high epa teams getting penalized for teaming up
Sophia De Medeiros
this year, the points tend to level off around 230-250 and scoring above this mark is very difficult. for top teams with 100+ epas, statbotics is expecting them to have 290-310 total points per match. however, this doesn't really happen due to how difficult it is to score more points after the ~250 threshold. when two top teams play together (like 1323 and 2910) and they don't score 311 points or whatever, their epa drops. however, for a team like 2056 which is far stronger than the other teams in their alliance, their epa climbs because statbotics isn't overestimating them while splitting the total points scored between another top team. maybe there could be some sort of function that corrects for this if the game is one where the points tend to level off at some point and recognizes that the total match score is probably not going to be team 1 epa + team 2 epa + team 3 epa if the teams are very strong.
Bob Croucher
I noticed this as well. A score cap is common in some games with limited time, space and game pieces. Statbotics has historical season max scores available to detect this phenomena. The max score ever without penalties this year was done by the above noted alliance twice (301 points). So, the predicted score could use this cap, and account for it in the ELO expectations. Statbotic's predictions began with that alliance's playoff run predictions as high as 328 points. By the end, as their EPA's fell, the expectation was still a lofty 320 points. This permitted 2056 to pass them both in highest EPA for the season. Perhaps that's okay, but it doesn't really reflect their ability to score when playing with other less capable alliances.