There's always thoughts rattling around in my head, and now being fairly deep into the fall season, it feels like a good time to introduce the TWISTR blog. There won't be a standard update cadence or anything to these, just more of a place for me to share some thoughts related to the ratings as they happen.
I've been thinking a lot about the TWISTR volleyball ratings recently, the heels of yeserday's Hamilton Southeastern–Roncalli contest feels as good a time as any.
When I was initially developing the TWISTR ratings algorithm, I struggled for a while to figure out a good way to integrate volleyball. Not only is volleyball fundamentally scored differently from other sports, but the scoring information that is available is less consistent and less useful than in other sports.
Take basketball, for an example. When we see that Gerstmeyer beat Garfield 75-65, we can safely assume that the game was pretty competitive. Similarly, when we see Gerstmeyer post a 75-30 win over Wiley, it's pretty safe to assume not only that there was a wide gap between the teams, but also that Garfield (who only lost 75-65 to Gerstmeyer) is probably much better than Wiley.[1]
In volleyball, however, both a 25-23, 25-20, 25-22 (75-65, functionally) Gerstmeyer win over Garfield and a 25-8, 25-10, 25-12 (75-30) one over Wiley will simply get reported in the data as a 3-0 win for Gerstmeyer. Like in the basketball example, these two contests had a completely different tenor to them, but in volleyball, they're going to come across as similar results to the model. The issue is two-fold — (1) set-by-set volleyball score data is difficult to access, and (2) set-by-set score data is not consistently reported to begin with.
In an ideal world, the TWISTR ratings would probably consider each set as its own contest and rate teams based on the number of points we'd expect one team to be favored by in an average set. This would allow us to say that Gerstmeyer is a little better than Garfield and much better than Wiley, consistent with the score-based approach used in other sports. Why do we prefer a score-based approach? It's almost always going to do a better job of predicting future success than ratings that don't consider margin of victory; against the same schedule, a 9-1 team that has nine blowout wins and one tight loss is probably better than a 10-0 team with all close wins.
The TWISTR ratings as they're constructed for volleyball really could be thought of a rating system that only considers whether a team wins or loses, just on a set-by-set basis. It's not the ideal, and it's kind of out of line with the TWISTR ratings for other sports, but until there's a consistent way to extract more detailed scoring data, it's what we have.
I've staged the TWISTR data personally for a couple of years now — posting them to a barely publicized Google Drive spreadsheet — and despite the aforementioned limitations, the model has still done a great job of identifying the top teams in each class.
A year ago, three of the four state champions finished at No. 1 in their class in the season-ending TWISTR ratings, and the fourth — 1A champion Fort Wayne Blackhawk — finished No. 2 in 1A after defeating No. 1 Tecumseh in five sets to win the title. 3A runners-up Bellmont and 4A runners-up Yorktown each finished No. 2 in their respective classes, meaning seven of the eight state finalists finished at No. 1 or No. 2 in their class (Linton-Stockton, who upset No. 3 Indianapolis Scecina — the top-rated 2A south team — at semi-state, was the lone exception).
TWISTR performance was strong in 2021 as well — in 1A, the model's No. 2 team Lafayette Central Catholic beat No. 1 Trinity Lutheran in five sets; 2A was won by No. 1 Andrean in five over No. 2 Western Boone, and 3A saw No. 1 Brebeuf Jesuit go the distance to beat No. 3 Bellmont. While 4A champion McCutcheon was ranked at just No. 7 in the class, the Mavericks were one of the 12 4A contenders rated within a "set" of each other in a year with a lot of parity at the top.
Returning to 2023, we see Hamilton Southeastern as a clear No. 1 team in the ratings — the Royals are a perfect 18-0 and have dropped just two sets all season despite facing TWISTR's top-rated schedule to date. Those sets were dropped in 3-1 wins over No. 3 Indianapolis Cathedral and No. 8 Yorktown, and they've swept No. 2 Bellmont, No. 4 Indianapolis Roncalli and No. 6 Providence. All signs point to this Southeastern team being generationally great and thankfully, the TWISTR ratings agree, with the gap between No. 1 HSE and No. 2 Bellmont being larger than the gap between No. 2 Bellmont and No. 10 Zionsville.
So, are the TWISTR ratings for volleyball exactly what I'd want? Not really — but in the meantime, we see Hamilton Southeastern sitting pretty at the top, which is more than enough to pass the sanity test.