Line Matching

alright here we are: - will be a website this year, not a google sheet - on-site chat instead of discord - same shit as every other season with projections, ranks, etc. - same price as last year ($60/mth, $300/season) - 15% discount on the first month on patreon til Oct 10th

Player Comparison: Connor Bedard vs. Sebastian Aho

Expected Home Goalie to Watch: Pyotr Kochetkov

Home Team Player to Watch: Sebastian Aho

Best Bets: CHI @ CAR

Game Preview: CHI @ CAR

added a small thing on the betting page

Bets: PIT now a bet NJD gap closed from 2.4% to 5.1% (still no bet) DAL gap improved to 5.2% (still no bet) MTL/UTA leans UTA now (leaned MTL) but no bet

Varlamov −0.9 Blackwood −0.5 Swayman, Soderblom and Dobes −0.4 Hellebuyck +0.6 Bussi and Cossa +0.4 Merilainen and Hart +0.3

Biggest drops (DK points): Thomas −0.7, Holloway −0.6, Snuggerud −0.5, McTavish −0.5, McMichael −0.5 Suzuki −0.7, Slafkovsky −0.6 Scheifele −0.6, Vilardi −0.5, Connor −0.4 Nugent-Hopkins −0.5, Raymond −0.4

Biggest gains (DK points): Kucherov +1.2, Point +1.1, Guentzel +1.1, Hagel +0.8 Jarvis +1.0, Aho +0.9, Svechnikov +0.9, Blake +0.7, Gostisbehere +0.7, Ehlers +0.6 Josi +0.9, Makar +0.7, Drysdale +0.6 Forsberg +0.6, Michkov +0.6

Teams with the biggest change in expected goals: CAR 3.17 → 3.48 TBL 4.33 → 4.54 PIT 3.36 → 3.55 NSH 2.76 → 2.92 COL 3.80 → 3.93 STL 2.78 → 2.55 MTL 3.22 → 3.08 UTA 3.20 → 3.31

based on those model changes2121, here are some of the impacts for tonight (not including any impacted by line changes)

6) finer bet sizing just using .05 steps in units, not a major change but increases ROI by 1%

5) large sim adjustment to how correlated a game is DFS result: fixed stacking in the simulator (wasn't stacking properly relative to actual contests), gap between actual correlations is down to 0.02 goalies save accuracy improves for DFS & also props

4) my optimizer templates (LM Cash/GPP/Short Slate) got adjusted slightly in how they handled ownership DFS result: LM Cash improves in 18/63 lineups, gets worse in 0/63 lineups; cash rate improves by 12.6% LM GPP cash rate improves by 2.8% no change to top-1% or win% on either

3) D-men blocks get curved betting result: no impact either way DFS result: error improves on all D projections by .11% and on blocks specifically by .29% 3+ block bonus is now at 19.2% projected (from 17%) vs. 19.3% actual

2) goalie projections are split slightly to account for goalies getting pulled. no prop history so bet impact doesn't exist prior to 2026, but log loss improved by .0044 this season and proj vs. actual improved in 13 of 15 seasons DFS result: 35+ save bonus more accurate

1) team ratings adapt faster than before. betting result on 23-26 seasons: +32.5u at 1/8 kelly (+1.5% ROI increase) log loss improved by .00104 overall, by .004 this season DFS result: skater error improved in every season (marginal)

made some back-end tweaks to the projection & betting model again, which i'll (vaguely) break down in this thread. getting pushed in a minute. all changes are made from OUT of sample games so it's not taking something it learned from a game to re-apply it to that same game:

someone on twitter the other day was arguing that because McDavid plays against top pairs/shutdown lines, that lets his teammates get softer minutes and that doesn't ever get quantified. it does, but it gets quantified in RAPM/Impact ratings, so here it is split out.