Meles
Bionic Poster
I'm going to OP this one differently than previous ELO threads in that the technical weeds of the discussion will follow the graph and conclusions:
1. Field strength has been weakening on grass (peak ELO of Wimbledon QFers) and we are in a weak era on grass (not what I wanted to here.
)
2. The early 80's are also weak, but are we really suprised in the era of Chris Lewis?
3. Late 80's and early 90's have the strongest field (a surprise for me as I've always held Connors, Mac, Borg era in higher esteem.
)
4. The field weakened through out the 1990s cratering in 2002 (Nalbandian in final.
) Sampras cleaned up in a weak era.
Considerations:
1. The peak ELO rating is from a list compiled at the end of 2012. Ratings were conservatively fudged for Raonic(2100), Cilc, Dimitrov, and Kyrgios. 2016 match data was not available from my source and I've avoided manually putting in the players as the trend is clear. The 1978-1980 ratings are very suspect. Borg's grass ELO is quite low in this list and I fudged it up to 2300. I don't have the tennis knowledge to begin to fudge the lesser players. We can make more corrections based on TTWer input, but I'd say 1978-1980 is a lost cause. Everyone starts out at 1500 at beginning of 1978.
2. I'd love to have current surface ELO at tournament time versus just peak, but that data is not availabe. For evaluating field strength, peak is probably very good and the players had to be in good form to make QFs. This is in the background, but QF evens out over the peak issues with the number of players.
3. I'd like to have an explanation for Sampras, but these are slam weighted peak grass ratings for players. They are set based so in bo5 a 3 set win carries more weight than a five set win. Did Sampras have some small grass events where he did very poorly every year? ELO rates Sampras as 3rd best of this era, but Edberg 2. Their ratings are almost identical.
I can post other data views on request, just ask.
The rank is the all-time grass rank (post 1978) of winner and runner-up. The QF and SF ELO are the averages of the participants.
Also I will be able to generate data for the three other slams very quickly. Wimbledon is the guinea pig as grass stats are notoriously unreliable, so ELO seems to be a great solution, but faces some of the same issues due to the shortness of the grass season. (Another reason that peak ELO may be the better rating for field.) Please ATTACK anything that you think ELO is over rating or under rating. Also the method of using peak ELO of QFs could be modified. I show graph of SF. R16 field strenght may be possible, but I stayed away as I thought it more likely to have players with ELO ratings outside the top 500 in R16. As it stands, Nastase was the only failed lookup for QFs (1978) and Janowicz in 2014. Doing R16 might trigger a lot more exceptions and guestimates.
Please see the predecessor to this thread for some initial discussion:
http://tt.tennis-warehouse.com/inde...ng-era-wimbledon-finals.569169/#post-10512433
Here is the best discussion I've seen of using ELO for tennis. The table with the odds of upset for the various different ratings is very good:
http://www.ighs-tennis.com/IGHST_Method.pdf (Yes, ELO is being used for tennis and is vastly superior to the current ranking system.)
http://www.mens tennis forums.com/11-statistics/211663-elo-ratings.html (get rid of the spaces; this has the four surface ratings used, slam weighted too.)
Slackmann's recent all time hard and clay ratings:
http://www.tennisabstract.com/blog/...-djokovic-and-roger-federer-and-rafael-nadal/
Slackmann's awesome github:
https://github.com/JeffSackmann/tennis_atp (all the match data of the open era in csv format for easy import.)
1. Field strength has been weakening on grass (peak ELO of Wimbledon QFers) and we are in a weak era on grass (not what I wanted to here.
)2. The early 80's are also weak, but are we really suprised in the era of Chris Lewis?
3. Late 80's and early 90's have the strongest field (a surprise for me as I've always held Connors, Mac, Borg era in higher esteem.
4. The field weakened through out the 1990s cratering in 2002 (Nalbandian in final.
) Sampras cleaned up in a weak era.Considerations:
1. The peak ELO rating is from a list compiled at the end of 2012. Ratings were conservatively fudged for Raonic(2100), Cilc, Dimitrov, and Kyrgios. 2016 match data was not available from my source and I've avoided manually putting in the players as the trend is clear. The 1978-1980 ratings are very suspect. Borg's grass ELO is quite low in this list and I fudged it up to 2300. I don't have the tennis knowledge to begin to fudge the lesser players. We can make more corrections based on TTWer input, but I'd say 1978-1980 is a lost cause. Everyone starts out at 1500 at beginning of 1978.
2. I'd love to have current surface ELO at tournament time versus just peak, but that data is not availabe. For evaluating field strength, peak is probably very good and the players had to be in good form to make QFs. This is in the background, but QF evens out over the peak issues with the number of players.
3. I'd like to have an explanation for Sampras, but these are slam weighted peak grass ratings for players. They are set based so in bo5 a 3 set win carries more weight than a five set win. Did Sampras have some small grass events where he did very poorly every year? ELO rates Sampras as 3rd best of this era, but Edberg 2.
Their ratings are almost identical.
I can post other data views on request, just ask.
The rank is the all-time grass rank (post 1978) of winner and runner-up. The QF and SF ELO are the averages of the participants.Also I will be able to generate data for the three other slams very quickly. Wimbledon is the guinea pig as grass stats are notoriously unreliable, so ELO seems to be a great solution, but faces some of the same issues due to the shortness of the grass season. (Another reason that peak ELO may be the better rating for field.) Please ATTACK anything that you think ELO is over rating or under rating. Also the method of using peak ELO of QFs could be modified. I show graph of SF. R16 field strenght may be possible, but I stayed away as I thought it more likely to have players with ELO ratings outside the top 500 in R16. As it stands, Nastase was the only failed lookup for QFs (1978) and Janowicz in 2014. Doing R16 might trigger a lot more exceptions and guestimates.
Please see the predecessor to this thread for some initial discussion:
http://tt.tennis-warehouse.com/inde...ng-era-wimbledon-finals.569169/#post-10512433
Here is the best discussion I've seen of using ELO for tennis. The table with the odds of upset for the various different ratings is very good:
http://www.ighs-tennis.com/IGHST_Method.pdf (Yes, ELO is being used for tennis and is vastly superior to the current ranking system.)
http://www.mens tennis forums.com/11-statistics/211663-elo-ratings.html (get rid of the spaces; this has the four surface ratings used, slam weighted too.)
Slackmann's recent all time hard and clay ratings:
http://www.tennisabstract.com/blog/...-djokovic-and-roger-federer-and-rafael-nadal/
Slackmann's awesome github:
https://github.com/JeffSackmann/tennis_atp (all the match data of the open era in csv format for easy import.)
Last edited:
This is kind of a weak era detente for all the various camps. I expect this may stir up a bit of a Sampras hornet's nest.
)
