Learning goals
By the end, members should be able to implement and explain the method, reproduce the worked example, identify failure modes, and distinguish a supported research conclusion from an unsupported betting claim.
Full teaching lesson
A handicap is a forecast supported by evidence, not a collection of interesting statistics. The professional workflow starts with an explicit target: Which market, which rule set, and at what time are we estimating a fair probability? A spread, moneyline, total, and player prop each require a different probability output. A claim such as “Team A is better” does not specify a price, outcome, or testable probability. Define all three before looking for supporting evidence.
Begin with an information cutoff. Record the precise timestamp when the analysis starts, the event identifier, the market, the available quote, and the data that would have been knowable at that time. Keep a separate log of later news. This prevents hindsight contamination: adding a starting-quarterback absence learned on Saturday to a Friday forecast is not an honest Friday backtest. Maintain a baseline forecast from a pre-registered method, then log justified adjustments one by one.
Organize research into five passes. Pass one is data integrity: confirm opponents, venue, home designation, game format, rest, injuries, and market settlement. Pass two is baseline ability: strength ratings and expected pace. Pass three is matchup-specific evidence: lineups, positional interaction, stylistic strengths, and game environment. Pass four is pricing: transform your forecast into a market-specific probability and compare with the executable quote after removing commission or vig as appropriate. Pass five is decision and audit: save the final estimate, sensitivity range, reason codes, and whether the opportunity cleared a pre-specified decision threshold. No qualifying opportunity is a valid outcome.
Use a versioned research sheet with columns for source timestamp, estimate, adjustment size, assumption, analyst confidence, and post-event audit notes. An injury adjustment supported by historical replacement performance is distinguishable from an anecdotal “team always struggles in cold games.” Assign the latter a lower evidentiary grade or leave it out. Keep correlated changes together: a missing quarterback can affect passing efficiency, turnover risk, total points, and receiver usage; treating each downstream effect as an independent penalty will double count it.
Worked example
Suppose a fictional model projects Home 26.0 points and Away 23.0: Home by 3.0 and a total of 49.0. Verified lineup information changes the Home offense by −1.5 expected points and the Away offense by +0.5, giving Home 24.5, Away 23.5, Home by 1.0, and total 48.0. Compare that forecast with the market only after choosing an outcome distribution. A projected one-point margin alone does not identify the probability that Home covers −2.5; estimate the distribution of scoring margins, including football's discrete final-score effects. If simulation estimates Home covers −2.5 in 46% of realistic scenarios, −110 requires 52.38% to break even, so the hypothetical price is unattractive. This is a methodological example, not an actual live game or line.
Quality controls
Before publication, a second reviewer should independently recreate the arithmetic, confirm each source existed at the recorded cutoff, inspect the adjustment ledger for double counting, and verify that the final decision used an obtainable price. Teaching content should show both rejected and accepted hypothetical ideas; otherwise members learn selection bias instead of process.
Exercises and answer key
Exercise 1. Baseline Home 28, Away 21. Revised Home −2, Away +1. What is the new projected margin and total? Answer: 26–22; margin Home +4, total 48. Exercise 2. An analyst learns of an injury after the event and retroactively changes the pregame prediction. What failed? Answer: timestamp integrity and information-cutoff discipline. Exercise 3. The projected margin changes materially but the analyst cannot explain why. What happens next? Answer: suspend the forecast, audit source and adjustment history, and do not invent an explanation.
Teaching application and ARBAI companion
The Academy teaches the structured method with saved examples and an assessment. ARBAI can walk a learner through an unfamiliar calculation, test a hypothetical assumption, or explain a confusing model output conversationally. ARBAI responses should not rewrite timestamped course case studies or substitute for reviewed teaching material.
Editorial review gates
Check every equation and answer, verify sport-specific assumptions, label all examples as hypothetical, review statistics and data-provenance claims, check content rights, and apply appropriate regional and safer-play review. No certain-return claims.
Try it in Sharp App: Odds screen
Questions? Ask ARBAI about this lesson · Ask our Sharps in Discord →
Sharp App surfaces +EV opportunities from public sportsbook data. We do not place, accept, or facilitate wagers. 21+ where legal. If betting stops being fun, call 1-800-GAMBLER or visit ncpgambling.org.
