Start with your race. Open the live race chooser or paste the exact TVG/FanDuel card URL. GAMELIN verifies the race identity and loads whatever runner and price data the source exposes. If the feed cannot supply a usable field, you can paste real runners instead. Every read is free, and the result belongs to the data available at that moment.
What the formulas do to that field
- Check the card: identify the track, date, race number and active runners; label the source, observation time and missing information. A posted schedule is not itself a usable runner table.
- Read the market: convert available prices into implied chances and normalize the field so tote overround does not masquerade as extra winning probability. Morning lines and current odds carry different labels.
- Build the model read: rank the runners using the signals actually available for this data mode. Class, recent speed, market position, agreement and coverage can affect conviction. A field missing compatible figures does not silently receive a Full Figures score.
- Separate ranking from price: show the projected winner, then compare modeled chance with the displayed price. A probability edge and expected return are different quantities; neither alone is a wager instruction.
- Apply the money gate: inspect source quality, field integrity, coverage, conviction, price and other safeguards. The result can be BET, a watchlist-level read, VERIFY, or PASS. Recheck the card and price near post.
Follow one illustrative price calculation
Suppose a runner has an assumed 25% win probability and the offered price is 5/2. The decimal return is 3.50 for each 1 staked. Its raw break-even chance is 1 ÷ 3.50, or 28.57%; the conditional expected return is 0.25 × 3.50 − 1 = −12.5% per unit. At 4/1, decimal return is 5.00 and the same assumed probability produces +25%. This is arithmetic with a fixed assumption, not an actual GAMELIN estimate or a BET call. Tote prices can change. Use the fair-odds calculator to test your own assumption, or open the Race Lab after an eligible real read to vary a hypothetical price while keeping that read's probability fixed.
How to interpret your result
Read the source and coverage label before the headline horse. The projected winner answers which runner ranks first in this particular read. The money decision asks a separate question: whether an eligible candidate, available price and quality checks justify a bounded primary WIN recommendation. A PASS does not erase the ranking; it says the money gate did not qualify a wager. An odds change, scratch or data update can change the read. A shared race URL loads a fresh read rather than replaying someone else's unsaved browser result.
Methodology explains how GAMELIN reaches a race decision.
This page documents identity, freshness, read modes and decision gates. For empirical support, use the Evidence page; for real pre-race records, use the Forward Test.
One race, five information states
| Mode | Available information | Monetary limitation |
|---|---|---|
| Entry Card | Posted runners and morning line | PASS; not a current-tote or complete-figure read |
| Live Odds | Current price information when available | No Conviction when fewer than two runners or less than 40% of the field has compatible figure pairs |
| TVG Provisional Conviction | Current prices plus compatible TVG class/speed pairs for at least two runners and 40% of the active field | Neutral imputation, coverage shrinkage, stronger value thresholds, 1% stake cap, and no inherited canonical performance claim |
| TVG Source Conviction | Current prices plus complete TVG average class and speed | Reduced stake cap; historical canonical performance is not assigned to this mapping |
| Full Figures | Current prices plus canonical required figures across the active field | Only a candidate; all quality and value checks must still pass |
Projected winner and betting candidate are separate
The projected winner is the highest-ranked runner under the available model read. A primary monetary BET is a separate decision that requires complete eligible inputs, appropriate Conviction and an acceptable price. At most one primary WIN candidate can qualify. Otherwise the money decision is PASS.
Do not interchange the numbers
Market probability is the normalized market share inferred from prices across the field. Model probability is GAMELIN's field-wide estimate. Probability edge is their difference in percentage points. Price expected value compares an estimated chance with the actual displayed return. They are related but not identical calculations.
Data definitions matter
Similar labels from different sources are not silently relabeled. TVG average speed remains TVG average speed and runs only inside clearly identified source-native or provisional modes. Canonical Full Figures retain their separate definition and historical evidence. A duplicate or ambiguous field still requires verification, while eligible partial coverage is shown with its exact count and an explicit uncertainty penalty.
Research is not a live result
The evidence page distinguishes starter-level research records, separately counted odds snapshots, historical model validation and international studies. A historical subset return is not a forecast, an individual-race guarantee or a typical-user earnings claim. A synthetic demonstration is not a real race prediction.
What remains to be proven
A prospective evidence record must preserve the decision and available information before the race and settle it afterward without rewriting the original call. An ordinary webpage timestamp alone does not establish that record. Publication of a guide or calculator does not expand the model's validation population.
Independent and free
The core race read remains free. Partner compensation cannot alter the analysis, and an outbound commercial prompt must not override PASS. GAMELIN supplies decision support; any permitted wager is the user's separate action with a regulated provider.
Sources and scope
Official references and GAMELIN's published evidence. Check current track and operator information before acting.