Behind the scoring
How Prop Scanner turns raw lines into a research score.
The data we collect, the model that scores it, and the limits of both, explained plainly so you can judge the evidence yourself.
From feed to score in four phases
Every prop goes through the same pipeline before it reaches the slate.
Ingest
Props, odds from major sportsbooks, injuries, box scores and game context are pulled on a schedule and stored, including a history of line changes.
Build features
Recent hit rates over the last 3, 5 and 10 games, recent averages, minutes trend, home/away, back-to-backs and line movement since the open.
Score
A logistic-regression model per prop type turns those features into a probability, calibrated so a score means roughly what it says.
Review & track
Open the reasons behind the score, compare alternate lines, log your decision, and later grade it and compare your price to the close.
01 / Data
A stored history, not a snapshot
Props are published days before tip-off, so the line you see today is only part of the story. Every change to a line or its odds is recorded, which gives each prop a true opening line, a movement trail and a closing line once the game starts.
Illustrative line history.
02 / Scoring
A calibrated model, with honest fallbacks
Players with at least five games of history are scored by a logistic-regression model trained on a large set of past props, with a separate model per prop type and Platt calibration. Thin-history props fall back to a simpler heuristic, and early-season samples are capped so a few games can't produce an inflated score.
Illustrative score inputs.
03 / Alternate lines
Compare every rung, not just the main line
The standard line is the most contested price on the board. Alternate ladders let you see the same player and market at several thresholds, so you can weigh a longer-odds higher line against the standard one with the hit rate for each in front of you.
Illustrative ladder.
04 / Verification
Closing line value and a results ledger
Single-day win rates are noisy. Pick tracking grades your logged decisions against real box scores and, when you log entry odds, compares your price with where the market closed.
Illustrative breakdown.
Transparency, including the limits.
No black boxes. Here is what the score is, and what it isn't.
Which sports and markets are supported?
NBA, WNBA and MLB. NBA and WNBA cover points, rebounds, assists, 3-pointers, PRA, steals and blocks, with alternate lines where available. MLB covers hitter and pitcher markets such as hits, total bases and strikeouts.
How often do lines refresh?
NBA lines refresh roughly every 5 minutes. WNBA and MLB props refresh twice an hour. Player stats and injuries update through the day, and closing lines are snapshotted every 5 minutes so closing line value can be computed after the game.
Can I change how props are scored?
Yes. The Model Builder lets you set your own weights for each sport, backtest them against stored history, and activate them. The default model is the one described above.
What is closing line value (CLV)?
CLV compares the price you logged with the price the market closed at. Beating the close consistently is a common sign of good price discipline, but it isn't a guarantee of profit, and a pick only gets a CLV when you log its entry odds on a game whose closing lines were recorded.
Does a high confidence score mean it will hit?
No. It is a calibrated ranking aid: higher scores have historically hit more often than lower ones, but any single prop can lose and no band is guaranteed profitable. Model History shows how each band has actually performed.
See it on today's slate.
Open any prop to see the inputs behind its score, then decide for yourself.