NEW PERK: 2021-23 WAR Cards (Patreon)

The season has begun so it's time to get the 2022-23 roll-out going. First up is a visualization that should be familiar to those who subscribed last year.

To access this viz:

• If your last name begins with A-E, click here 
• If your last name begins with F-L, click here 
• If your last name begins with M-R, click here 
• If your last name begins with S-Z, click here 

The way this model (and most other WAR and RAPM models) works, it needs a lot of data input into it for the outputs to be meaningful. What makes WAR effective is that it processes a huge amount of information about linemates, opponents, deployment, etc. to work out a player's impact in various areas of the game. When the sample you use is too small, the model is basically flying blind.

To account for this while still providing stats from the young new season, Patrick and I landed on a compromise: using a combined sample of last season and this new season. This isn't just adding last season's numbers to this season's numbers, but adding the new season's data into the pool of data that the WAR model is built on. In fact, because there are so many new combinations of players, the results become a lot stronger.

These cards compare the new results (as we call them, 2021-23) with the ones built solely on last season. They are programmed to update every morning and will automatically integrate players as they reach the magic 200 TOI threshold.

Next up is to get everything else moving: the Stats Central, the Tableau Viz, and of course the player cards updated with these numbers. I'm working full time (in an unrelated field) for the full season this year, but I'm doing my best to keep up with things.

Cheers,

Jack