Fade The Chalk
Live site: https://kyle-vo.github.io/fade-the-chalk/ — Today's board, Moneyline, past days with results, and the Track scorecard.
Contrarian MLB home run / NFL anytime touchdown board. Two layers per player:
- Model % - what the numbers say.
- MLB: regressed season HR/PA x hitter platoon split x starter HR/BF (platoon-aware, damped)
x bullpen HR rate for the late PAs x park factor x weather (temp, wind in/out) x run environment
x a tiny hot-hand nudge, over expected plate appearances for the lineup slot.
- NFL: spread + total -> implied team points -> expected offensive TDs -> player's regressed
share of team TDs (2025 rates, positional priors, roster-normalized, injuries removed).
- Heat - how obvious / over-bet the name is today (leaderboard rank, hot streak, narrative park,
bad pitcher, primetime, big favorite). Type a real public-bet % into the board and it replaces heat.
Verdicts: SLEEPER (edge, nobody on him), VALUE, TRAP (crowd on him, no edge), FADE
(public 60%+ and negative edge), CHALK (hot name, no price entered yet).
Run it
.\run.ps1 # today's MLB slate + current NFL week, opens output\board.html
.\run.ps1 2026-09-12 # specific MLB date (run again after lineups post, ~2-4h before first pitch)
.\run.ps1 -Sport nfl # refresh one sport only
Then open output\board.html, expand "Paste the book's odds board", paste lines like
Kyle Schwarber +165 62% (odds, then optional public %) and hit Apply. Edge, verdicts and the
nasty score recompute live. Entered numbers are remembered in the browser.
Files: fetch_data.py (public MLB StatsAPI + ESPN feeds -> data/), model.py (-> output/board.json),
build_html.py (-> output/board.html), serve.py (optional local preview on :8765).
Requires Python 3 with requests.
Knobs worth touching
model.py PARK - HR park factors (100 = neutral). Update yearly.model.py heat formulas - what "the crowd is on him" means to you.build_html.py verdict thresholds (edge >= 4 pts, heat < 45, etc.) and the nasty score.
