Wednesday, February 12, 2014

It is only February, but the 3 strike DQs have begun in USTA League

In USTA League play, new players self-rate at a certain level based on answering a series of questions.  When answered correctly, this usually results in the player beginning play at the proper level.  However, sometimes, whether deliberate or not, a player self-rates too low.

To try to avoid uncompetitive and unfair situations where a self-rated player is playing at too low a level, the USTA has a 3-strike DQ process where if a player's dynamic rating exceeds a level specific threshold 3 times, they are disqualified at that level and promoted to the next.

I recently learned about an unexpected DQ after just 3 matches and was asked to generate an Estimated Dynamic NTRP Rating Report for the player.  The report confirmed that my estimated ratings agreed that 3 strikes had occurred.

While there very may have been other DQs already in 2014, this was the first I did a report on.  Since the DQ was in just 3 matches, the chart is kind of boring so I won't show it, but each match rating was above the strike threshold resulting in each dynamic rating being over as well.

I'm often asked why strikes occur and the simple answer is that the player has demonstrated that they are playing well above level.  In this case, the player played with a lower rated partner and beat some higher rated opponents.  The specifics varied a bit from match to match, but the consistent theme was that he carried the weaker partner to wins and that is a recipe for getting strikes.

As is usually the case, there are fears and concerns that the player won't be able to compete at the promoted level.  This usually isn't the case, although the player likely won't win most of their matches at the higher level.  This doesn't mean they aren't now at the right level though.

Players like to win and if they have a losing record for a season they think they should get bumped down.  But what many like to forget is that for every winner there is a loser and losing a match to a strong opponent may be expected and the loss doesn't mean their rating should go down.  Not all losses hurt a rating, a loss could actually raise your rating, and similarly not all wins raise a rating.

So if you get DQd or bumped up, relish the opportunity to raise your game and embrace the challenge.  It is the DQ or bump wasn't appropriate, your rating will come down and you'll move down to the appropriate level, but given enough matches and as long as players aren't trying to manipulate ratings, they work for 99% of players.

Tuesday, February 11, 2014

More interesting USTA League stats - Comparing the 40 & over and 18 & over divisions

In 2013, the USTA introduced new age divisions in the Adult leagues, a new 40 & over league to go along with the existing 18 & over league, and the prior 50 & over league became a 55 & over league.  In effect, the 18 & over league was split into two leagues, one that was the same as before and another only for those 40 or over.

I don't know the exact driver for it but it was probably some combination of an an interest by middle aged USTA members for more league play (and the corresponding opportunity to make more money through league fees) and a response to complaints from older "under 50s" about having to play the young guns in league play, or perhaps even young players complaining about having to play the old guys.  Regardless, depending on your section and district and how it was implemented, there have been those that love the new division and those that aren't so enamored with it.

A recent thread on Talk Tennis started in support of the 40 & over division elicited some good discussion about the merits of it and also if it was resulting in a 40 & over 4.5 no longer being the same as an 18 & over 4.5.  Some say that an 18 & over player will generally be more athletic and have more stamina and will have an advantage over the same rated 40 & over player.  Others say that the algorithm looks at scores and if the 40 & over players are indeed losing, their ratings will go down accordingly.

I fall in the latter camp, that a 4.5 is a 4.5 regardless of age, and if the older 4.5 can't hack it, they will ultimately get bumped down to 4.0.  Note that this hypothesis requires that there be some cross-play between the 40 & over crowd and the younger players for these adjustments to be made, but there are a large number of the older players that continue to play in the 18 & over league so I believe this is the case.

Now, if I'm right, what about the perception that the over 40 players don't do well when they face the younger players in the 18 & over division?

My first response is that this perception may not be accurate and may be skewed by a handful of results that people focus on from playoffs at the local, district, and national level where some young player crushes and older player 1 & 1 and thus not representative of comparing the play of different ages.  It is probably more an indication of the under self-rating and sandbagging that is more likely to take place in the 18 & over division and an indication that the young player should be rated higher.

But another important point is that when you compare two players that are at the same level, they could be at opposite ends of the half-point range and thus if the younger player is say a 4.45 and the older player is a 4.05, the result is expected to be a fairly easy win by the younger player, but both players can be correctly rated as 4.5s.  In fact, given age does catch up with all of us, it is probably reasonable to hypothesize that the distribution of actual ratings for 4.5 rated players is a bit lower for the over 40 players than for the under 40 players.

To test this hypothesis, I took a look at the data I have to estimate Dynamic NTRP ratings to see if any trends stuck out.

First, I took a look at the average dynamic rating I calculated for all players that ended the 2013 year rated a 4.5 and found it to be 4.23.  This makes sense, you'd expect the average to be somewhere around the middle of the range, and the fact that it is a little lower than the exact mid-point just means the distribution of ratings is a little biased towards the lower half, also something you'd expect at a fairly high level like 4.5.

Next, I changed the players I looked at to only those that played in the 40 & over division.  The average dropped to 4.21.  Two hundredths isn't a huge difference, but it is significant enough that I think it supports the hypothesis.  Note also that the 4.23 above includes these 40 & overs, so if they were to be taken out, the average for the under 40s would be a bit higher and gap would be more than two hundredths.

So I would contend that a 40 & over 4.5 is not any less a 4.5 than a 22 year old 4.5, but they very well may have a slightly lower actual rating and that is the reason for the perception that the old guys of the same NTRP level can't keep up with the young guys.  It is just the nature of a rating system where a given level has a range of 0.5 rating points that this will happen, particularly at a higher NTRP level.


Sunday, February 9, 2014

USTA League Points Per Position being trialed in Middle States

I wrote back in December about a possible Points Per Position system the USTA sent a survey out about.  In summary, in an effort to combat stacking courts/lines, the lower numbered courts would be worth more which would encourage captains to play their strongest players on those courts and not stack the higher numbered courts with stronger players.  It was just a survey, but we already see it being implemented at least on a trial basis.

It appears that at least the 18 & over 4.5 flight in the Delaware district of the Middle States section is going to employ some variant of the proposed new scoring.  From what was shared by a player from that district on Talk Tennis, the scoring will be 3 points for court 1 singles and doubles, 2 points for court 2 singles and doubles, and 1 point for court 3 doubles.  This means that a total of 11 points are available per team match.

Using this scoring system, a team would be able to get the team win by winning just court 1 singles and doubles (6 of the 11 points), meaning that the team match could be won without winning the majority of the courts.  This means that a team could have just 3 strong players play on these courts, weak players on every other court, and win their league.

The net result is that while stacking is now discouraged, sand bagging and getting players to self-rate too low so a team can get just 3 ringers is encouraged more than it was before.  And in my opinion, this is a bigger issue than stacking.

As I wrote before, whether this works or not was going to depend on the points/weights for each court, and the 3/2/1 system probably doesn't work the best.  A suggestion from another Talk Tennis poster is to go with a 5/4/3 system.  This would mean there are 21 points available, and winning the two court 1's would only get to 10 points, not enough to win the team match.  A team would still be required to win 3 courts to win the team match.

So what would a 5/4/3 system accomplish then?  The accumulated points for the season would be used as a first or second tie-breaker.  This would mean that a team that wins matches 3-2 by winning the lower numbered courts would win the tie-breaker over a team that wins matches 3-2 by "stacking" and winning the higher numbered courts.  Thus stacking is discouraged as it could hurt a team come tie-breaker time.

Having just the tie-breaker affected may not be perceived as doing enough to discourage stacking though.  If that is the case, I'd offer my other earlier suggestion of not allowing players playing up to play on a lower numbered court than a player at level.  For example, a 4.0 team with a 3.5 playing up would not be allowed to play the 3.5 on any court other than 3 doubles or 2 singles if they were the only 3.5.  This would eliminate a very common stacking scenario where a player playing up is sacrificed on a lower numbered court.

It will be interesting to see how the trial works in Middle States.  What do you think?

Friday, February 7, 2014

Some interesting USTA League NTRP stats from the USTA

In a February newsletter, the USTA had an article on the NTRP system and included a number of statistics.  Given my interest in these stats and my past blogs about them, I found some of them interesting.

Over 240,000 players were issued ratings, nearly 84% stayed at the same level. Of the 16% that moved, 10% went up, 6% went down.

Self-rated players had 60% stay at their self-rate level, 18% went up, 22% went down.

The self-rated stat is interesting with more being bumped down than up. 18% going up is still a large number mind you so there is certainly still a lot of self-raters that under-rate or improve as they play and get bumped up.

And the overall stat is interesting with nearly twice as many players being bumped up than bumped down.  I'd like to see this stat from past years to see if this is a consistent trend or not.

Tuesday, February 4, 2014

Estimated Dynamic NTRP Ratings on Social Media

Since you are reading this, you already have found me on social media one way or another.  But I thought'd I'd post a summary of the sites you can find me on in addition to the blog.

You can follow Schmidt Computer Ratings on Twitter at @computerratings.  You'll get updates when new entries are posted to the blog plus that is a good way to interact with me and ask questions.

You can like NTRP Ratings on facebook.  This will also get updates as new blog entries are written and is a great way to share what I post with your friends or teammates on facebook.

I write the blog using Google's Blogger and so as you might imagine there is a Google+ page.

And while it may be old school now, if you use an RSS reader, you can get the blog updates using the RSS feed.

So pick your poison.  I'd love to hear from you.