Player Props

How to use the Player Props board

Every row pairs two independent signals for one player, market and line: what our prediction model expects in this fixture, and what the player has actually done recently. The edge is in reading them together.

The columns

Model %

The probability our model gives the player to clear the line (e.g. 2+ shots) assuming they start and play most of the game. Built from their recency-weighted per-90 output, blended with a positional baseline, scaled for this match-up, and calibrated against historical results.

Fair odds

The model probability expressed as decimal odds (1 / probability). This is the break-even price: a bookmaker price above fair odds means the model sees value; below it, the model says the price is short. We deliberately don't show bookmaker odds — player-market prices arrive late and patchily, so compare against your own book.

Hit rate

How often the player actually cleared this line in their recent games (the 7/10 under it is the raw count). Uses club and international games together, most recent first. Change the window with the Last 5 / 10 / 15 toggle.

Form

The same recent games as squares: green cleared the line, red didn't, the number is the actual count that game. In the detail panel, minutes played are shown under each square — a "miss" in a 20-minute cameo means little.

Conf

How much data sits behind the model number. High: a regular starter in a league we fully cover. Med: decent but partial data. Low: thin or missing club coverage — the number leans on positional averages, treat it as a sketch. Cards can never show High: bookings are inherently low-predictability.

Score

A blend of Model % and hit rate for ranking, weighted by confidence (high confidence trusts the model more). It's a sorting aid, not a probability.

Δ flag

Model and recent form disagree by more than 25 points. Not a red flag — a reading prompt. Open the detail panel: often the model sees a soft or brutal match-up that raw form can't, or the form run came against very different opposition.

Suggested workflow

• Browsing a match: pick the fixture, pick a market, and scan high-Score rows where Model % and hit rate agree and confidence is Med+.
• Hunting across fixtures (Analyst): choose "All fixtures", narrow by competition or date, then set minimum Model % and minimum hit % — e.g. model ≥ 60 and hit ≥ 60 on shots 2+ surfaces the bankers; model ≥ 30 on goals surfaces priced-out scorers.
• Before trusting any row: open the detail panel and check minutes in the form chart (is this player actually starting?) and the "Why this number" block (is the sample real or a fallback?).

Honest limits

• The model assumes the player starts (~85 minutes). It doesn't know lineups — check team news for rotation risk; substitutes' probabilities are overstated.
• Players from leagues we don't cover (and tournament debutants) show Low confidence — their number is a positional estimate, not personal form.
• Knockout fixtures without decided teams show "No preds yet" — we don't invent numbers for placeholder ties.
• These are model estimates for research, not betting advice. Stakes are your own.