Thursday, December 13, 2018

The International Tennis Number - What is it and how does it compare with NTRP?

Those of us that play USTA League have our ratings life centered around NTRP, the rating system used by the USTA for level based league and tournament play.  The NTRP system works well for the most part and results in competitive matches between players of similar ability.

But there are other ratings systems out there.  One from the International Tennis Federation (ITF) is the International Tennis Number (ITN).  From their Web-site:
The International Tennis Number (ITN) is a tennis rating that represents a player’s general level of play and is recognised internationally
This sounds very similar to NTRP, and from what I've read and can tell, the goals are basically the same, to have a way for players of similar ability to play more tennis and grow the game.  So what is different about the ITN?

The big thing is the scale and direction.

With NTRP, player ability increases as the rating gets larger, e.g. a 3.5 is better than a 3.0, and a 4.5 better than a 4.0, and the scale more or less goes from 2.5 to 7.0 in half point increments.

The ITN on the other hand is on a 1 to 10 scale without half point designations, but in reverse order.  A beginner is a 10, and a world class player is a 1.  Moving "up" is actually moving to a smaller numbered ITN.  This gives you 10 levels which is more or less equivalent to the 10 levels the NTRP gives you from 2.5 to 7.0.  In fact, the ITF has a conversion chart that pretty much just maps the levels one to one in opposite order.

So in the end, either system should result in a rating that lets you identify and find other players of similar ability.  But how they go about calculating the rating is entirely different.

Those that regularly read my blog know that I calculate Estimated Dynamic NTRP Ratings and do reports for players wanting to know how their rating is improving or how their results affect their rating.  I've done a little bit (an understatement) of research on how the algorithm works and have written an FAQ with a lot of good information and answers, but the short form is that the algorithm looks at the current dynamic rating of all players in the match to arrive at an expected score, and then compares this with the actual score to determine how to adjust each player's rating.  Do better than expected, your rating goes up, do worse than expected, and it goes down.  It is important to note that the score does matter, not just winning or losing as game differential is considered by the algorithm.

The ITN on the other hand documents a calculation that is more focused just on wins and losses and the level of your opponent.  For example, it calls out that a win against a same level opponent yields 1 point, a win against one level higher is 2 points, and so on.  Correspondingly, losses against at level opponents are -1, against one level down -2, etc.  Wins against lower level opponents or losses against higher level opponents are not given point values.

The ITF does not host a site or calculate a ratings for players, but outlines the above approach and recommends that ratings be updated at least yearly if not more often.  The update involves taking the player's current ITN and adjusting it up or down by the average points per match, and then rounding it.

For example, if a player played only at-level players and won 75% of their matches, they would get 0.75 points per match and regardless what level they were at, would move to the next better level.  For example purposes say a player is an ITN 7, subtracting the 0.75 would be 6.25 and rounding that would be a 6 and they would move "up" a level to their new ITN of 6.  The number of matches played and frequency of updates will obviously be a factor in how much a player can or does move.

For the most part, the two different systems track each other, but because they are calculated differently there are some deviations where players may end up with an NTRP level and ITN that are not equivalent per the conversion chart.  This doesn't mean that either one is right or wrong, they just use different criteria and so end up at a different result.  NTRP focuses on scores, ITN focuses on wins.  NTRP uses the dynamic rating of players in a match, ITN just looks at the ITN level of the players.

Where is the ITN used?  It has not really been adopted in the United States to date, but perhaps that will change.  The conversion chart linked above shows that Great Britain's system is very similar on a 1 to 10 scale but with added point levels, and from what I hear ITN is used in other European countries to some degree.

The example calculation/algorithm the ITF documents is I think intentionally simple as it was written over 10 years ago and they wanted organizations and national federations to have a low technical bar to reach to implement it.  I can think of a number of ways to improve the simple calculation and have done so as a fun comparison in the past.  If there is interest, I will perhaps publish some ratings lists using ITN or include an ITN in the reports I do.

But what do you think?  Is there a place for the ITN?  Should there be an international system rather than every country/organization having their own?  And should a rating system consider scores and actual current ratings?  Or should the focus be on wins regardless of score and just look at the level of the players?

Please leave comments here or on Facebook, or e-mail me with your thoughts.

Wednesday, December 12, 2018

USTA Adult League Participation by Section 2013 thru 2018 - A few sections not declining, but most are

I wrote about the general Adult league participation trend last week, and noted the continued decline, but the trend is not the same across all of the sections.  It seemed that looking at the trend by section would make sense, so here goes.

The chart below is the largest I've made, but I didn't want to create 17 separate ones for each section, so you get it all in one.  The chart shows the participation in the Adult leagues from 2013 thru 2018.

One thing really stands out, and that is that Southern has far more participation than any other section, which makes sense given it cover so many more states than any other section.  But seeing in a chart really makes the point.

But Southern is also the section that is showing the greatest decline from over 81K to just over 71K in the 5 years this analysis covers.  But nearly every section has declined, a few exceptions being Southern Cal, up over a thousand over the period, Pacific Northwest up several hundred, but down a bit from 2017 to 2018.  Southwest was largely flat, and a few others had small declines, but most have dropped several hundred to thousands over the period.

What do you think?  Are you seeing participation trends like the above in your area?

Friday, December 7, 2018

What if NTRP ratings were gender neutral? A hypothetical NTRP level distribution

As part of my year-end stats series of posts, I looked at year-end distribution by level, overall and by men and women, focusing on the change from 2017 to 2018.

What I didn't show in a single chart was how the men's and women's counts compare in the same chart.  Here that is.

We can see that more women play than men, and women make up well over half the population at a level until you get to the 4.0 level, when the men get close to 50% and at 4.5 there are more men than women.  No big surprises here, the most players are at the 3.5 level by a significant margin, then 4.0, then 3.0.

A sometimes discussed topic is how men and women of the same level compare, or what a gender neutral league might look like.  Yes, there is Mixed league today, but that isn't really gender neutral as male/female pairs play each other, you never have a man playing a woman in singles or unbalanced doubles.

The USTA does not include any play between the genders for Adult/C ratings, so there is nothing in the NTRP system to tell us that a 3.5 man and 3.5 woman are or aren't the same.  But it is generally understood that a male of a level will on average be stronger than the female of the same level.

A common hypothesis is that a male player of the same level is about one level (0.5 NTRP rating) higher than the equivalent female, e.g. a 3.5 male and 4.0 female are probably of similar ability.  The reality is that that isn't true at every level, at least it isn't an exact 0.5 difference, and certainly there are exceptions, but it is a rough rule of thumb one can use, and one exact level makes my analysis below easy, so we'll go with it.

So we can imagine for a minute, what the distribution by level might be if we had gender neutral ratings by simply adding 0.5 to all the male players, e.g. shifting the blue segments of the chart above down a column.  Rather than making you do the mental exercise, here is that chart.

By moving the 3.0 men to 3.5, and 3.5 men to 4.0, etc. we see a change in the distribution.  There are now nearly as many "4.0s" as "3.5s", and nearly as many "4.5s" as "3.0s".  We also see that women would out number the men through "4.0" and only at "4.5" would there be more men than women.

If we had gender neutral ratings, then gender neutral leagues could more easily be played and it appears that the "4.0" level would be very competitive with nearly equal men and women being eligible for the level.  All the others would have a significant bias towards more women (3.5 and below) or men (4.5 and above).

I don't think any of the above is going to happen any time soon.  NTRP is not suddenly going gender neutral nor are gender neutral leagues in the offing, at least to my knowledge, so all of this is just a hypothetical what if.  But seeing the distribution above, would you be interested or willing to play in a gender neutral league?  Or do you like things the way they are now with Mixed the only league where men and women play and then only as pairs against each other?

Thursday, December 6, 2018

Analyzing 2018 USTA NTRP year-end ratings - bump rates by league/division played

A common question I get is whether a 55 year-old 4.0 is the same as a 25 year-old 4.0.  The short answer is yes, they are supposed to be.  But there is a lot behind that answer.

I say yes because the NTRP system is based on match results.  If a 55 year-old plays a 25 year-old, the result of that match will be factored in to each player's rating and the relative ratings between them should be accurate.  If the 25 year-old 4.0 is better than the 55 year-old 4.0 and wins easily, the 25 year-old's rating will go up, the 55 year-old's will go down.  If either goes up or down enough, they could be bumped up or down at year-end.

Correspondingly, anyone that plays either of those players will have their rating calculated appropriately.  If the 55 year-old is a "low" 4.0, then wins against that player won't rate as high as a win against the "25 year-old "high" 4.0.

Now, for this to work, there must be some amount of play between the older/younger players so the groups are "connected" so to speak.  This generally happens enough, a fair number of 55+ players do play 18 & Over, or the 55+ players will play 40 & Over and there will be 40+ players they play there that play 18 & Over.

But there still seems to be a perception amongst some that the ratings don't get it right, and some 55+ players have inflated ratings and when they play against a 25 year-old of the same level, it isn't a competitive match.

Now, there is a lot more to a match than just comparing ratings, one player may be a "high" at-level player and another "low", so a 6-2,6-2 win for the 25 year-old may be expected.  Similarly, if they play singles and the 55+ player has achieved their rating from primarily doubles but isn't as strong a singles player, that may explain it.  And playing styles and how those match up is always a factor.

Generally, you might expect players ratings to gradually drop as they age and so the older players may be in the lower part of the range for their level.  But at the same time, if they are going down and just got bumped down, they may be in the higher part of the range for their level..

All of this begs the question, what are the bump rates among players of different ages?  Unfortunately I don't have player's ages, but we can use the league/division they play in as a proxy for age.

Below are charts that show the bump up/down rates for players based on the division they played in.  Note that players aged 40+ and 55+ can play in the lower age divisions so they may be counted potentially in each division.

First, for the women.

We do see that players that play 18 & Over are bumped up a lot more, nearly 2.5 times more, than down.  The 40 & Over division is similar but closer to 2 times more.  But the 55 & Over has a significant drop and there are slightly more bump downs than up.

This is largely as you'd expect given I noted above that older players will eventually see their skills or mobility diminish, and you'd expect younger players to have a higher ceiling to improve into.  It is perhaps surprising those in 40 & Over are still bumped up over twice as much as down, but that is likely due to some players picking up the game in their 40's and having room to improve.

Next, for the men.

We see a very similar trend.  So similar I don't really have much to add.

But the above is across all levels and that is where you may see new players at any age still improving along with those on the decline.  What about if we look at the higher end of the levels, say the 4.5 level for the men?

Interestingly, we see more bumps down than up in every division, although the ratio definitely gets larger as you move up in age.  It is especially interesting and surprising for at least me, that even for those age 55+ that there are more than 2% that play in that league bumped up.  If we are to believe the ratings, more than 2% of 55+ players are still improving, and improving enough to cross the threshold to be a 5.0!

I'm certainly not going to say that older players can't improve, but improving from a 4.5 to a 5.0 is a pretty tall task and 1 in 40 4.5 rated players age 55+ doing it seems a pretty big ask.

One might ask if the USTA is inflating or keeping some older player's ratings too high?  I know I've heard from some older players that have been bumped up saying there is no way they can play against younger players at 5.0, it almost becomes a safety issue, and there are few or limited playing opportunities for a 55+ 5.0 player.  The result is they can't play much and are stranded, or have to somehow find a way to play younger players and get killed (hopefully not physically!) to get their rating adjusted and get bumped down.

I don't know that the USTA does this, but I could see some reasoning being that if you don't bump players up, or let them be bumped down, it is unfair to the lower rated at-level players to have stronger ones beating up on them, and you end up with a glut of players with too large a range at a given level.  One solution to avoid the glut at a level would be to bump more of those lower rated within players down, but since everyone has at least a bit of an ego, those bumped down might not like the idea of that or having to play at the lower level won't be fun for them and decide to quit playing USTA League.

What do you think?  Are older player's ratings inflated a bit?  Is there not enough "connectedness" between the age groups for the system to work as hoped?  Or is the USTA just keeping players rated at higher levels to avoid a glut of 55+ players at the 3.5 and 4.0 levels that don't want to be bumped down?

Update: For completeness, here is the same chart as above for the 4.5 women.

Analyzing 2018 USTA NTRP year-end ratings - 3.0 men bump rates by section

Continuing with my analysis of year-end ratings and bump rates, here are the bump rates for the 3.0 men by section.

At approaching 60% of 3.0 men bumped up, Caribbean again leads.  This is a huge number of bump ups, I don't know I've ever seen such a high rate at the 3.0 or above level.  So we again do the chart without Caribbean.

There are some high bump rates here too, higher than the 3.0 women in fact, most sections around/over 15%.  Hawaii is approaching 30% though which is quite high, and New England, SoCal, Florida, and Texas are around/over 20% too.  There are still some bump downs, but very few.

More to come.

Analyzing 2018 USTA NTRP year-end ratings - 3.0 women bump rates by section

Continuing with my analysis of year-end ratings and bump rates, here are the bump rates for the 3.0 women by section.

Caribbean again leads, and is well over 45% at this level, but if anywhere, the lower levels are where you'd expect high bump rates.  Regardless, we again do the chart without Caribbean.

There are a lot of higher bump up rates here.  Every section other than Middle States and Eastern are over 10%, and Northern and Southwest well over 15%.  There are still a not insignificant percentage bumped down though, most sections around 3-4%.

More to come.

Tuesday, December 4, 2018

Analyzing 2018 USTA NTRP year-end ratings - 4.5 men bump rates by section

Continuing with my analysis of year-end ratings and bump rates, here are the bump rates for the 4.5 men by section.

Caribbean again leads, at just over 45% at this level which is surprising given that as the levels goes higher the trend typically switches from more bumps up to more bumps down.  So we again do the chart without Caribbean.

Here we see the largest percentage bump goes to Hawaii 4.5s being bumped down at over 13%.  They still had over 3% of their 4.5s bumped up though.  Middle States, Midwest, and New England also had a lot of bump downs, so look for them to perhaps have strong 4.0 teams next year.

The largest percentage of 4.5s bumped up goes to SoCal, and almost more up than down.  Mid-Atlantic, Missouri Valley, and NorCal were close behind.

More to come.