Week 2
Fez_Keyreb@mingle/Yunyi
| Fez_Keyreb | mingle/Yunyi (home) |
|---|
| Win prob | 49% | 51% |
| Spread | mingle/Yunyi -1 |
| Total (O/U) | 155.5 |
home most-likely bans Earth/Space
Week 2
anurag@GidTheKid2
| anurag | GidTheKid2 (home) |
|---|
| Win prob | 50% | 50% |
| Spread | anurag -1 |
| Total (O/U) | 156.5 |
home most-likely bans Biology
Week 2
dan.k.memes@xpoes
| dan.k.memes | xpoes (home) |
|---|
| Win prob | 50% | 50% |
| Spread | dan.k.memes -1 |
| Total (O/U) | 155.5 |
home most-likely bans Biology
Week 2
jonathan@James W
| jonathan | James W (home) |
|---|
| Win prob | 49% | 51% |
| Spread | James W -1 |
| Total (O/U) | 156.5 |
home most-likely bans Physics
Week 2
Connor Chang@czz
| Connor Chang | czz (home) |
|---|
| Win prob | 51% | 49% |
| Spread | Connor Chang -1 |
| Total (O/U) | 160.5 |
home most-likely bans Chemistry
Week 2
ChessFun@cryo
| ChessFun | cryo (home) |
|---|
| Win prob | 52% | 48% |
| Spread | ChessFun -3 |
| Total (O/U) | 158.5 |
home most-likely bans Biology
Method
Post-Week-1 player ratings -> subject-resolved tossup-contest simulation -> home-court category ban -> Normal margin/total. Model-only (no market blend); the per-game markets are untraded platform seeds. Home removes one tossup of the category that most helps it (usually the away team's strongest); soft softmax over bans (lambda=1, a prior until actual bans are observed).
Caveats
- Rosters = full drafted roster; players without a Week-1 rating use their combine ability (cross-season proxy). A handful of unmapped/no-ability roster slots are still excluded.
- jonathan and Fez_Keyreb had the Week-1 bye, so they are still at the combine+draft seed.
- Home-court effect is modest here: the contest beta is still small, so removing one tossup moves the line only a little. Week 1 hinted home matters more; we won't size it up until we have actual ban data.
- Only 1 week of game data: ratings still lean on the combine+draft seed.
- Spreads and win-probs are widened by a confidence factor (gain 2.5x) so opening lines aren't pick'em; the raw model is more conservative and the AMM will correct. Totals are NOT widened.