TLS Website

Discussion in 'Adult League & Tournament Talk' started by dblsonly, Mar 21, 2013.

  1. dblsonly

    dblsonly New User

    Joined:
    Jun 20, 2012
    Messages:
    14
    Hello,

    Has anyone here looked at the website www.tennisleaguestats.com? It appears to show you what your USTA rating is to the hundredths percentage point. It links to your USTA results for the 2012.

    I ask because 3 of my friends were bumped to 4.5 a year ago. Their TLS rating shows as 4.02, 4.07 and 4.14. They have tried to computer appeal but were not successful. Two of them have not had very good results at 4.5. Can someone who is a 4.02 (barely a 4.5) do a written appeal? On what grounds could it be granted?

    Thanks for your responses!
     
    #1
  2. dblsonly

    dblsonly New User

    Joined:
    Jun 20, 2012
    Messages:
    14
    Another note

    One other thing, my TLS rating is a 4.13 but USTA says I am a 4.0. There are a just few others on the Wisconsin list that have a TLS rating higher than their USTA rating. I am wondering how I can be a 4.0, when my friends have a lower TLS rating but are 4.5's. Confusing...
     
    #2
  3. tennis_tater

    tennis_tater Semi-Pro

    Joined:
    Oct 22, 2010
    Messages:
    456
    I've never seen the TLS website before and since there ratings aren't out for my section, I can't really give my thoughts on their ratings for players I know.

    As far as your first post, maybe your friends can't appeal because they are bench marked?

    And if they got bumped to 4.5, then I would think the ratings provided by TLS would have to be considered clearly wrong.

    As far as teh post above, since you are rated as a 4.0, my understanding of that is that your dynamic NTRP # is then somewhere between 4.00 and 4.49.So you being a 4.0 rated player with a 4.13 rating would be accurate. Once you accumulate enough wins to get over 4.5 rating, you will be a 4.5 until your dynamic rating hurdles the 5.0 mark.
     
    #3
  4. goober

    goober Legend

    Joined:
    Jun 9, 2004
    Messages:
    8,491
    That just shows that TLS are estimates. Supposedly if you are .05 away from the lower level an auto appeal is granted. The 4.02 guy is actually higher than 4.05, but he he likely is very close.
     
    #4
  5. beernutz

    beernutz Hall of Fame

    Joined:
    Aug 16, 2005
    Messages:
    4,428
    Location:
    expanding my Ignore List
    I am pretty sure a 4.0 year end rating means you had a dynamic rating of between 3.50 and 3.99.

    His having a 4.13 dynamic rating as calculated by this website and a 4.0 year end rating could be because the dynamic rating is just an estimate and isn't accurate. Or even if that dynamic rating estimate is accurate, if he is looking at his dynamic rating for 2013, just because he has a 4.13 now doesn't mean he had one at the end of 2012 when the year end NTRP calculation was run by the USTA.
     
    #5
  6. dblsonly

    dblsonly New User

    Joined:
    Jun 20, 2012
    Messages:
    14
    Rating

    Actually, according to this website, you are considered a 4.0 when your rating is between 3.51 and 4.00.
     
    #6
  7. AR15

    AR15 Professional

    Joined:
    Mar 3, 2006
    Messages:
    1,125
    Location:
    Gulf Coast, USA
    The website said ratings weren't available for the Southern Section.
     
    #7
  8. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    I believe the USTA has posted before that the cut points are 3.50 - 3.99 (4.0) and then 4.0 - 4.49 (4.5). Not that the 0.1 really makes a difference when you are talking about estimated ratings.

    I thought it was interesting and I looked through the list for folks I knew. Most of the ratings seem reasonable ... there is a guy (who is better than me) that the USTA still claims is a 4.0, but the LTSA site had him rated higher at 4.5 ... which is more accurate as far as I am concerned.

    The more interesting question is who has time for this nonsense and how do they plan to make money from it.
     
    #8
  9. beernutz

    beernutz Hall of Fame

    Joined:
    Aug 16, 2005
    Messages:
    4,428
    Location:
    expanding my Ignore List
    Yes, I was off by one-hundredth of a point at each end. I believe the rest of my previous explanation still applies though.
     
    #9
  10. Angle Queen

    Angle Queen Professional

    Joined:
    Jul 27, 2010
    Messages:
    839
    Location:
    On the deuce side, looking to come in
    Any clue to what the blue/red numbers mean? I can't find a legend on the site.

    Interesting, though.

    EDIT to add: I think "red" is an unexpectedly "higher" rating (than the analysis or actual would seem to suggest) while "blue" is the opposite...an unexpected "lower" rating (than analysis or published numbers).

    Other analysis to follow....
     
    Last edited: Mar 21, 2013
    #10
  11. Angle Queen

    Angle Queen Professional

    Joined:
    Jul 27, 2010
    Messages:
    839
    Location:
    On the deuce side, looking to come in
    I have run a Ladies Combo team for the past two years. And it's fairly common knowledge that to succeed in that Division, you need top-of-the-NTRP folks at both NTRP levels. That is, for say a 6.5 team, you'd want high 3.5s playing with high 3.0s. Reality is, though, it was easier to get those at the high-end of the lower NTRP to play...and hope you still had some good (if not necessarily high-end) upper NTRP to go with them.

    Seems I did a nice job of putting together my team (and the actual results bore it out...we finished a very close third out of 10). My higher leveled NTRP ladies were middle-of-the-pack (or better!)...and my lower ones...sometimes got the actual bump or are certainly on the verge of it for this year.

    And, as I look through many of the names at the levels I'm most familiar with, the numbers confirm what I've guessed at. Some of the ones I've scratched my head at on why they're still where they are or why they were bumped up/down...showed just how close they are to the cutoffs.

    Still...very, very interesting stuff. Bottom line, though, wouldn't really change a thing for me as a player or captain...if I knew NTRP down to the hundredth.
     
    Last edited: Mar 21, 2013
    #11
  12. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    AQ .... while I might not change a thing as a captain if this site were to update their ratings periodically (say every month) I think I would check back often. However, I am not sure that would be good for my psyche ...

    Interesting to note is that I have a different rating for each of districts I play in ... and the break out does not include sectionals ... so obviously this ranking is missing some of the equalization and benchmarking that the USTA does between sections.
     
    #12
  13. Angle Queen

    Angle Queen Professional

    Joined:
    Jul 27, 2010
    Messages:
    839
    Location:
    On the deuce side, looking to come in
    ^^ Yeah, Dizz. I'm with ya on that. And perhaps, maybe it's better just with "after the fact" analysis. I wouldn't want to be tempted to waste so much time looking at numbers that tell me what I've already figured out (even though the confirmation is, well, nice).

    And here's the ding to MY psyche...the site still has me at 3.5 (and close enough to probably be in the appeal zone...which I've also "guessed" at). Maybe I should? Nah...I'm playin' where I'm playin. Done.
     
    Last edited: Mar 21, 2013
    #13
  14. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    ^^^ what I think is funny is that according to this site

    We have a guy that is decent player but did not make our starting 8 in playoffs .... he did not get the bump.... but according to this site he had a better year end rating than the 5 guys from our team that were bumped.
     
    #14
  15. J_R_B

    J_R_B Hall of Fame

    Joined:
    Jan 27, 2010
    Messages:
    2,668
    Location:
    Newtown, PA
    An auto appeal will not be granted if you played in 6 or more matches regardless of how close you are. I don't know if this applies in this case or not, but it is possible that the person in question could have a 4.02 rating and not be granted an appeal.
     
    #15
  16. J_R_B

    J_R_B Hall of Fame

    Joined:
    Jan 27, 2010
    Messages:
    2,668
    Location:
    Newtown, PA
    I just checked this website. One of my players (who was 3-2 at 4.0 for the season and currently 4.0 C) was given a TLS rating of 4.46, or almost a 5.0. LOL. This is somewhat interesting, but a loose approximation at best.
     
    #16
  17. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    While this is very interesting and gives me some idea of how I would react if the USTA did this ... it is obviously is not balls on accurate.

    However, I will say that the ones which appear to me to be wildly off are the players that only played a few matches. It seems that the players who played 8, 9, or 10 + ,matches are all pretty close to where I think they should be.
     
    #17
  18. McLovin

    McLovin Hall of Fame

    Joined:
    Jun 2, 2007
    Messages:
    3,250
    Another thing to consider is that these ratings are broken down by Area, so the number you find next to your name doesn't tell the whole story.

    For example, a friend of mine plays in multiple leagues in multiple areas. For one area, his TLS rating is 4.43, which would have him as a 4.5. But in another league, he's a 4.73, which puts him at 5.0. The 3rd league doesn't have a rating since it was mixed/combo only.

    He was bumped to 5.0 this past fall.

    Also, according to the site, I should have been bumped to a 5.0 (4.65 TLS rating), but I wasn't. Maybe it was that I only had 5 matches in? Either way, I'm fine as I'd get my butt kicked at 5.0...
     
    #18
  19. tennismonkey

    tennismonkey Semi-Pro

    Joined:
    Jun 18, 2010
    Messages:
    741
    +1. i noticed the same thing. i also noticed that my rating in virginia was a good deal higher than my rating in maryland. which corresponds to my observation over the years that virginia is a tougher league.
     
    #19
  20. J_R_B

    J_R_B Hall of Fame

    Joined:
    Jan 27, 2010
    Messages:
    2,668
    Location:
    Newtown, PA
    There are actually 3 significant problems with the calculation:

    1. Unknown Formula. Obviously, this is just an approximation of the actual dynamic NTRP formula since the real formula is not public.
    2. Multiple Ratings. As someone mentioned, the calculation is done by district, so if you play in multiple districts (which is common here), you get multiple ratings. I am 3.93 in NJ (9-1 - only loss to Mr mcwizzl...), and 3.88 in PA (5-1).
    3. Post-Season. The calculation only appears to use regular season matches, so for players that played at districts/sectionals/nationals, the ratings are probably a little understated.
     
    #20
  21. cknobman

    cknobman Legend

    Joined:
    Sep 7, 2007
    Messages:
    6,209
    Location:
    Saudi Arabia
    For the players in my area I checked about 15 different people (guys on my team or guys I have known for years) and all of the ratings seemed to be very accurate.

    There were a few people that got a EOY bump to 4.0 that I know are not 4.0's and according to this website it shows their dynamic ratings are indeed below 3.5.
     
    #21
  22. schmke

    schmke Professional

    Joined:
    Jan 20, 2010
    Messages:
    1,371
    I did some spot checking of my estimates against the TLS ratings several months ago and found that mine were a bit more accurate. Here is what I found when comparing the numbers for someone:

    Player A - A 4.0 last year, TLS had him as a 4.04 that should have been bumped up to 4.5 but he wasn't. I had him as a 3.97 that should have stayed a 4.0 which he did.

    Player B - A 4.0 last year, TLS had him as a 3.21 that should have been bumped down to 3.5 but he wasn't. I had him as a 3.45 that should also have been bumped down but I was a lot closer to keeping him at 4.0 and in fact, at 3.45 he may have been within the USTA's threshold to stay up.

    Player C - A 4.5 last year, TLS had him as a 4.71 that should have been bumped up to a 5.0 and he was. I had him as a 4.60 that should also have been bumped up.

    So, one where we agreed and were both right (C), one where we disagreed and I was right (A), and one where we agreed and both were wrong (B), but I was closer to being right and perhaps within the USTA's tolerance so was perhaps right after all.

    Also, FWIW, you were a 4.0 last year and TLS had you at 3.54 so you should have stayed a 4.0 according to their rating, but I had you below 3.5 saying you should be bumped down to a 3.5 and you were. So I was more accurate here too.​

    In the end, both my and TLS' numbers are estimates and won't be exact, but I've found that mine are actually quite accurate.

    See http://computerratings.blogspot.com/search/label/tennis for the various reports I offer and can generate.
     
    #22
  23. tennismonkey

    tennismonkey Semi-Pro

    Joined:
    Jun 18, 2010
    Messages:
    741
    definitely an interesting site and yours too schmke.

    if you look at the ratings garnered for playing up, it seems that if you really want to get bumped up -- would just playing the higher league, almost regardless of wins, accomplish that?
     
    #23
  24. kylebarendrick

    kylebarendrick Professional

    Joined:
    Nov 16, 2006
    Messages:
    1,042
    Location:
    Northern California
    Regarding appeals... When they did the big bump up a few years ago, they also instituted a rule such that if you played more than 6 (or 6 or more, I can't remember) matches that counted towards your rating, then you would not be able to appeal. The logic was that this was enough matches that the computer rating should be considered accurate enough that an appeal mechanism was unwarranted.

    I don't know if it is still in place, but I thought it was a great idea. Before that rule, I saw many cases of people going 12-2 year after year, getting bumped up, appealing back down and doing it all over again.
     
    #24
  25. beernutz

    beernutz Hall of Fame

    Joined:
    Aug 16, 2005
    Messages:
    4,428
    Location:
    expanding my Ignore List
    I am going to speculate and say yes, playing up will much more likely result in your dynamic NTRP increasing because you can lose at the higher level and potentially your NTRP could still go up. OTOH if you lose at your current year end level, depending on the NTRP of your opponent, it seems much more likely you'll decrease your dynamic NTRP. If you win playing up, the increase would be more than what you would experience by winning at your current year end level. So the odds of your dynamic rating going up rise dramatically imo when you play up.

    I now await someone who is not speculating to actually answer the question.
     
    #25
  26. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    And don't you forget it brotha!
     
    #26
  27. goober

    goober Legend

    Joined:
    Jun 9, 2004
    Messages:
    8,491
    Where did you hear this info on 6 matches or more? I can say with absolute certainty that 2 players I know were granted auto-appeals both played well over 6 matches- probably close 12-15 matches each during the calendar year.
     
    #27
  28. goober

    goober Legend

    Joined:
    Jun 9, 2004
    Messages:
    8,491
    So yes you can be denied an autoappeal at 4.02 if you were benchmarked. There is nothing about 6 matches unless that is something they added recently
     
    #28
  29. J_R_B

    J_R_B Hall of Fame

    Joined:
    Jan 27, 2010
    Messages:
    2,668
    Location:
    Newtown, PA
    I ccan't find any mention of it in league rules, which leads me to believe it was something we were told at the captain's meeting (which means it may be local).
     
    #29
  30. kylebarendrick

    kylebarendrick Professional

    Joined:
    Nov 16, 2006
    Messages:
    1,042
    Location:
    Northern California
    Were those 6-12 matches all in leagues that count towards your rating? Was this recently?
     
    #30
  31. Topaz

    Topaz Legend

    Joined:
    Jun 18, 2005
    Messages:
    6,806
    I think the colors are changes in ratings...red is a bump up, blue is a bump down?
     
    #31
  32. Mongolmike

    Mongolmike Professional

    Joined:
    Jun 5, 2009
    Messages:
    996
    Location:
    Ohio
    I just noticed, and maybe others can confirm, but the TLS site for 2013 only shows league play... no tournament results. When I check myself in Tennislink, the 2 tourneys I played in 2013 show up, but TLS only has recent league results.

    I just double-checked 2012... same thing.

    I was under the assumption that stand-alone USTA "sanctioned" tournaments (not league connected) results influence your rating. I know some tourneys are not official USTA tourneys... but did I assume wrong? Are tournament results NOT counted in your rating tabulation?
     
    #32
  33. goober

    goober Legend

    Joined:
    Jun 9, 2004
    Messages:
    8,491
    USTA tourneys count toward year end ratings dependent on section. Some use it and others don't. It does not count toward dynamic ratings and does not affect DQ. Only NTRP tourneys count not open/age group (at least in my section)
     
    #33
  34. Mongolmike

    Mongolmike Professional

    Joined:
    Jun 5, 2009
    Messages:
    996
    Location:
    Ohio

    Oh, ok. Thank you. I don't think the TLS site uses them at all.
     
    #34
  35. Cindysphinx

    Cindysphinx G.O.A.T.

    Joined:
    Aug 31, 2006
    Messages:
    15,093
    What do you think would happen if I sent this link to my teammates?
     
    #35
  36. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    yea ... no one should think the TLS ratings have any basis in reality when in comes to the USTA rating. They are interesting, but that is all they are ... Interesting.

    Based on the comments from this board and my general observations ... the TLS website only used 2012 results, did not include playoff matches, and treated the same player in different sections differently. IE ... I Have a different rating in each of the three section I played in.

    In looking at their estimated ratings I would wager that their estimated rating only includes the won-lost games record of you and your opponents that occurred in 2012. There is no consideration of where you started the year and the rating of your opponent. The is also no year end equalization between sections.
     
    Last edited: Mar 26, 2013
    #36
  37. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    I think it is curious that in another thread some one already quoted this website in their bid to appeal a rating.

    I sent this site to my regular partner and he immediately replied back that he was upset his rating was lower than mine. He offered reasons why he thought this was so and immediately poo-pooed the site.

    I think it is just a snapshot of all the trouble the USTA would invite if they made something like this official.
     
    #37
  38. goober

    goober Legend

    Joined:
    Jun 9, 2004
    Messages:
    8,491
    I actually did. Some didn't like how low their actual ratings were compared to where they thought they should be or compared to team members that they thought they were better than. But most people didn't really care too much. We were only looking at it to see who was close to a bump and how close.
     
    #38
  39. Cindysphinx

    Cindysphinx G.O.A.T.

    Joined:
    Aug 31, 2006
    Messages:
    15,093
    I would be lying if I said I didn't immediately check out all of my friends to see where they stood in comparison to me.

    For those who were higher and IMO not much better than me, I consoled myself by noting that the site probably had me lower than USTA because it was unaware of my awesome post-season runs.

    I then sucked my thumb and pouted for a while, then noticed that my DC numbers gave me a boost and my NOVA day results were ignored.

    Feeling better, I then checked out women on the team who feel I don't appreciate their Mad Tennis Skillz. I concluded my view of relative strength on the team is correct.

    So. No good can come from this site.
     
    #39
  40. ian2

    ian2 Semi-Pro

    Joined:
    Nov 18, 2010
    Messages:
    409
    Clearly TLS ratings are not (cannot) be identical to USTA ratings, due to reasons other posters referred to (dizzlmcwizzl, in particular). But looking at the data for the players I know, these ratings seem quite reasonable. And I'd guess they correlate to USTA ratings fairly closely for people who mostly play league, not tournaments. But I wonder, does TLS account for opponents' ratings, or W/L only - and still arrive at a rating that is close to USTA (in most cases)?

    BTW, did anyone notice that if you click on a number in "Detail Ranking Order" column, it shows players' W/L (matches and games) for each league they played? Obviously the W/L info can be gathered on tennislink but having it summarized like that is nice.

    Interesting... and rather disappointing when I look at my own numbers :)
     
    #40
  41. Angle Queen

    Angle Queen Professional

    Joined:
    Jul 27, 2010
    Messages:
    839
    Location:
    On the deuce side, looking to come in
    I sent the link to a few like-minded friends, some who are teammates, others who are not. Don't know that I would ever sent it en masse...except if I was ready to start a mutiny. LOL.

    Most of us, after looking at it objectively, realized it for what it is: a really good approximation of one's USTA rating. Still, it did instigate some rather honest evaluation of the State of the Club (as I am a member of one).

    I think if USTA were to do such a more detailed analysis, it'd generate lots of said conversations....then we'd all go back to where we were. 'Cause that's where we, well, are.
     
    #41
  42. Mongolmike

    Mongolmike Professional

    Joined:
    Jun 5, 2009
    Messages:
    996
    Location:
    Ohio
    Madness?

    [​IMG]
     
    #42
  43. damazing

    damazing Rookie

    Joined:
    May 22, 2009
    Messages:
    252
    Using the numbers from this website and people's thoughts on ratings in general - what would the expected scores be for opponents that are .25 apart, .5 apart, .1 apart?

    I know I heard somewhere that the system sees a score of 6-3, 6-3 as competitive and a score of 6-2, 6-2 as not competitive.

    I was trying to see if my matches this year in doubles have helped or hurt my ratings based on the detail provided. I understand that its the number of total games won and not the result that the system looks at so here are my results and the combined ratings of me and my partners versus my opponents:

    My team 7.56 combined rating - 12 games won
    Opponents 7.71 combined rating - 4 games won

    My team 7.56 combined rating - 12 games won
    Opponents 7.50 combined rating - 4 games won

    My team 7.56 combined rating - 12 games won
    Opponents 7.76 combined rating - 4 games won

    My team 7.50 combined rating - 10 games won
    Opponents 8.15 combined rating - 9 games won
     
    #43
  44. schmke

    schmke Professional

    Joined:
    Jan 20, 2010
    Messages:
    1,371
    In general, it appears you are beating teams at or above your rating as listed on the site, and so your ratings would be going up. However, your, your partner's and your opponent's ratings change after each match played so you aren't really looking at the rating that the NTRP system is using for all but the first match, and perhaps not even the first if it wasn't your opponent's first match of the season.

    If you want a more accurate view of how your rating has changed match by match, contact me and I can probably generate an Estimated Dynamic NTRP report for you. See the examples on my blog at http://computerratings.blogspot.com/search/label/tennis or the example report at http://sites.google.com/site/computerratings/usta-tennis/example-report. I presently only post a full ratings list periodically for districts in the Pacific Northwest section, but can usually generate a report for most any USTA League player. Contact me at computerratings@techrunning.com if you are interested.
     
    #44
  45. damazing

    damazing Rookie

    Joined:
    May 22, 2009
    Messages:
    252
    Thanks for that - I forgot that it recalculates ratings after each match. I'll probably ask for a report, but it will be closer to the end of the adult season.
     
    #45
  46. goober

    goober Legend

    Joined:
    Jun 9, 2004
    Messages:
    8,491
    I know it is a free site, so you can't complain, but I wish they would update more frequently like once a month. Last update in my section was 2/9. Having exact "estimated" ratings is only helpful if they are relatively recent.
     
    #46
  47. TFC

    TFC New User

    Joined:
    Nov 5, 2013
    Messages:
    2

    I noticed the same thing, plus one other serious problem -- in my opinion. I was studying the TLS ratings (as of Sep 24) for Montgomery County, MD, where I play. I was interested in knowing my likelihood of being bumped up to 4.0 next year. So I specifically tried to analyze the populaton of men who began the year at 3.5.

    I noticed what seemed to me to be a very unusual distribution of TLS ratings for these players (I have no prior familiarity with the TLS system, but as an epidemiologist I am quite familiar with population distributions). Of 131 players with enough data to have a TLS rating, only 28 are at 3.30 or above, and only 9 are at 3.40 or above. Since 3.5 players are near the middle of the range for players who play in USTA, I would have expected the TLS ratings to be somewhat evenly distributed throughout the 3.5 range (in which case there would be about 26 between 3.40 and 3.49, but instead there are only 2). Only 9 players are above 3.38, which appears to be the dividing point for the clear "outlier" range.

    So I began to study the differences between the top players in the "normal range" (3.30 to 3.38 ) and the 9 players above that range. I looked especially at the top 15 in the "normal range" because I have played with or against most of them (and for those that I didn 't know I looked up their record). What characterizes these top 15 players in the "normal range" is that 14 of the 15 (including me) have played a lot of matches and have records that are very good this year and who I would expect to be among the ones most likely to move up.

    In marked contrast, among the 6 lowest rated players in the "outlier range" (3.46-3.53) , the number of matches is not only comparatively few (31 for all 6 combined), but each one of them have played ALL of their matches at the 4.0 level. And yet, their records in these matches are not very good, even considering that they are played at the 4.0 level. Their overall record is 4-27, and most of the losses were massacres. If not for noting that their TLS rating is near the top of the heap for those who started the year at 3.5, I would have said that few if any of them had much of a chance of getting bumped up this year. One of them I'm pretty sure is near the very bottom of the 3.5s in skill level. Another has a record of 0-2 this year, with only 6 games won in the two matches, and yet a TLS rating of 3.51. I can't see how any logical rating system could move these guys up to 4.0.

    So my conclusion is that the TLS system (and perhaps the NTRP rating system which it attempts to mimic) has a serious bias in evaluating the results of matches played between players with highly disparate ratings. And this conclusion makes sense from what else I know about the system. For example, I think that the idea of saying that a player playing against another player (or doubles team) withe a 0.5 higher rating is most likely to lose 0-6, 0-6 is highly unrealistic. The higher rated team is not likely to take the lower rated one very seriously, and in all probability will be playing far from peak level.

    So it seems to me that far too much weight is given to these types of matches, thus creating a situation where it is far easier to move up a level by playing a very small number of matches against far higher rated opponents than it is to move up by compiling a very strong record over many matches against opponents rated close to one's own rating.

    You seem to have given this rating system a lot of thought, and I am interested to know your opinion on this.
     
    #47
  48. schmke

    schmke Professional

    Joined:
    Jan 20, 2010
    Messages:
    1,371
    Good idea. Did you only play in Montgomery County? I've found many (most?) in the MD/DC/VA vicinity play in multiple areas and so the limitation of TLS doing separate ratings in each area is significant and may skew your analysis.

    I did a similar distribution analysis with my ratings. See http://computerratings.blogspot.com/2013/10/interesting-usta-league-stats-histogram.html. Mine did not seem to have any peaks/valleys that seem odd, which makes me think my algorithm is reasonable at least. And other analysis makes me believe they are pretty accurate too.

    E-mail me (computerratings@techrunning.com) who this is, TLS may be way off on this player, but my ratings have correctly explained other strange bump ups/DQs like this and I'd like to check what I have for the player.

    I don't believe the algorithm itself has a bias. Your observation of how it handles matches between disparately rated players is valid though. If a high rated player consistently plays players rated well below them and doesn't go out of their way to play their best, their rating will go down.

    I don't believe this is an accurate, at least the implication that you can't reasonably get bumped up by playing at level. Does playing up give greater opportunity to get bumped up? Sure, and if that is all you do and the better opponents gift you games, your rating may be higher than it should be, but this is probably not a common case.
     
    #48
  49. Orange

    Orange Rookie

    Joined:
    Sep 14, 2011
    Messages:
    209
    I found the statistics to be interesting but inaccurate in both directions.

    I looked up my rating and compared it to two players I have recently played in singles ladder matches. I recognize that these are individual match-ups, not composites as the ratings are, but the differences are stark.

    One woman is rated .1 below me (both on the same level). I beat her 6-0, 6-1 a few weeks ago and 6-1, 6-1 within the past week.

    Another woman is rated .51 below me. We have played 71 games in three different singles ladder matches within the last month. I won 36 and she won 35. There is no possible way that she is really .51 points below me.

    The website does provide a quick overview of matches of opponents and was very interesting to see.
     
    #49
  50. dizzlmcwizzl

    dizzlmcwizzl Hall of Fame

    Joined:
    Mar 24, 2010
    Messages:
    2,178
    Location:
    DE
    @TFC

    For years we have known and discussed at length that the surest way to get bumped up is to play up. The has long been suggested by both anecdotal and quantitative evidence.

    The information the USTA has published on their system and the data TLS provides also seem to support this result. In trying to perform my own mimicry of the algorithm I have also found my biggest bumps in rating occur when I am playing against the highest rated opponents. In short, if you are playing against someone much better than you, simply doing better than breadsticks will raise your rating. Getting bagels is expected and does not hurt much. If the worst result does not hurt your rating, and even modest results help ... playing up means you will move up.

    The one thing I suspect however, is TLS was initially skewed because they assumed everyone started at mid level last year. Meaning every 4.0 was thought to be a 3.75at the beginning of 2012. This year I believe their ratings to be more accurate because they were able to use last years ratings as a starting point. I suspect if they stick around in 2 or 3 more years their ratings will be very reliable for most veteran USTA players.
     
    #50

Share This Page