dominated her results in Martina's domination adjuster being raised. I find this argument far more persuasive than Chris dominated everyone else during this time because the whole world had become weaker than during the Court Era and Chris alone was the sole player at the old level.



Tim,

1. You've made two huge assumptions based on putting the cart before the horse. A lack of interest in the Australian Open does not equal a lack of interest in the Grand Slam. The only way you can logically make that statement is if a player had been capable of attaining the GS but had opted not to. No-one did so the argument is invalid. Your second assumption is that, if a player (Borg is the only option) had been shooting for the Grand Slam he would have been met with weak opposition. The reality is that other of the top players would have followed suit.

2. The Australian Open was played in December during the years 1971, 1972, 1973, 1974, 1975, 1977, 1978, 1979, 1980, 1981, 1982, 1983 (actually started in November), 1984 (Nov), 1985 (Nov).
That encompasses all of Borg's active years and takes us right through Evert's career up to the change of venue. Can you point out how anyone could be chasing the Grand Slam by winning the last one of the year?

3. The Australian Open wasn't played at the start of the year (apart from once, in 1977) until well after Borg had retired.

4. No, you didn't prove anything, you just said Evert challenged her. I showed you the results which contradict that statement.

5. On the contrary, you're the one who is trying to devalue the prestige and worth of the Grand Slam in order to justify your ratings. I've only pointed out to you that your claim, that the Grand Slam was not held in the same regard, is false.

6. Somewhere along the line you don't seem to be grasping what I'm saying. Very simply - not playing the Australian Open is no reflection on the status of the Grand Slam (the winning of all four tournaments in a calendar year). It is only a reflection on the status of the Australian Open. Inferring otherwise is just.....inane.

other

1. The calendar year, something observed by all of us, is not an 'artificial horizon'. The system you're following is.

2. Again, the system is artificial because you've introduced a framework which is totally alien and arbitrary.

3. A) I didn't say it was the most significant indicator of talent, although I would have said it is the pinnacle of achievement in our sport - it is. B) I didn't say that.

4, 5 & 6. Just don't work. Too much credit is given to Evert when she didn't offer the same challenge to Navratilova that King, Goolagong, Jones, Wade, etc offered to Court. That Evert was better than everyone except Navratilova says less about Evert and more about her opposition. 1983/84's top ten of Navratilova, Evert, Mandlikova, Shriver and Turnbull isn't as strong as 1970's King, Casals, Richey, Wade and Jones.

Tim, if nothing else, I think you need to go back and read up on the dates of the Aus Open during the 1970's and the reasons why certain players (the best ones), wouldn't, couldn't or weren't allowed to enter events like the Australian Open.

urban
12-06-2007, 09:12 AM
Only to the Grand Slam problem. Andrew has a point here, that the real Grand Slam was a big target even in the 70s. When the AO turned the date to December in 1977, it became the last station of a possible GS. Borg certainly would have played Australia, if he had come through at USO. And there were a lot of public comments by Connors and Mac, that they would have followed Borg on his feet to the last hurdle. Call it honor or pride, to give Borg the toughest competition available. Connors made the famous phrase at Wimbledon in 1978, that he would follow the SOB to the end of the world.

Nickognito
12-06-2007, 03:18 PM
There was competition. Today Nadal and Roddick would say: 'i'm happy to live in an era where a player like Federer can make the Grand Slam, I'm flattered and I hope he can do it.

:D

c.

CyBorg
12-06-2007, 04:03 PM
so maybe wuornos' numbers didn't watch old matches :)

'70s matches aren't old. Maybe you didn't watch old Hoad matches...

But you can keep trying to figure out why.

c.

Wuornos' numbers rate Borg very high over a three-year spread. But of course this doesn't matter to you as they don't align with your predisposed biases.

I've seen Hoad player a number of times. I'm still trying to figure out if you've seen either he or Gonzalez play on hardcourts. Good job ducking that one.

Nickognito
12-06-2007, 05:49 PM
Cyborg, it's not necessary to have seen everything, i'm not God, i'm not 100 years old, and I read books written by people who know tennis better than me.

Anyway, Borg in a five-years spread is far behind Lendl, for example. So very poor performance for a player usually considered as the best of open era and very better than Lendl. In 5 years, Borg won 8 slams vs Lendl's 6. He had less losses tha Lendl. So why he's so far behind Ivan? For the relative weakness of his opponents. Or, if not, why?

RELATIVE weakness.
c.

Wuornos
12-07-2007, 02:37 AM
Tim,

1. You've made two huge assumptions based on putting the cart before the horse. A lack of interest in the Australian Open does not equal a lack of interest in the Grand Slam. The only way you can logically make that statement is if a player had been capable of attaining the GS but had opted not to. No-one did so the argument is invalid. Your second assumption is that, if a player (Borg is the only option) had been shooting for the Grand Slam he would have been met with weak opposition. The reality is that other of the top players would have followed suit.

2. The Australian Open was played in December during the years 1971, 1972, 1973, 1974, 1975, 1977, 1978, 1979, 1980, 1981, 1982, 1983 (actually started in November), 1984 (Nov), 1985 (Nov).
That encompasses all of Borg's active years and takes us right through Evert's career up to the change of venue. Can you point out how anyone could be chasing the Grand Slam by winning the last one of the year?

3. The Australian Open wasn't played at the start of the year (apart from once, in 1977) until well after Borg had retired.

4. No, you didn't prove anything, you just said Evert challenged her. I showed you the results which contradict that statement.

5. On the contrary, you're the one who is trying to devalue the prestige and worth of the Grand Slam in order to justify your ratings. I've only pointed out to you that your claim, that the Grand Slam was not held in the same regard, is false.

6. Somewhere along the line you don't seem to be grasping what I'm saying. Very simply - not playing the Australian Open is no reflection on the status of the Grand Slam (the winning of all four tournaments in a calendar year). It is only a reflection on the status of the Australian Open. Inferring otherwise is just.....inane.

other

1. The calendar year, something observed by all of us, is not an 'artificial horizon'. The system you're following is.

2. Again, the system is artificial because you've introduced a framework which is totally alien and arbitrary.

3. A) I didn't say it was the most significant indicator of talent, although I would have said it is the pinnacle of achievement in our sport - it is. B) I didn't say that.

4, 5 & 6. Just don't work. Too much credit is given to Evert when she didn't offer the same challenge to Navratilova that King, Goolagong, Jones, Wade, etc offered to Court. That Evert was better than everyone except Navratilova says less about Evert and more about her opposition. 1983/84's top ten of Navratilova, Evert, Mandlikova, Shriver and Turnbull isn't as strong as 1970's King, Casals, Richey, Wade and Jones.

Tim, if nothing else, I think you need to go back and read up on the dates of the Aus Open during the 1970's and the reasons why certain players (the best ones), wouldn't, couldn't or weren't allowed to enter events like the Australian Open.

To be honest I'm fed up with all this now. I'm not sure whether your deliberately misunderstanding to try to annoy me or whether you are genuine.

First of all your final comment. I am well aware of when the Australian Opens were. However it is irrelevant to a system that operates off a rolling horizon and is far more concern if you operate of a calendar year. I am also aware that you said the opposite in a previous post that the fact the dates changed didn't matter if I was operating off a calendar year.

How can I put this. If there is a lack of interest in the Australina Open, what does that do to the value of a Grand Slam?

I have introduced a framework that is alien and arbitrary. I assume you mean because it has a rolling horizon. Go and tell the ATP to amend their rolling horizon for rankings. I didn't introduce it, they did, and the NO 1 status calculated on this basis, which is essentially a moe simplistic version of DOT is valued.

God all mighty. I can't believe that your arguing that the whole world got weaker during the time of Navratilova and Evert. I have to hold my hands up to that. DOT does assume constant population standard over time at around the 50th position in the rankings. Personally I find it more likely that Chris and Martina were good than everyone else was suddenly significantly worse. I'm not arguing that Martina didn't dominate Evert more than Court dominated King. She did. The question is do we place evrything on this head to head or do we look at Chrissies other results, conclude she was very very good and mratina was better. DOT looks to the latter on teh basis that statistically it was more likely.

You do understand they we are discussing the old rating methodology don't you as I abandoned the new one at post #18 when you convinced me the new methodology was flawed. The ratings you are saying are incorrect and that we are discussing are the following ones. I thought you liked them and they were calculated using the methodology you are arguing against. The following list shows all players obtaining a rating above 2700 DOT.

1 Margaret Smith Court 2880
2 Steffi Graf 2880
3 Martina Navratilova 2878
4 Monica Seles 2837
5 Chris Evert 2823
6 Serena Williams 2797
7 Billie Jean King 2781
8 Arantxa Sánchez Vicario 2769
9 Martina Hingis 2762
10 Evonne Goolagong 2752
11 Venus Williams 2750
12 Justine Henin 2749
13 Hana Mandlíková 2747
14 Lindsay Davenport 2731
15 Gabriela Sabatini 2721
16 Jennifer Capriati 2713
17 Ann Jones 2705
18 Amélie Mauresmo 2701

Sorry but unless we're talking at cross purposes I think we'll have to agree to disagree on this. I believe the above output shows great quality in looking and evaluating results much further than the basic data. Notice Seles above Evert despite a much shorter career and Court above both Graf and Navratilova. Notice Avonne Golagong in the top 10 above Henin, because of the sparsity of talent in the women's game at the moment, etc etc.

Sorry we can't agree

Regards

Tim

Wuornos
12-07-2007, 02:53 AM
It's been a while since I've followed up on this stuff... quite a few changes, I see. The only real question I have--which you may have already answered in another thread--is this: why the move to these DOT ratings over your earlier ELO system? As impressive as your DOT ratings are, it was ELO that really blew me away; I still think it is the most elegant and satisfying statistical method I have yet seen. Was the development of these new ratings a response to popular opinion, or did your own philosophy change? Thanks, and keep up the excellent work.

P.S. If you can come up with any sort of reasonably reliable system for dealing with the sparse data of the pre-open years, such that even a rudimentary all-time ranking could be developed, then I will be REALLY impressed!

Hi Chaog and thanks for the kind comments.

Yes the ELO System was interesting but I kept facing two criticisms which ELO didn't handle particularly well. The new system is more of a hybrid system and does incorporate some aspects of ELO.

The main problems were.

1. People were used to one year horizon and by and large I don't think people could understand that ELO ratings didn't automatically decay overtime given inactivity. E.g. Using ELO Lindsay Davenport would return to the game at the same level as she left a year ago.

2. People were used to a volume based system and they couldn't appreciate that someone who had a smaller number of very good results would be rated better than a high activity player. E.g. ELO always rated Jelena Janković lower than people felt she deserved.

The new system still takes into account many of the aspects of ELO. It is still a system that is based upon quality of results and looks at the opposition faced by a player rather than fixed points for individual tournaments, although this is part of DOT too. The fixed hoizon of one year in the official rankings was just to short to be considered statistically reliable and therefore DOT is based over a longer period of 4 years with the current year accounting for about a half of a players current weighting and the ealier three years making up the other half. This results in a more realistic decline in relation to inactivity and gives a better indication of standard when a player returns.

Like ELO the population is monitored for consistency and sometimes the top of the poplulation can be stronger or weaker. The population however always remains constant.

If you have any further questions let me know, but by and large I feel this system is probably more reliable than ELO.

A desription of methodology can be found at:

http://tt.tennis-warehouse.com/showthread.php?t=165420.

Ignore the top posts in this thread, I have abandoned that methodology as flawed and returned to version 2.3 which was my best. I think it is unlikely that this methodology will change again now, or if it does changes will be comparitilvely minor. The peak DOT Ratings for men of the open era exceeding 2700 are as follows:

1 Roger Federer 2841
2 Rod Laver 2806
3 Ivan Lendl 2786
4 Björn Borg 2771
5 Mats Wilander 2764
6 Pete Sampras 2762
7 John McEnroe 2738
8 Ken Rosewall 2736
9 Jim Courier 2734
10 Boris Becker 2726
11 Stefan Edberg 2716
12 Rafael Nadal 2711
13 John Newcombe 2710
14 Jimmy Connors 2706


Take care and keep in touch.

Tim

Nickognito
12-07-2007, 02:56 AM
Glicko rating solve the first problem, the second one is not a problem, it's the right thing ;-)

c.

Wuornos
12-07-2007, 03:04 AM
Thanks Wuornos.
I'm only repeating myself and say that in my opinion DOT rankings do not rank in the right order players careers, because they give domimance too much importance. And I would like to 'abuse' your ranking :). I.e. I think that your ranking is so well made that iit's wishable to be able to integrate it with other considerations in order to make a better overall ranking. Maybe it's not up to you, but i think it's possible to improve your method. I don't know how, at the moment, i'm sorry. And obviously others can 'abuse' your ranking saying that it's already perfect in describing an open era ranking ;-)

Regards

c.

Thanks Nick

Yes an amalgamated system using different time bands was what I was trying to do with version 3.2, but to no availe. It didn't really look right. I'm really coming up against a bit of brick wall now. At first improving the system was comparatively easy and then as it got better, improvements became incresingly difficult to come by. I'm hoing for a sudden revelation at some point to take things forward the next step but my gut feeling at the moment is this is about as good as it's going to get.

Thanks for all the feedback. Appreciated.

Tim :)

Wuornos
12-07-2007, 03:13 AM
I use "well" as a relative term. Obviously Borg did do very well in these years compared to what most players have achieved in their career. However Borg's best year only reaches number 10 on Tim's ranking list, which relative to Borg's reputation and generally held status is not very high at all.

No, No, NO. That system was dropped as being flawed.

In Post #16 I reverted back to the old system whose methodology was far sounder. The old methodology was desribed at http://tt.tennis-warehouse.com/showthread.php?t=165420.

The old methodology gives the follwing list as all male singles players with a peak DOT in excess of 2700.

1 Roger Federer 2841
2 Rod Laver 2806
3 Ivan Lendl 2786
4 Björn Borg 2771
5 Mats Wilander 2764
6 Pete Sampras 2762
7 John McEnroe 2738
8 Ken Rosewall 2736
9 Jim Courier 2734
10 Boris Becker 2726
11 Stefan Edberg 2716
12 Rafael Nadal 2711
13 John Newcombe 2710
14 Jimmy Connors 2706

Regards

Tim