01 The receipts
The single rule we won’t bend: every featured pick is written to our database the moment it fires and graded after the game ends. There is no “quietly delete the misses” option in our pipeline. The track-record card on the homepage reads from the same database that fires Telegram alerts to paid members — if it shows up there, it shows up here.
What this rules out. No retroactive deletions. No selectively-reported windows. No “lock of the day” we cherry-pick at the end of the week. If we sent a pick on Tuesday and lost, that loss is on the homepage Wednesday.
02 Sharp Signals
A Sharp Signal is a quantitative pattern in the betting markets — line movement, public-vs-sharp money split, cross-book disagreement — that historically preceded a winning bet. Each signal is a named detector running on live line-movement and public-money data. Every signal must pass three gates before it’s eligible to fire featured picks:
- Backtest passes: across the historical training pool, the signal hits at ≥ 55% win rate at ≥ +5% ROI with p < 0.05.
- Out-of-sample (OOS) holds up: when we subtract live-fired games from the backtest pool, the signal still passes. This catches signals that look great only on data they were tuned on.
- Edge-decay guard: live performance is monitored every day. If win rate or ROI drops materially below backtest, the signal is paused or demoted automatically.
Signals that fail any of the three are not featured. Signals that fail badly enough are removed from the roster. The model leaderboard shows the current roster with every model’s live-vs-backtest divergence visible.
03 Game Reports
A Game Report is a long-form, single-game write-up by a separate model that doesn’t look at line movement. Where Sharp Signals say “the market is screaming X,” Game Reports say “here’s how the matchup, recent form, and the specific lineup tonight argues for Y.”
The two streams are independent on purpose. When they agree, that’s a confluence pick (we mark it). When they disagree, we publish both views and let you decide. Our backtest of split games (n=22) showed neither side had a significant edge over the other, so we don’t bury one feed’s call to make the other look better.
Every Game Report carries one of six outcomes once the game is in: hit, miss, push, pass (the analyst recommended no bet), postponed, or pending (game hasn’t completed yet). Only hit and miss count toward the win rate. pass picks track separately so they don’t inflate the pending count.
04 ROI math
The per-signal ROI surfaced on each Sharp Signal card (Pro & Sharp tiers) and the aggregate ROI on the homepage track-record card (all visitors) are computed two ways depending on how much live data we have:
- Real-odds primary: when a signal has at least 30 picks with the captured American odds at fire time, the displayed ROI is computed pick-by-pick from those real prices. This is the honest number — if we got +130 on a winner, that pick contributes +1.30 units, not the +0.91 a -110 placeholder would imply.
- -110 fallback: below 30 captured-odds samples, single-pick variance dominates and a real-odds ROI would be misleading. We fall back to a -110-implied estimate derived from the signal’s win rate. Each card carries a label so you can tell which is which.
Aggregate stats on the homepage (the All-Time ROI pill, the Combined Track Record card) use the captured price on every pick where one was logged, with -110 used as a fallback only on the small fraction of rows where the price wasn’t captured. Pushes, voids, and split games (the two feeds on opposite sides of the same market) are excluded from win-rate math.
One known limitation: as of May 2026, only 2 of our 65 active signals have crossed the 30-captured-odds threshold. The other 63 still display backtest -110 estimates. Those numbers will become real-odds-derived as live data accumulates — usually within 30-60 days for a regularly-firing signal. We don’t hide which is which.
05 Sample size disclosure
Bigger sample sizes mean more reliable numbers. Our all-time track record (April 2026: ~520 graded picks since launch) is a meaningful sample, but it’s six weeks of one season. Across multiple seasons, with regime changes and bookmaker model updates, the picture could change. We don’t pretend otherwise.
Every aggregate stat we display includes the sample size that produced it. Every card on the Sharp Signals page can show the underlying signal’s n. If we’re ever showing fewer than ~100 graded picks for a given window or signal, treat the number as directional, not predictive.
06 Pauses & reactivations
Any feed on this site — a single signal, a sport, or an entire stream — can be paused or reactivated at any time. We treat that as normal operating procedure for a quantitative system, not an exception. Some of the reasons we’ve paused things in the past or could pause them in the future:
- Regime shifts — regular season → playoffs, rule changes, a new ball, a coaching-strategy wave that the training data didn’t see.
- Edge decay — live ROI or win rate diverging materially from backtest expectation (see When something stops working for the exact thresholds).
- Data-source disruption — a feed we depend on (line moves, public-money splits, injury reports) goes offline or changes shape.
- Bookmaker model updates — when the market re-prices around a pattern we’ve been exploiting, that pattern can stop paying.
- Sample-size guardrails — a new or low-volume signal can be benched until it has enough live data to evaluate honestly.
When something is paused, picks stop firing publicly but still write to the database for forensics. When we reactivate, the historical record stays attached — we don’t reset the books to make the comeback look cleaner.
A worked example: in April 2026 our NBA team-bet Sharp Signals dropped from 62.4% (regular season, n=85) to 41.7% (early playoffs, n=48) — a 20.7-point swing across spreads, moneylines, and totals. We pulled the public feed rather than ship picks we knew were negative-EV, kept logging them privately for forensics, and committed to retrain on regular-season-only data with a separate playoff-aware model before bringing it back. That’s the template for how any future pause gets handled.
Why this disclosure exists. Most pick services bury their pauses and quietly tweak the model until the public stats come back to baseline. We’d rather tell you up front that regime risk is real, that we may pull or restore a feed without warning when our monitors say to, and that the historical record stays intact either way.
07 What you see on each pick
Pick cards are tier-aware. We pack different layers of context onto the card depending on what your plan unlocks — the underlying pick (side, price, score) is the same across paid tiers; the analytics layer expands as you move up.
- Edge sees, on every pick: the bet, the captured American odds at fire time, the best-of-book price across our 5 major US sportsbooks, live score as the game plays, and the resolved final score.
- Pro adds, on each pick from an active-roster model: the model name, its historical win rate, ROI (real-odds where verifiable), p-value, sample size, and the confluence / dispute chips that mark when two independent models agree (or disagree) on the same call. Pro also unlocks the cross-tier per-model leaderboard.
- Sharp gets everything in Pro plus each signal three minutes before it’s released to Edge and Pro, the daily model deep-dive page, and unlimited AI Betting Analyst queries.
See the full feature comparison for the per-tier breakdown.
What stays proprietary
To preserve whatever edge the signals have, we don’t publish:
- The exact computational definition of any individual signal.
- Which features feed each model, beyond the high-level taxonomy on the analytics page (Pro & Sharp).
- Real-time positioning — we surface sharp / public splits at fire time, not minute-by-minute.
If we published the exact recipe, the books would price around it inside a week. The track record, the methodology, the grading rules, and the per-pick metadata your tier unlocks are all on the table — the recipe is what stays in the kitchen.
08 When something stops working
Every active signal is monitored on a rolling window. The criteria for automatic pause:
- Live ROI ≤ -10% on n=5+ picks (fast pause)
- Live ROI ≤ -5% on n=20+ picks (slow pause)
- Live win rate ≤ 52% on n=20+ picks
- Live performance diverging by ≥ 10pp from backtest expectation
Paused signals stop firing featured picks immediately. They still write to the database for forensics. We re-evaluate paused signals every two weeks; some come back, some don’t. The roster on the analytics page rotates — once a signal goes a full week without firing, it drops off the public leaderboard automatically.
If you have questions about how a specific number was computed, or you want to see the raw graded record, email hello@sportsgods.com. We’ll send you the export for that signal.