Sunday, February 3, 2019

Introducing a new report - Predicting flight winners/standings with USTA League regular season simulations

I haven't written in a couple weeks and that is because I'm been busy cooking up a new report!

Last year the USTA introduced a new format for Nationals that led me to come up with a new way of simulating the results from flight play and identifying the most likely record for each team, and the chances of all possible records.  I used this to predict who would advance to the semis with the new format with a fair amount of success.

Now, I'm offering reports with the same sort of simulation for regular season flights.  The report will show the most likely record for each team and thus a predicted flight winner and chances of being top-2 where the top-2 advance on to playoffs, but also show the other records each team ended up with in the simulation and the chance of that occurring.

I can do these simulations at any time, pre-season or during the season, the latter taking into account the current record and changes in player's ratings.  The simulations are done by taking a representative strength for each team from a top-N average from my Estimated Dynamic NTRP Ratings, and then simulating the full schedule a million times with a random variation in how a team performs relatively to their representative strength.

Naturally, who specifically plays and the match-ups go a long way to determining team wins and upsets can occur as well, and some teams with new unrated players are hard to predict, but the simulations give a very good idea of how things will end up.

Here is an example from one of the 4.0 sub-flights in the 40 & Over league here in Seattle last year showing the final record compared to the pre-season predicted record.

TeamActualPredicted
PSC-Coe8-18-1
CP-Katz7-28-1
BC-Antezana6-34-5
BELL-Caldwell5-46-3
BTA-Smed5-47-2
FC-Basha4-54-5
EDG-Namba4-55-4
MC-ODB Blasters3-61-8
IND-Nomads2-72-7
GC-Murray1-80-9

We can see the simulation predicted the correct record for 3 of the teams and was just one off on 4 of the others, and no more than 2 off for any team.  Given all the variables involved, I'll take being within one game on 70% of the teams and off no more than two games as good prediction accuracy.

If you are interested in getting a flight simulation report, contact me and I'll get you more details and pricing.  Note that this is a new report so I'm offering it at a very low introductory price so get one now!

And if you are just wanting to scout your flight, I continue to offer Flight Reports, pre-season or in-season as well, that give you summary information on each team including overall and top-8 averages and for teams that have played matches, played averages by court.

Let me know what you think!

No comments:

Post a Comment