# Did we reach a consensus on whether a double bagel counts in your rating?

Discussion in 'Adult League & Tournament Talk' started by Maui19, Jan 28, 2013.

1. ### Maui19Hall of Fame

Joined:
Sep 18, 2010
Messages:
1,671
I've seen the question asked about whether a 6-0, 6-0 result is included in calculation of a player's rating. I have seen the following answers: yes, no, and maybe. Does anyone know for sure whether a double bagel does or doesn't count?

Thanks.

2. ### dizzlmcwizzlHall of Fame

Joined:
Mar 24, 2010
Messages:
2,200
Location:
DE
In a typical non-answer answer, I will say I think so ... now.

It used to be that you could find references of the double bagel not counting on almost every section's website. Taken word for word, it was almost like they cut and pasted from the same official USTA document.

However, now it is hard to find a reference regading rating calculation at all. The only sites that still have this reference appear out-dated.

Additionaly we have have had two folks claiming some level of inside knowledge claiming that they have changed the formula and they now count.

Bottom line is no one knows for sure, but my magic 8-ball says signs point to yes ... they count.

3. ### Maui19Hall of Fame

Joined:
Sep 18, 2010
Messages:
1,671
I'll trust your magic 8-ball. Thanks.

4. ### beernutzHall of Fame

Joined:
Aug 16, 2005
Messages:
4,539
Location:
Make TT better, use the Ignore List!
So which end of the double bagel were you on? :twisted: j/k

Well I just saw your other post so I retract the question your honor.

Last edited: Jan 28, 2013
5. ### CindysphinxG.O.A.T.

Joined:
Aug 31, 2006
Messages:
15,360
I don't know if they count, but they certainly should.

Why on earth should a losing player be better off ratings-wise if they lose -0 and -0 than if they lose -0 an -1?

6. ### gameboyHall of Fame

Joined:
Jul 10, 2009
Messages:
1,620
Double bagel is tough for calculation since you cannot accurately formulate exactly what the difference is between winning and losing player.

Let me explain...

Let say you have a player rating scale from 1 to 10. Let's also say for every 1 rating difference between players equal the number of game difference in a set. So, when a 10 rating player plays 7 rated player, the sets will most likely end up 6-3.

This works fine as long as the difference in rating is 6 or less, but what happens when the difference is more?

If a 10 rated player plays a 4 rated player, the expected result is 6-0. But what about a 3 rated player plays 10 rated player? Obviously, it will be 6-0. But that means that the loser could be a 4, 3, 2, or 1 rated player. You won't know the difference. The same goes the other way. If a known 2 rated players loses 6-0, you won't know whether or not the opponent is 7, 8, 9, or 10 rated player.

That is why counting double bagel games are tricky.

7. ### CindysphinxG.O.A.T.

Joined:
Aug 31, 2006
Messages:
15,360
Math is stoopid.

8. ### schmkeHall of Fame

Joined:
Jan 20, 2010
Messages:
1,809
Excellent example. And at the risk of more math comments from Cindy, I'll expand on it a bit.

As noted above, when the expected result of a match is greater than 6-0, a 6-0 score doesn't tell you much of anything. However, if the expected result is 6-1 or closer, then 6-0 does have meaning.

So, IMHO, double bagel results should be thrown out when the expected result was a double bagel or worse, but if the expected result is closer than that, they should be counted. I hope this is what the USTA is now doing.

9. ### NumbersGuyRookie

Joined:
Jun 5, 2012
Messages:
176
That's why the computer throws out ALL results when the players are more than 0.5 apart and thus the expected result would be greater than 0 and 0. And yes double bagels counted in YE12 ratings and going forward. I trust my USTA source.

10. ### CindysphinxG.O.A.T.

Joined:
Aug 31, 2006
Messages:
15,360
How does this work for doubles?

11. ### NumbersGuyRookie

Joined:
Jun 5, 2012
Messages:
176
Sorry. I should have said when the teams are more than 0.5 apart. So, say, two 4.0s against a 4.0 and a 2.5. What I'm not certain about, is the 0.5 threshold the difference between the Levels (e.g. 4.0-3.5=0.5), or the Dynamic Ratings (e.g. 3.83-3.30=0.53).

12. ### schmkeHall of Fame

Joined:
Jan 20, 2010
Messages:
1,809
I would think it has to be the difference of the dynamic ratings.

13. ### schmkeHall of Fame

Joined:
Jan 20, 2010
Messages:
1,809
So, just to confirm, your source tells you that a close win or even a loss by the higher rated player is thrown out if the higher rated player is more than 0.5 above the lower one?

Throwing out a 0 and 0 match in this scenario makes sense, but if the match is closer than that, and particularly if it is two close sets or even an upset, one can make the case that the match should count.

14. ### dizzlmcwizzlHall of Fame

Joined:
Mar 24, 2010
Messages:
2,200
Location:
DE
You could also argue that by not counting these results sandbagging would be more difficult.

15. ### NumbersGuyRookie

Joined:
Jun 5, 2012
Messages:
176
Yes.

And I suspect the USTA's logic is that a match should not count if there is only one direction that players' ratings could possibly move as a result of it. And yes it makes tanking more difficult, which is a good thing.

16. ### schmkeHall of Fame

Joined:
Jan 20, 2010
Messages:
1,809
Yep, you are right. With most of these issues, there are two sides to the coin. One just has to decide which side they want to err on. In this case, do you want to reward the lower rated player if they play great and do better than expected? Or leave open a way for higher rated players to still win easily but help keep their rating down?

Given that most people might play just 3-8, matches a year, I've leaned towards trying to count every match to get all the data I can. But I can see how these matches between unmatched opponents could be considered meaningless, or at least less meaningful, and there shouldn't be that many played, although with folks playing up, there might be more than you think. It would just be a pity to not count them at all.

17. ### chatt_townHall of Fame

Joined:
Mar 17, 2010
Messages:
2,125
Why would it have not ever been included. I thought all scores counted. That's why when there were defaults I would tell the captain to put the scores in under someone else's name because I was always thinking them putting 0 and 0 scores would screw me over. Maybe they have some way of knowing it's a default and don't count that toward you but if you go out and double bagel someone in a match, I don't see why it would have not counted.

18. ### beernutzHall of Fame

Joined:
Aug 16, 2005
Messages:
4,539
Location:
Make TT better, use the Ignore List!
When someone wins by default, there is no opponent entered for the winning player. It just says 'N/A Default' in tennislink so presumably the computer doesn't take those scores into account when computing someone's DNTRP.

19. ### schmkeHall of Fame

Joined:
Jan 20, 2010
Messages:
1,809
Correct. Defaults don't count and I believe certain other incomplete matches (retirement without completing a set I believe is one, probably others) don't count too. So there is no need to worry about defaults entered in the system hurting (or helping) your rating.

And to the earlier question of why they ever would not have counted, there was a time (up until 2012?) that USTA FAQs and other documents clearly indicated that 0 and 0 matches were not counted and most believe that to mean they were thrown out entirely. It does appear that this has changed for 2012 and going forward and they are counted when the players involved are rated close enough for the score to be meaningful.

20. ### Maui19Hall of Fame

Joined:
Sep 18, 2010
Messages:
1,671
So when I'm paired with another 3.5 and we're playing two 4.0s, the results won't count? Sheesh, if this is true and they're throwing out double bagels, I've probably only had 5 matches that count in the last two years.

21. ### schmkeHall of Fame

Joined:
Jan 20, 2010
Messages:
1,809
From what NumbersGuy is reporting, yes.

However, I'd assume it is using actual dynamic ratings and not just your NTRP level that is used.

So if you and your partner are in the high end of the 3.5 range, say 3.4, and your opponents are even in the middle of the 4.0 range, say 3.75, the difference isn't too large and the match would count. It would only be if you were in the low end of a 3.5, say 3.1 and were playing a high end 4.0, say 3.9, that the difference would be too large and the match wouldn't count.

22. ### NumbersGuyRookie

Joined:
Jun 5, 2012
Messages:
176
That's how I would see it. The way I rationalize the USTA's non-use of matches where you're more than 0.5 apart (caution! math geeks only!) is as follows.

If it were expected that I would beat a guy, say, 6-3, 6-3 (i.e. we're about 0.25 apart), and I played him many times with neither of us getting better or worse, then sometimes I'd win 3 and 3, sometimes 2 and 2, sometimes 4 and 4 (among many other possible outcomes of course), with an average result of 3 and 3. Now let's say I play a guy who I'm 0.5 better than. I'd be expected to win 0 and 0. Sometimes I'd win 1 and 1 or 2 and 2, but there would be no offsetting 7-0, 7-0 or 8-0, 8-0 scores to get the overall average result to 0 and 0. So my rating is virtually guaranteed to go down, and my opponent's to go up. You can argue that the matches are data points, but they're biased data, and I'd argue that biased data are worse than no data. By this logic, I expect that matches where you're exactly 0.5 apart are also disregarded.

23. ### schmkeHall of Fame

Joined:
Jan 20, 2010
Messages:
1,809
<mathgeeks>
By the same logic, should matches between players 0.45 apart be disregarded too? The expected score might be 6-1, 6-0, but like you say, sometimes it might be 6-2, 6-1, or even 6-3, 6-2 sometimes, but there is no 6-0, 7-0 or 7-0, 8-0 to balance it out.

I'm not disagreeing with your logic, just trying to figure out where one would reasonably draw the line.
</mathgeeks>

24. ### NumbersGuyRookie

Joined:
Jun 5, 2012
Messages:
176
Good point. They could choose to draw it at less than 0.50 apart. But I should think not at more than 0.50 apart where even achieving the expected result results in a change of rating.