0.0%
Target ATS
4 seasons, out-of-sample
0.0%
Trailing ATS
adjusted for real spreads
0
Models
deep learning + statistical
0+
Games Validated
across 4 seasons
Four college basketball seasons (2022-26), tested on data the models never saw. Trailing ATS adjusted for real sportsbook line differences. The target pattern misses ~30% of the time — expect losing streaks within any given month.
Chalk compares its model prediction against the live 2H market line. Every element is a decision input.
Kansas trailing by 9 at halftime. Trailing regime — historically the highest-edge window.
Model projects Kansas to outscore by 1.2 in 2H. Market expects 5.5. A 4.3-point disagreement.
BEARISH — model is less optimistic on Kansas than market. Large divergence flags a potential opportunity.
Calibrated range on the second-half outcome. No false precision — the uncertainty is part of the signal.
Big 12 game. Power4 conference games have the strongest historical trailing edge.
Kansas came back to win 74-72. The 2H swing was 11 points. Home covered the spread.
The same game, 11 minutes into the second half. Chalk is now a live terminal tracking your position against the spread.
Early in the half, every game looks the same. Around the 9-minute mark, winners separate from losers — that's when hedging and hold decisions actually matter.
Toggle alerts on Chalk's published edges or your own custom strategies. SMS fires within seconds of halftime — no app to check, no dashboard to refresh.
CBB Chalk Edge: ARIZ @ KU
trailing, BEARISH
HT: 42-33
chalk.is/game/8050...
Other tools give you picks. Chalk gives you a research lab. Slice four seasons of college basketball by conference, regime, and margin range — then prove your thesis before you risk a dollar.
487 matching games
Top 20%
63.2%
ATS
54.1–72.3%
+18.4%
ROI
97 games
Top 40%
58.7%
ATS
51.4–65.9%
+11.2%
ROI
195 games
Top 60%
55.1%
ATS
49–61.2%
+4.8%
ROI
292 games
HMAC-signed webhooks at halftime. Discord and Telegram bots in minutes. Live position polling with Greeks. Automated P&L tracking — the full pipeline from signal to settlement, one API key.
Halftime signal hits your endpoint in under a second.
# Incoming webhook payload (POST to your endpoint)# Headers:# X-Chalk-Signature: sha256=9a4f2e...# X-Chalk-Timestamp: 1711843200{ "event": "signal.matched", "data": { "game_id": "401820795", "away_team": "Duke", "home_team": "Kansas", "halftime_margin": 9, "regime": "trailing", "signal": "BEARISH", "model_implied_2h": -1.2, "market_2h_line": -5.5, "divergence": 4.3, "filter_name": "Power4 Trailing Bearish" }}7-day free trial on every tier. Cancel anytime.
Three things: the gap between Chalk's prediction and the live second-half market line, the regime — trailing, close, or leading — which tells you where the historical edge is strongest, and a confidence range on the second-half outcome. Every element is a decision input.
Eight models working together — deep learning trained on play-by-play data from 60,000+ games, plus statistical models that capture how scores evolve through a game. They specialize in different things: momentum, volatility, comeback dynamics. The output isn't a single number — it's a full probability range on the second-half margin, so you see the confidence behind every signal.
Within seconds of halftime. Chalk runs all eight models against the live 2H line the moment it's posted — regime, divergence, and confidence range appear on your dashboard instantly. SMS alerts fire immediately. During the second half, Cover Dynamics updates in real time as every basket changes your position.
We studied every target pattern game across four seasons. Before 9 minutes into the second half, eventual winners and losers look identical — both are noisy. Around 9 minutes, trajectory plus current position start to separate outcomes dramatically. That's the inflection point where hedge-or-hold decisions actually matter.
Every result is tested on data the models never trained on — no peeking. We adjust for the difference between exchange lines and what you'd actually get at a retail sportsbook. The target pattern is deliberately narrow — specific conferences, specific game situations — not a broad claim that works everywhere. It misses ~30% of the time. We publish the methodology and the loss rate because the edge is real enough to survive scrutiny.
The same framework as options pricing, applied to spread positions. Theta measures time decay on your cover margin. Delta measures score sensitivity. Gamma tracks how fast Delta changes. Vega quantifies your volatility exposure. Together they answer: how is my position aging, and what does the next basket do to it? Available on the Syndicate tier.
Billing pauses automatically after late June. No charges July through September. Resumes in October with email reminders and one-click cancel.
College basketball and NBA. Second-half spreads at halftime — that's where the gap between model and market is widest and the historical edge is strongest. Chalk is purpose-built for this one market and fits into your existing workflow for line shopping and bankroll management.