A short, written routine beats last-minute hunches.
Minutes before kickoff, alerts ping: injury updates, celebrity picks, and a friend's “he always comes alive in big games.” It's tempting to chase those stories; Super Bowl player props are high-variance and information-overloaded, so a vivid anecdote can hijack judgment.
A compact, documented routine—three quick checks or under five minutes—replaces gut swings with a repeatable habit. Write the prop, note the market baseline, record one short reason and the planned stake. That small ritual reduces doubt, curbs recency bias, and makes choices more defensible under pressure.
- Rate confidence 1–5 to compare choices objectively.
- Cap each prop stake at 1–3% of the Super Bowl bankroll.
- Keep a one-line log for postgame review and learning.
Create a single-game baseline from season data
Convert season and per-game numbers into a realistic single-game expectation by choosing a sample window, scaling for game context, and adding opponent and variance adjustments. The result should be a reproducible projected distribution that can be compared directly to prop lines.
-
Pick a sample window
Balance recency and sample size: use the last 4–8 games for role changes or a full season when roles are stable. Exclude games where the player missed snaps due to injury or was limited by obvious game-script distortions (blowouts).
-
Scale season/per-game numbers to one game
Start with per-game rates (targets, carries, yards) and scale by expected snaps and team pace for the matchup; factor in projected game script (score margin expectation). For the conversion step, consult the detailed conversion method for formulas and examples.
-
Apply opponent adjustment and simple variance model
Adjust the expectation by a defensive multiplier (e.g., opponent target or rush-rate allowed) derived from recent data, then choose a distribution: Poisson for low-variance counts, negative binomial or a two-part model when variance is larger. Record each assumption in a one-sheet so projections stay reproducible and auditable.
Run simple checks before trusting a number. Compare implied per-snap rates to season norms; if the projected per-snap rate is far outside historical ranges, the model likely overreacted.
Simulate 1,000 game outcomes from the chosen distribution and inspect median, 10th, and 90th percentiles. If the bookmaker line sits well inside the simulation spread, the prop may be fair; if it lies in a thin tail, consider alternative inputs.Pick a metric and stick to it
- Per-snap rate
Measures production per offensive snap (targets/snap, yards/snap). Useful when playing time fluctuates — see the explainer on per-snap vs per-game stats for when it outperforms totals.
- Per-route rate
Focuses on opportunities while a receiver is actually running routes (targets/route). Best for predicting catches and yards when snap share is stable but route participation varies.
- Per-game totals
Simple averages of final box-score counts. Good for quick benchmarks but vulnerable to outlier games and usage shifts.
- Usage metrics
Targets, route share, and snap share capture opportunity, often beating raw totals when a player’s role changes; prioritize these for prop lines tied to volume.
- Receptions vs receiving yards
Receptions rely on targets and catch rate; yards depend on depth of target and yards after catch. For more nuance, consult the discussion of their correlation.
State the chosen metric before modeling. Metric choice changes projected outcomes.
If volume is uncertain, use per-snap or per-route rates. If role is stable, per-game totals are simpler and often sufficient. For receptions vs yards props, model targets and depth separately then derive both.Changing the metric mid-analysis is the most common source of inconsistent projections.
From script to volume: adjust baseline plays
Start with the single-game baseline (season rate scaled to one game). Convert a verbal script — run-heavy, tracking a lead, garbage-time comeback — into simple percentage shifts on that baseline.
Use short rules of thumb as initial multipliers:
- Run-heavy: -15–25% pass attempts, +20–40% rush attempts for primary backs.
- Pass-heavy (trailing): +20–40% pass attempts, RB targets decline ~10–25%.
- Blowout (leading): starters rest; veteran work share falls ~30–60% in fourth quarter.
Apply these changes to per-player volumes (targets, routes, snaps), then recompute totals and per-play averages. For refining passing outcomes, tie pass-attempt adjustments to expected yards per attempt — see the practical method to adjust passing yards for game script.
Coach tendencies shift the multipliers: conservative play-callers damp variance, while aggressive or historically game-specific coaches tilt toward passing or RB usage. Consult historical big-game usage to pick multipliers that match the coach profile: coach big game player usage history.
Start with the midpoint of the multiplier range.
Adjust only one major axis (pass vs. run) per iteration.
Reweight for quarter-specific play-calling rather than averaging across the whole game.
Weather adjustments for the script
Apply fixed weather deltas to the script-derived baseline to avoid guesswork; when conditions worsen, follow simple rules rather than ad‑hoc changes. For full guidance, see the weather adjustment rules.
- Wind ≥20 mph: passing yards −12%, pass attempts −10%, rushing attempts +7%.
- Wind 10–19 mph: passing yards −6%, pass attempts −4%, rushing attempts +3%.
- Rain (light <0.3″): passing yards −5%; heavy ≥0.3″: passing yards −12%, increase short targets.
- Cold <40°F: passing yards −5%, rushing efficiency +4%.
- Combine effects additively but cap total adjustment at ±25%. Apply changes to attempts first, then per-attempt efficiency.
Apply percentage adjustments to projected attempts before altering yards per attempt. Round final multipliers to the nearest whole percent and document the choice.
Turning last‑minute noise into clear adjustments
Quick decision tree
Map common practice labels to expected snap‑share moves and act reproducibly.
- Out / DNP — treat starter share as transferred to the backup (backup gets ~100% of starter snaps).
- Doubtful / Likely out — assume the backup takes 60–80% of the starter's usual snaps; starter may take limited situational snaps.
- Questionable / Limited — reduce starter snaps by 20–40%, assign backup 20–50% depending on role.
- Active (full practice) — no change to baseline.
Apply the change to targets
Convert snap adjustments into target and volume projections by multiplying baseline per‑snap rates (targets/snap, routes/snap) by the new snap share. If pass‑game share shifts, reallocate expected targets proportionally across receivers.
For step‑by‑step backup math, consult the backup snap projection guide. For translating reports into prop tweaks, see the practice report checklist.
Checklist:
Confirm which snaps are offensive vs. special teams. Check historical backup usage in similar injuries. Cap combined adjustments to ±30% to avoid overfitting.Why model touchdowns separately — and how to convert usage into odds
Why do touchdowns need different treatment?
Touchdowns are rare, high-variance events driven more by red‑zone and goal‑line usage than by season totals. Treating them separately focuses the model on opportunity and context; see the estimating touchdown probability guide for the full approach.
How to convert red‑zone chances and a player’s usage into touchdown odds?
Compute expected TDs = team red‑zone trips × player's share of red‑zone opportunities × team red‑zone TD rate. Convert expected TDs (λ) into probability of at least one TD with P=1−exp(−λ); e.g., 6 trips×0.4×0.5→λ=1.2 → P≈1−e^{−1.2}≈70%.
How should two‑minute and overtime be adjusted?
Apply small multiplicative adjustments to usage: increase pass‑targeted players in two‑minute (e.g., ×1.2–1.5) and reduce rushing TD chances (×0.6–0.9); treat overtime as lower overall scoring—scale expected TDs down (commonly ×0.5–0.8). See the walkthrough on two‑minute and overtime adjustments for specifics.
What’s a compact way to document these probabilities?
Record: RZ_trips, player_RZ_share, team_RZ_TD_rate, λ (expected TDs), P(≥1)=1−e^{−λ}, two‑minute_adj, overtime_adj. Keep one line per scenario for reproducibility and quick comparison.
How to turn the probability into betting odds?
Use the computed probability P as the fair probability for the market; implied decimal odds = 1/P. Compare to market odds to decide value after accounting for vig and variance.
Single‑line template: RZ_trips | RZ_share | RZ_TD_rate | λ | P(≥1) | 2‑min_adj | OT_adj
Example: 6 | 0.40 | 0.50 | 1.20 | 70% | ×1.3(pass) | ×0.6 Save scenario lines for baseline, two‑minute, and overtime so comparisons remain reproducible.Turn projections into actionable bets — step by step
-
1. Import lines and odds
Collect every book's line and price into columns: Book, Prop text, Line, Price (decimal or American), and timestamp. Include a column for market type (total/binary/moneyline) to drive later calculations.
-
2. Convert prices to implied probabilities
Convert odds to implied probabilities (standard formulas for decimal/American). Remove vig by normalizing the summed probabilities across outcomes so the total equals 100%.
-
3. Turn market probability into the market expectation
Use the market probability directly for binary props. For numeric totals, map the market cumulative probability to an implied mean using the projection's distribution (a normal approximation from projected mean and SD). For a ready template, run the quick spreadsheet walkthrough.
-
4. Compute edge and size suggestions
Add columns: ProjectionProb, MarketProb, Edge = ProjectionProb − MarketProb. Flag rows where Edge exceeds a chosen threshold (e.g., 5 percentage points) and compute stake suggestions (Kelly fraction or flat-percent) in adjacent cells.
-
5. Track exposures and results
Log placed bets with Book, Stake, Odds, Correlated exposures, and Result. Maintain running P&L and per-player max exposure to avoid over-concentration and to iterate on projection calibration.
Track correlation and caps. Correlated bets (same game or same player) can multiply risk quickly. Keep simple rules:
Set a per-player and per-game exposure cap. Record every bet and update P&L after settlement. Limit Kelly stakes when the model has limited data; prefer a fractional Kelly or fixed percent.Choosing a Plug‑and‑Play Model vs Building In‑House
-
Vendor deliverables and provenanceExpect documented inputs, outputs, sample code, update cadence and a small backtest. See buying a plug-and-play model for typical vendor promises and checklist items.Look forclear data sources, sample runs, and scheduled updatesAvoidopaque predictions with no provenance
-
Transparency and validationRequire feature lists, methodology notes, and out‑of‑sample or season‑level backtests with calibration metrics. Validation artifacts make spot checks and drift monitoring feasible.Look forbacktests, error bars, and access to intermediate featuresAvoidproprietary claims without verifiable tests
-
Integration, latency and update cadenceConfirm API formats, latency, and how often lines or inputs refresh; faster cadence helps late adjustments but can reduce stability. Check compatibility with existing spreadsheets or bet-tracking tools.Look forclean API, example client, and clear update scheduleAvoidmanual CSV drops or unpredictable release timing
-
Cost, support and customizationBalance licensing, per‑update fees, trial access and support response times against in‑house development costs. Ensure contract allows rollback or export of historical results for audits.Look fortrial period, export rights, and SLA for supportAvoidopaque pricing and no trial or export options
Repeatable prop checklist
-
Baseline
Set one data window and build a single‑game baseline with opponent and variance adjustments.
-
Metric choice
Pick per‑snap, per‑route, or per‑game up front; model receptions, yards, and targets separately.
-
Script adjustments
Translate script and role expectations into percent multipliers for snaps, routes, and carries; cap combined impact.
-
News rules
Map report labels (e.g., ‘limited', ‘questionable') to reproducible snap‑share multipliers and apply them immediately.
-
Touchdowns
Convert red‑zone share to expected TDs and use a Poisson conversion to get TD probability.
-
Cross‑book check
Compare implied probabilities across books, flag edges, and size stakes by edge and exposure.
Cap weather + script effects at ±25%.
Log and iterate
- Record every projection, adjustment, and bet outcome.
- Review misses monthly and recalibrate multipliers and news conversions.
- Treat baselines as versioned experiments, not fixed truths.
A simple log plus regular review turns repeatable rules into improving edges. Iterate modestly, track outcomes, and let documented results guide refinements.
