An alert is only useful if it is actionable before the market moves. Every Desk alert ships four mechanics — strategy, entry trigger, unit size and invalidation — so you can place or skip it without further research. Click any underlined term to open its definition in the glossary below.
Part 1
The named pattern the bet comes from — not a hunch. Example: Sharp-vs-soft prop divergence on rebound overs after a starter is ruled out.
Part 2
The exact price and market state required. Example: take Over 8.5 rebounds at -115 or better while the sharp fair sits at 56%+.
Part 3
off your bankroll and , then capped by your max-stake rule so no single ticket can distort the log.
Part 4
The condition that kills the bet before you place it: price moves through fair, the number moves against you, or the player is downgraded.
Scenario gallery
Illustrative numbers, run through the identical pipeline: strip the sharp book’s hold, convert to a fair price, measure the soft book against it, size at quarter-Kelly on a $2,000 bankroll, then grade the price against where the market closed. Compare how favourites, long-shots, props and efficient markets behave — and how line movement decides whether an edge was real.
Player points — over
Props hold the most vig, so the de-vig step moves the fair number the furthest. The soft book never repriced until close — that gap is where prop edges live.
Moneyline — heavy favourite
Big favourites need a large price improvement to produce a small edge, and Kelly demands a huge stake for it. This is why the risk band caps break-even probability.
Puck line — underdog +1.5
The market drifted our way after we bet. The realised CLV is bigger than the edge we logged — the strongest confirmation signal the model gets.
Team total — over
Edge cleared the floor at post time, but the line ran against us into close. Negative CLV on a bet that looked +EV is how a stale sharp quote shows up.
Spread — pick'em
A near-efficient market: after de-vigging there is barely anything left. Correctly, no alert fires — most of the board looks like this.
Alert threshold used here is the default 2.5% edge. Change it in the alert simulator above to see which of these would still fire for you.
Availability
Each sport maps to one competition on the odds feed. If that competition is out of season — or the market isn’t one the feed carries — the pick is priced by the model instead of a live sharp quote. Everything below is the rule set the engine follows.
A provider key, a sport that is in season, and a sharp book posting a two-sided quote on that event. When all three hold, the fair price is de-vigged from real market numbers.
Out of season, between the 6-hour free-tier syncs, when the provider is rate-limited or returns incomplete markets, or for any market the feed doesn't carry (props, spreads, totals, exotics).
The board header shows the source of the last sync — "live market feed" or "model feed" — and a banner appears whenever the engine has fallen back. A manual Resync always attempts live first.
| Sport | Feed competition | Season | Status now | Priced live | Model-priced |
|---|---|---|---|---|---|
| NFL | NFL regular season + playoffs | Sep – Feb | In season — live eligible | Moneyline; in-game stat props — passing/rushing/receiving yards (plus longest completion/rush/reception and the pass+rush and rush+rec combos), completions, attempts, pass/rush/rec TDs, interceptions, receptions, kicking points, field goals, tackles, sacks, anytime TD; game derivatives — 1st-half and quarter spreads/totals, team totals, alternate numbers | Spread, Team Total |
| CFB | NCAA FBS regular season + bowls/playoff | Aug – Jan | In season — live eligible | Moneyline; in-game stat props — passing/rushing/receiving yards and the rush+rec combos, receptions, rush attempts, pass/rush/rec TDs, interceptions, anytime TD; game derivatives — 1st-half and quarter spreads/totals, team totals, alternate numbers | Spread, Team Total |
| NBA | NBA regular season + playoffs | Oct – Jun | Off-season — model only (1mo) | Moneyline; in-game stat props — points, rebounds, assists, threes, steals, blocks, turnovers, field goals, free throws, 1st-quarter points, double/triple-double and the combo lines (PRA, P+R, P+A, R+A, B+S); game derivatives — 1st-half and 1st-quarter spreads/moneylines/totals, team totals, alternate numbers | Spread |
| MLB | MLB regular season + postseason | Mar – Oct | In season — live eligible | Moneyline; in-game stat props — strikeouts, pitcher outs, hits allowed, earned runs, walks, pitcher win, total bases, hits, runs, RBIs, H+R+RBI, singles, doubles, walks, batter strikeouts, stolen bases, home runs; game derivatives — first-5 moneyline/run line/total, 1st-3 and 1st-inning totals, team totals, alternate numbers | First 5 Under, Team Total |
| NHL | NHL regular season + playoffs | Oct – Jun | Off-season — model only (1mo) | Moneyline; in-game stat props — shots on goal, points, assists, goals, power-play points, blocked shots, goalie saves, anytime goal; game derivatives — 1st-period moneyline/puck line/total, team totals, alternate totals | Puck Line, Team Total |
| MMA | UFC + major promotions | Year-round | In season — live eligible | Moneyline (fight winner) | Fight Total Rounds, Method of Victory |
| Soccer (US) | MLS | Feb – Dec | In season — live eligible | Draw No Bet (derived from the two-way quote); in-game stat props — shots, shots on target, assists, anytime goal; game derivatives — 1st-half goal totals, both teams to score, team totals, alternate totals | Over/Under Goals, Both Teams to Score, Asian Handicap |
| Soccer (Intl) | Premier League (UCL and other comps priced by model) | Aug – May | In season — live eligible | Draw No Bet (derived from the two-way quote); in-game stat props — shots, shots on target, assists, anytime goal; game derivatives — 1st-half goal totals, both teams to score, team totals, alternate totals | Over/Under Goals, Both Teams to Score, Asian Handicap |
Model-priced picks use the same de-vig, edge and Kelly math as live ones — the difference is the input. Treat them as a rehearsal of the workflow, not as bettable prices.
A single NFL moneyline is priced by every desk and hammered by sharp money for days. A third-string tight end's receptions line is priced by an algorithm, published late, and moved mostly by news rather than volume. That asymmetry in attention — not any secret model — is why prop markets carry more exploitable error than the main markets on the same game.
A pick with no trigger price is untradeable: by the time you see it the number may already be through fair. Every alert therefore carries the strategy it came from, the exact price and market state that make it live, the stake as a fraction of bankroll, and the condition that voids it. If any of the four is missing you are guessing at three of them.
Full maximises long-run growth but assumes your estimate is exact, which it never is with a prop. Quarter-Kelly stakes a quarter of that amount: on a 3% edge at -110 that is roughly 0.3% of bankroll rather than 1.2%, giving up a little growth in exchange for a drawdown you can actually sit through. A hard max-stake cap sits on top so no single ticket can distort the log.
An alert dies when the soft price moves through the fair number, when the line itself moves against you (a 6.5-rebound over is a different bet at 7.5), or when the player's status changes — a downgrade, a coach comment, a late scratch on the other side of the rotation. Checking status at entry removes most of the tickets that look like bad luck later but were actually stale.
Prices longer than +398 are excluded regardless of computed , because margin skew and thin two-sided liquidity make the fair estimate unreliable exactly there. Edges beyond roughly 12% are treated as a data problem rather than an opportunity: in a market this heavily watched, a genuine 30% mispricing on a mainstream prop essentially does not exist.
On a full slate a disciplined screen produces a handful to a few dozen qualifying prices across all sports. Services showing hundreds daily are almost always comparing to each other, or skipping the entirely. Fewer, better-priced alerts with mechanics attached beat a long list you cannot act on before the number moves.