@thegrantedwards the results in the dashboard and google sheets link should be consistent now,
made an update to the model for better year to year comparisons
Results are now an ensemble of the player similarity based model and ML models to estimate tier probabilities. This improves PR-AUC across higher tiers to 0.37
The question of the pace of a pitch has bugged me for a long time. I wrote down an exploration of the pace of pitches, comparing different countries and discovering three kinds of surfaces. And yes, "fast" pitches really mean "bouncy".
Read more: https://t.co/IQpOKOetBb
Used ML based models to estimate a players probability of making the nba and then derived the other outcome probabilities using the similarity based approach.
Updated top-60 for the 2026 draft
https://t.co/gnKZlZNiqI
Used ML based models to estimate a players probability of making the nba and then derived the other outcome probabilities using the similarity based approach.
Updated top-60 for the 2026 draft
https://t.co/gnKZlZNiqI
Two new additions to the DARKO spreadsheet (https://t.co/mVkemwg9tS):
1) A pace RAPM estimate, which tells you how much impact each player is projected to have on a team's average length of possession, broken out by offense vs. defense.