New Tennis Ratings Site
- Thread starter ronray43
- Start date
OnTheLine
Hall of Fame
Sorry, fat fingers ... score was 6-4 ; 6-4 !!
navigator
Hall of Fame
Just played at the Rome Tennis Center a couple of weeks ago in the ITF North American Regional Championships (Seniors). Great staff, very well run. Courts are in great condition. Two negatives: (1) No clay courts (baffling), and (2) Rome is in a beautiful part of the world, but... it's boring as hell there; zero to do. I don't know how players who travel with their significant others are going to convince them to come to Rome for a week. If you're not playing, it's pretty dreadful.
FearTheJester
New User
Does anyone have an email for an admin at tennisrecord.com? The site keeps a running dynamic rating but doesn't break it up by season.
I don't have an e-mail but tried to reach out to them. I've been able to backdoor engineer their formula a tiny bit. Assuming you have played more than 3 matches your current rating will often be the average of your previous three ratings and your last match score. It seems that new players or self-rate players are given a rating smack dab in the middle of the range (i.e. 3.25, 3.75, 4.25 or 4.75) not just for their first match but at the beginning of new seasons as well. This isn't correct (obviously) and taking this shortcut negatively affects the ratings for everyone. I'm not sure how they generate a match rating (AWR or ALR) but it doesn't seem too far off from previous posts (see link below). They are netting the actual result from the predicted result and adding or subtracting from your previous rating. I would assume people much smarter than I can do a better job of backdoor engineering the formula they are using
https://tt.tennis-warehouse.com/index.php?threads/question-on-dynamic-ratings.116331/
https://tt.tennis-warehouse.com/index.php?threads/question-on-dynamic-ratings.116331/
NumbersGuy
Rookie
Good catch. They also still disregard 6-0, 6-0 matches. Perhaps the way we're supposed to communicate with them is via this forum.
ncgator
Rookie
I received the update on new features and updates from news@tennisrecord.com
J011yroger
Talk Tennis Guru
NumbersGuy
Rookie
Easy to get any player's email, plus their phone number and USTA number, from TennisLink if you know how. Probably why they're changing their security protocols.
J011yroger
Talk Tennis Guru
I'm such a loser. As soon as I play a match I keep clicking refresh to see if it is in the 5.0 range.
J
J
D
Deleted member 23235
Guest
lol yeah, me too.
also making me avoid doubles (ie. with weak dubs partner)
similar to the bucket list item of running a marathon,... getting bumped to 5.0 is my current goal (if even for just a season).
J011yroger
Talk Tennis Guru
Ya, I'm like I'm bageling people at #1 singles, what more do you want from me.
J
D
Deleted member 23235
Guest
i wonder if it's harder to go up by beating folks lower than you... ie. beating a 4.1 6-1,6-1 might be harder than losing 4-6,4-6 to a 4.9 (which might make your rating go up also)
but hypothetically you win an "easy" match against a 4.1 say 6-3,6-3.. your rating might not change or even go down? (maybe you were off that day, maybe you were trying tsomeone new, maybe that player was on a hot streak, matchup issue, etc...)
I think you're right that you're better off losing playing up a level than beating lower players at your level.
I just did a quick run through of my local area last year. 75% of players bumped up played up a level. Only 25% of players bumped played only at level for the adult season.
J011yroger
Talk Tennis Guru
My plan is to beat the best players at my level and the ok players at the next level. Losing isn't part of my strategy.
J
D
Deleted member 23235
Guest
lol, yeah, always try to win... but sometimes your opponent has a say in that
just saying, that from a "fastest-ratings-bump" perspective... might be better off always playing the best players (eg. even if it means losing when on an upper NTRP level)
i find winning 0,0 or 1,1 sometimes difficult cuz my mind starts to wander, or i'm generally less engaged... so i might go up 4-0, then end up winning only 6-3 (when i should have won 6-1)
Last edited by a moderator:
OnTheLine
Hall of Fame
One note on the league section ... .when I go to look at how my teams rate, I find that ALL teams connected with my club (over 30 different teams) are all linking to one particular 6.0 Mixed team from our club. Doesn't matter how you drill down on it, (picking women, picking men, picking 4.5, 4.0, 3.5, picking 40+, picking 18+) doesn't matter, all teams from our club are this Mixed team on the site.
Some programming glitch, it doesn't seem to effect other club's teams, just ours.
Some programming glitch, it doesn't seem to effect other club's teams, just ours.
S&V-not_dead_yet
Talk Tennis Guru
Sounds like the "All roads lead to Rome" bug.
Nostradamus
Bionic Poster
it needs to be completely Separate rating for Singles and Doubles. if you mix them all together, then it is completely foolishly false. Singles only, and Doubles only, that is how it should be done. If you don't play singles, you don't get singles rating.
I always love seeing the doubles expert going out to sandbag by playing singles once the local league is wrapped up.
OrangePower
Legend
Haha funny you should say that... I was watching a match over the weekend where that's exactly what was going on - older guy, dubs specialist, great record and probably near top of level, playing at #1 singles. His team had already clinched their playoff spot, so likely just trying to manage his rating. Problem is, opponent is self rated, and looking to keep scores close to avoid possible DQ. Fun times!
Your record is quite good this year. You should be close to getting bumped. Just play a couple more 5.0 league matches but try to avoid the 5.5 players. LOL.
J011yroger
Talk Tennis Guru
Amazing times when guys with current ATP points are losing in 5.0. I'm more of a second doubles guy than a futures circuit guy lol.
J
I am trying to understand the algorithm, and here is a comparison of two singles matches:
Name / Incoming rating / Score / Match rating / Resulting rating
Adam ------------ 3.30 / W 6-2, 6-2 / 3.55 / 3.36
Bob --------------- 3.37 / L 2-6, 2-6 / 3.11 / 3.31
Charlie ------------ 3.34 / W 7-6, 7-6 / 3.15 /3.28
Dan ---------------- 2.95 / L 6-7, 6-7 / 3.04 / 3.00
1. Don’t pay attention to outgoing ratings, I am curious about how match ratings work.
As you see, Bob and Charlie both lost within a couple weeks. The winners' incoming rating for both matches is roughly the same: 3.30-3.34. So, on paper, they played against the same opponent.
Now, Bob was a paper favorite before his match, and is supposed to narrowly win, but instead he loses badly, 6-2, 6-2. Dan does somewhat the opposite: he is supposed to get blown off the court (being 0.39 point lower), but loses a very close match, albeit in straight sets.
Based on scores, shouldn’t Bob’s match rating be something like 2.9 (for posting a big loss to a 3.3 player) and Dan’s match rating be something like 3.2 or even 3.25 (for playing an even match with a 3.3 player)? But no, Bob still has a match rating of 3.11, higher than Dan's 3.04.
2. Let’s look at Charlie-Dan match again.
Assuming the incoming ratings somewhat accurately reflect the strengths, the score should have been a blowout, which didn’t happen. So either Charlie played a very poor match that day, or Dan played lights out, or little bit of both. Since USTA (or, in this case, Tennisrecord), doesn’t know which happened, it should assume it was exactly same amount of both. Therefore, match ratings should be set the at same distances from incoming ratings. For example, if you say Charlie underperformed his rating of 3.34 by 0.15, then Dan must have overperformed his rating of 2.95 by 0.15. Tennisrecord doesn't seem to think so. Charlie is assumed to have underperformed his rating by 0.19 while is Dan assumed to have overperformed his rating by only 0.09. It looks like whenever a higher rated player underperforms, the system tilts toward him having a bad day instead of other scenarios. In case of Adam-Bob, both players are somewhat evenly matched, hence it’s assumed to be “almost equal amount of both" scenario.
Name / Incoming rating / Score / Match rating / Resulting rating
Adam ------------ 3.30 / W 6-2, 6-2 / 3.55 / 3.36
Bob --------------- 3.37 / L 2-6, 2-6 / 3.11 / 3.31
Charlie ------------ 3.34 / W 7-6, 7-6 / 3.15 /3.28
Dan ---------------- 2.95 / L 6-7, 6-7 / 3.04 / 3.00
1. Don’t pay attention to outgoing ratings, I am curious about how match ratings work.
As you see, Bob and Charlie both lost within a couple weeks. The winners' incoming rating for both matches is roughly the same: 3.30-3.34. So, on paper, they played against the same opponent.
Now, Bob was a paper favorite before his match, and is supposed to narrowly win, but instead he loses badly, 6-2, 6-2. Dan does somewhat the opposite: he is supposed to get blown off the court (being 0.39 point lower), but loses a very close match, albeit in straight sets.
Based on scores, shouldn’t Bob’s match rating be something like 2.9 (for posting a big loss to a 3.3 player) and Dan’s match rating be something like 3.2 or even 3.25 (for playing an even match with a 3.3 player)? But no, Bob still has a match rating of 3.11, higher than Dan's 3.04.
2. Let’s look at Charlie-Dan match again.
Assuming the incoming ratings somewhat accurately reflect the strengths, the score should have been a blowout, which didn’t happen. So either Charlie played a very poor match that day, or Dan played lights out, or little bit of both. Since USTA (or, in this case, Tennisrecord), doesn’t know which happened, it should assume it was exactly same amount of both. Therefore, match ratings should be set the at same distances from incoming ratings. For example, if you say Charlie underperformed his rating of 3.34 by 0.15, then Dan must have overperformed his rating of 2.95 by 0.15. Tennisrecord doesn't seem to think so. Charlie is assumed to have underperformed his rating by 0.19 while is Dan assumed to have overperformed his rating by only 0.09. It looks like whenever a higher rated player underperforms, the system tilts toward him having a bad day instead of other scenarios. In case of Adam-Bob, both players are somewhat evenly matched, hence it’s assumed to be “almost equal amount of both" scenario.
Last edited:
J011yroger
Talk Tennis Guru
Uh oh, some weird calculations this week. I put up a better result than my rating, and my rating didn't change. My friend put up a worse result than their rating and their rating went up.
J
J
J011yroger
Talk Tennis Guru
The site also gives more credit to the higher rated player in doubles.
J
J
TennisLeagueAnalytics
New User
It seems that rating accuracy is not the goal of the site. The model seems to be a "click bait" model. I've analyzed the numbers fairly well and they don't add up. Just look at the prediction of who should be bumped up and in one district it was wrong 95% of the time. But again, as long a players keep clicking.......
TennisLeagueAnalytics
New User
That would be nice but hard to do in practice because you'd have to also calculate the rating with the opponents' singles/doubles separated ratings. The permutations would get out of hand fairly quickly unless simplified and then, well, you might as well just use the all inclusive rating and make the singles/doubles (and mixed, combo, tri-level for that matter) adjustment elsewhere. That's the approach I take and it works very, very well.
Last edited:
TennisLeagueAnalytics
New User
That's just inaccurate, if true, but I'm curious about how you determined the "credit". The higher rated doubles player does give an advantage.
There's a very nice chart (that I can't reference here), that was posted recently in a usta league blog that shows the probability of winning given a difference in dynamic ratings (player strength rating) for doubles. Those likelihoods were based on 100,000 2016 matches.
In God we trust, everyone else bring data.
Moveforwardalways
Hall of Fame
Do you run the competing site tennisleaguestats.com?
TennisLeagueAnalytics
New User
Sorry, that's not my site.
J011yroger
Talk Tennis Guru
Find a doubles match, look at the player's match rating for it. Click on their partner, look at the partner's match rating. The player with the higher rating will have the higher match rating.
J
Why wouldn't he have a higher rating? It's not like the two players played exactly even.
TennisLeagueAnalytics
New User
Thanks. What's the definition of "match rating"?
J011yroger
Talk Tennis Guru
J011yroger
Talk Tennis Guru
I think the USTA officially said once that the differential in dynamic rating between doubles partners is maintained in the match ratings for doubles matches, so that if the partners started the match with a differential of 0.15, then their differential in the match rating also should be 0.15 regardless of the result. Is that what you are seeing?
J011yroger
Talk Tennis Guru
Anecdotally, I'd say 90% of the time they are consistent when calculating the most recent rating. The formula is rather simple: the last match rating you received averaged with your last three ratings. In the attached image the average of every number in red equals the number in green. How do they come up with a match rating? That's the question I'm trying to figure out.
I also agree with the assessment the the differential in dynamic rating between doubles partners is maintained in the match ratings for doubles matches. If Player X's match score is .15 higher than his last overall rating the same should be true of his partner. Also, that should be true of their opponents only in reverse. Each opponent's match rating should be .15 lower than their respective last overall rating.
J011yroger
Talk Tennis Guru
Crawling up at a snail's pace!
J
J
OnTheLine
Hall of Fame
and doubles partners do not get "equal credit" for a win.
Last match played, score Win 6-1; 6-0 My partner's eNTRP went up .08 and mine only went up .02 . our starting rating was only .03 different from each other, I had been .03 "higher" and now partner is .03 higher ... bizarre.
Yes the question is how is the match rating determined .... the rest is an overly simplified formula it seems.
Last match played, score Win 6-1; 6-0 My partner's eNTRP went up .08 and mine only went up .02 . our starting rating was only .03 different from each other, I had been .03 "higher" and now partner is .03 higher ... bizarre.
Yes the question is how is the match rating determined .... the rest is an overly simplified formula it seems.
OnTheLine...just to be clear, what are you calling eNTRP (the number on the left or the right?). The overall rating (number on the right) won't go up equally with your partner after each match but your match rating (what I call AWR or ALR) should go up or down from your last overall rating by the same amount.
The good news is there's a lot of options to review where you stand
Free: TLS, tennisrecord websites
Paid: schmke, TennisLeagueAnalytics
Only the USTA really knows whether you will be bumped but its fun to look at. I like tennisrecord of all the options since it provides the most detail for free.
Free: TLS, tennisrecord websites
Paid: schmke, TennisLeagueAnalytics
Only the USTA really knows whether you will be bumped but its fun to look at. I like tennisrecord of all the options since it provides the most detail for free.
Seems like the formula works here as well (Average of 4.38, 4.39, 4.43 and 4.53 is approximately the overall rating of 4.43)Crawling up at a snail's pace!
J
J011yroger
Talk Tennis Guru
Similar threads
