A bit more on the recent WTN algorithm change / adjustment

I wrote a few days ago about the recent significant change to WTN ratings that the USTA sent a notice out about, and many players saw their WTN get larger (worse) overnight, some in the high single digits, but others went up 10 or more.

I don't have any detailed analysis yet, but do have a few more observations.

The USTA published an FAQ about the changes that is worth a quick read, but it had a few interesting explanations.

First, some insight into why the change was made:

These enhancements to the ITF WTN algorithm will reposition cohorts of players on the current scale of 40-1 and ensure that players from all over the world are more accurately aligned, particularly with regard to different age groups. While ratings are likely to be adjusted, players will likely see movement relative to their age group.

So, they seem to have concluded that different age groups of players were rated correctly relative to each other.

Going on, they get more specific:

Adult players aged between 19-29 should expect only minor changes to their ITF World Tennis Number. Larger changes begin to occur for players over 30 years old that regularly play against players of a similar age. Adult players over 30 will see their numbers move down the scale (towards 40), this is to create some balance with Junior players and to ensure full use of the 1-40 scale.

This is consistent with my anecdotal observations that USTA League players around me, for the most part age 30+, have seen their WTN get larger and closer to 40.  I haven't looked at any juniors, but believe their statement to be true.

And from what I see, USTA League players are by and large are all in the mid-20s and higher.  I see 4.5 and 4.0 men in the mid-20s and the majority of league players are at 3.0-4.0, so that means the majority of league players are being squeezed into the 25-40 range.  That is quite a bit less granularity than we had before (15-40) so we will see how that works out.

Regarding juniors, the FAQ states:

Junior players aged 10 and under will most likely experience a movement down the scale, (towards 40) with players who are aged 17 and 18 moving up the scale (towards 1). The player ratings are now more of an accurate reflection of age, with a general incremental increase expected from aged 10 upwards.

I've also heard from players in the UK and US looking at comparisons with players in Europe that prior to this change, there seemed to be a disconnect in that what appeared to be similarly skilled players from Europe and the US had significantly different WTN ratings, those in the US having lower numbers, so this adjustment seems to have addressed that, at least somewhat.

All of this highlights the challenge of a single algorithm that tries to put all tennis players in the world on the same scale.  Whether it is age groups or countries or other geographic grouping, you will inevitably have islands, or cohorts as the FAQ says, that all play each other, but have little to no play with other islands or cohorts.  When this happens, the algorithm has very little data to go on to reasonably ensure players from those different groups are appropriately rated relative to each other.

This is probably the biggest challenge for the WTN, and as such one would have hoped a lot of effort would have gone into reviewing this and getting it right prior to launch.  If adult USTA League players were rated in the teens and similar recreational players from Europe were in the 20s, that should have been one of the first things that was reviewed and stuck out.  Similarly if all juniors regardless of age were clumped together, or juniors relative to adults seemed off, this could have been observed far earlier and address before WTN was launched.

There are ways an algorithm can address this, notably by looking at the matches that are played between the islands of players, and giving them more weight in determining the relative ratings for the group of players they represent.

This is effectively what the USTA does with NTRP, using the matches from Nationals to "connect" the different sections, and have those results and how well a section did trickle down to players from the section.  If a section does very well at Nationals, that means relatively speaking the players from that section at that level were rated too low relative to other sections, so player's ratings may go up and more bump ups will occur.  It can go the other way too, and I've pointed out the effect of this in many of my year-end rating analysis'.

I'm guessing the WTN doesn't inherently do this as part of the algorithm, as it seemingly required this significant adjustment to compensate for what they were observing.  Hopefully they still applied the principals of leveling across cohorts and it wasn't entirely manual.

Another challenge the WTN has, and one that doesn't seem to have been addressed yet and wasn't mentioned in the FAQ is gender neutrality.  The algorithm claims to be gender neutral, i.e. just like a WTN 25 male should have a competitive match with another WTN 25 male, either should also have a competitive match with a WTN 25 female.

It is generally accepted that NTRP which is not gender neutral has about a one level (0.5 NTRP) difference between the genders, e.g. a 3.5 male and 4.0 female will be similarly skilled, but looking at WTN ratings one will regularly see men and women of the same NTRP level with similar WTN ratings or even the women having lower (better) WTN ratings.  Similarly, you will see men with higher (better) NTRP levels having higher (worse) WTN ratings.

This means WTN really can't be used to get cross-gender competitive matches today.  And WTN is really handcuffed now when it gets to trying to solve this as it now has to fit both men and women into this compressed 25-40 range I noted above.

For example, looking at men's WTN ratings, leaving room for the 2.5 and 3.0 women to have higher WTN ratings than 3.0 men, and leaving some room for complete beginners not even playing matches yet, the WTN range all recreational men have to fit into is more like 25-35 at best, and perhaps even less.

The change made to WTN may very well make it better than before.  But the challenges of truly being a world tennis number are still there.  It will be interesting to see how it continues to mature.