Platform Tennis Index

What is PTI?

Platform Tennis Index (PTI) is a performance rating developed by the APTA. PTI is an objective data driven measure of rated match performance. Using this rating, league administrators and tournament directors are able to organize competitive play amongst players of similar performance levels.

PTI 2 Description & FAQs

Why are we transitioning to a new PTI algo?

Thanks to the work of Jake Karetsky, Demian Johnston and the APTA’s PTI Algo Committee,  the APTA now has a rating algorithm that has been proven mathematically to be more accurate than our already quite accurate algo (PTI).In addition, the new algorithm requires virtually no selective manual intervention and hones in on accurate ratings for players far more quickly than the previous algorithm.

The new PTI algo (which we will continue to call “PTI”, but for the purposes of this document we will refer to as “PTI2”) is an even more accurate predictor of match outcomes than the original PTI

What are the main differences between the two algos?

• PTI2 rewards winning sets first and foremost vs PTI which rewarded winning matches and % of games won
  • With PTI2, for example, if you win in straight sets, your rating cannot get worse, regardless of set score or rating differential
  • Conversely, significant underdogs who win a set may improve their rating for that, whereas in the past they did not
  • Percent of games won still impacts ratings adjustments, but only in the magnitude of the rating change. Winning a set 7-6 will not result in the same rating change as 6-0.
• PTI2 weights all sets and matches the same; it eliminates the match scaling we had with  PTI where main draw tournament matches count more… or casual matches count less
• Unlike with PTI, the math for PTI2 is complex. While it will be very difficult for most of us to follow the exact details of the algo, we are creating a calculator that allows you to plug in player PTIs and a match score to see how the algo adjusts. Calculator link
• PTI2 almost entirely eliminates the need for manual adjustments:
  • PTI2 automatically adjusts a new player’s Start Rating after a handful of matches if a certain threshold of PTI movement is hit
  • PTI2 has a much more dynamic and complex version of PTI’s original confidence factor… which eliminates any need for manual adjustments of established players
• With PTI2, ratings moves after each match may not be the same for every player on the court.  
  • This is due to a factor that PTI2 accounts for where PTI1 did not:  volatility of performance by each individual player (tentatively called Sigma at this time).  The higher the volatility of a player’s performances, the larger their PTI2 move will be for any given match.  This factor is assumed to be relatively high at the beginning of a player’s recorded career which is why PTI2 is able to move much more quickly for over-performing or under-performing new players.  Over the course of future matches, this volatility factor will lower to an equilibrium point around which it will tend to oscillate.  However, this equilibrium point is different for each player.  Thus, the amount one player’s PTI2 moves due to a given match can be different from the moves that the other players experience.
  • PTI2 will not allow for “Confidence Date” adjustments.  Instead, the volatility factor (Sigma) is dynamic.  It is sensitive to significant changes in player performance and will automatically rise when it measures a statistically significant change in volatility of performance, thus allowing their PTI2 to move more than previously.

Why are players’ PTI2s different than PTI?

 Short answer:  the math is different, so everything is different.  They are close because PTI1 wasn’t terrible, PTI2 is better.  

Longer answer: While the APTA has taken steps to have PTI2 rating exist in the same range (most players being between 0 and 80), it is actually a fundamentally different rating system.

It is reassuring that PTI2 and PTI1 track so similarly, it implies that both systems do a reasonable job of rating the vast majority of players.  In fact most players will find that their PTI2 rating number is very close to their PTI1 number.

However, PTI2 has been tested to be statistically better than PTI1.  

• PTI2 correctly predicts the winners of matches more often than PTI1.
• PTI2 better handles rating new players
• PTI2 more quickly responds to a sudden change in player performance 
• PTI2 has eliminated match type scaling (tournaments and casual matches) and the possibility for a team to have a worse rating despite winning in straight sets.
• PTI2 shifts from game orientation to set orientation for consistency.
• PTI2 more consistently handles single set play (for additional play formats)

As a result of this underlying mathematical change, it is unlikely that anyone will have an identical rating to the end of last season. It is our belief that the new updated ratings are, in fact, better than before.

Of course, it is very important to reiterate that PTI2 (just like PTI1) is NOT precise to the digits displayed.  Almost every PTI is meant to be in the middle of a range of performance that we players deliver match to match.  

If you are rated 52 and I’m rated 54.  It’s more accurate to say that you typically play in the 49-55 range and I typically play in the 51-57 range.  We are not statistically different beyond your good days are likely better than mine and my bad days likely worse than yours.

Overall are PTI2s higher or lower than PTIs?

We tried to keep the overall scale of PTI2s fairly similar to the original PTI. Initially, we had communicated that players below 50 were more likely to see their PTI2s rise and players above 50 would see their PTI2s fall. We have since modified the scaling of PTI2s so that on average players don’t veer too far from their original PTI. 

That said, everyone’s PTI2 (with some rare exceptions) will be different than their PTI as the algo is different and it now includes an automated start rating adjuster for new players.

Where (and why) will we likely see the biggest changes in PTI2 vs PTI?

The players whose PTI2s differ the most from PTI will likely fall in the following categories:

• Newer players without extensive match history
  • PTI2 adjusts people quicker than PTI
• Players whose start rating was off
  • PTI2 has a built in start rating adjuster that will do a much better and more consistent job finding and correcting players with inaccurate start ratings
• Players who have gone on long winning or losing streaks, especially straight set streaks
  • PTI2 counts all matches (none are skipped); this results in players seeing more movement in their PTI2s than they did under PTI

Where do we expect to see “outliers” under PTI2 (ie; where are the edge cases where players may become temporarily mis-rated under PTI2)?

Unlike with PTI, in PTI2 there are no “skipped” matches. While heavy favorites can’t improve their PTI2 meaningfully by beating big underdogs… they also can’t ever see their PTI2 get worse if they win in straight sets.

The reverse of that is also true. Big underdogs cannot improve their PTI2 just by playing close sets vs heavy favorites. They need to win a set to improve their PTI2

Therefore, it’s possible that there will be edge cases where players for a variety of reasons find themselves in divisions or on teams or spots in their line where they are always expected to win or lose. If/when that happens, we will likely see those players become overrated or underrated UNTIL they get to a division/team/line where they are more appropriately matched (at which point their PTI2 will quickly (much quicker than in PTI) adjust back to a more accurate level

In the event a player is believed to be mis-rated, what is the process for having that player adjusted? 

While we expect a dramatic reduction in the need to adjust any players’ ratings, there will be the occasional “outlier” (see above) where an adjustment may be necessary. The APTA is centralizing that rating adjustment process; leagues will no longer be making any adjustments locally. The APTA’s Algo Committee has created an email address,  pti@platformtennis.org, that League heads can use to alert the Algo Committee about potential outlier ratings. (please note that the only lever for adjusting a rating under PTI2 is the start rating; the confidence factor which existed in the original PTI algo is no longer an adjustment option as it is built into the new PTI2 algo and functions in a much more dynamic way… one of the most significant enhancements of the new algo)

How do third set superbreakers (in lieu of the third set being played out) get treated by PTI2?

A third set superbreaker should be scored the same way it has been… 1-0 for the team that wins the tiebreaker. PTI2 will treat that set (regardless of the actual tiebreak score… 10-0, 10-8, etc,) as if the set score were 7-6.  

Note: To reiterate, game score really only has an impact on the magnitude of “adjustment” made for winning/losing a set.   This answer reflects that winning or losing a superbreaker is NOT a decisive result but rather a very close result.   As such, a superbreaker has as low an impact as a 7-6  

Why does PTI2 eliminate the match scaling we had with PTI where main draw tournament matches count more… or casual matches count less?

Match type scaling wasn’t necessary to produce a more predictively accurate system.   ‘Predictive Accuracy’ is our primary method to determine if one system is better or worse than another (What percentage of matches does the better rated team win). PTI2 is “better” / more accurate than PTI without having matches treated/scaled differently.  In fact, back testing of PTI2 revealed that scaling casual and tournament matches the same as regular matches produces more predictive accuracy than if they were scaled differently.

Where can I see my PTI2?

PTI2s will be displayed the same way they have been displayed in the past

• Many leagues have a Ratings tab on their league site that allows players to see detailed match and ratings history for everyone in the league
• Some leagues have opted to hide/suppress the ratings tab. For those leagues, a player can still see their own PTI by creating a profile on the site and logging in (PTI will be listed under the players “profile” link/page). Players will not be able to view anyone else’s PTI on that league site.

I have played most of my matches with the same partner; why did their new PTI improve while mine got worse (or didn’t improve by as much)?

Short answer:  the difference will be in how you performed in the matches where you didn’t play together

Long answer:  for two players with a fair amount of history, your ratings will move up and down together when you play together. They will only move independently when you play apart.

For the new PTIs, we have completely replaced the algorithm behind the ratings system in such a way that the calculations end up in the same number range.  More simply, it's brand new math that mostly agrees with the old math.  This is great news because it provides further confidence that our old algorithm was pretty good, and that we have been able to improve on it.

The PTI2 algorithm is more sensitive to changes in match performance than our older algorithm and that accounts for the different handling of historical matches.   We're confident that this approach results in a statistically significant improvement in Predictive Accuracy of the system (% of matches in which the lower rated team wins).

If you are/remain confident that your rating relative to your partner’s is off, the easiest way to remedy that is to play rated matches with a different partner… or even better, play (and enter a casual match score) against your regular partner. 

How do I enter a casual match score? 

Video: Entering a Casual Match score

Why are PTIs not updating immediately when I post a score? 

For the time being, PTIs are not updating “live” (as soon as scores are entered). PTIs will update overnight. So if your match score from last night’s league matches weren’t posted til the day after (or if a score has been edited), PTIs won’t update til the next morning

Who is in charge of the PTI algorithm?

The APTA PTI algorithm committee is made up of 9 people from 7 different leagues across the country. This committee monitors how the algorithm is working and reviews suggestions for tweaking the algorithm during the off-season.  For more information or if you have any questions about PTI, please contact the PTI committee at pti@platformtennis.org.