Methodology
Where our data comes from, which matches count, and what we leave out. Definitions and formulas for every statistic are in the metrics reference.
Data sources
Official results and our own calculations are kept separate. Elo, OPR, and every other rating on this site are FTC Secrets calculations, not FIRST rankings or predictions.
- FIRST FTC Events API
- Team profiles, event details, event rankings, and awards, when the server has FIRST credentials configured. FIRST publishes event data after a tournament concludes, so it is not a live scoring feed.
- FTCScout
- Match participants, surrogate, no-show, and disqualification flags, final alliance totals, and game-specific score breakdowns. It is the current source for all match data, and each match keeps its raw source record and timestamp for audit. FTCScout also stands in for profiles and events when FIRST is unavailable.
- 2026–27 BIOBUZZ competition manual
- Season rules and ranking point definitions.
Update schedule
A scheduled job runs once a day at 04:00 UTC. It pulls new and corrected matches, refreshes event analytics, and saves a daily rank snapshot. When any match is new or has changed, it replays the whole season to rebuild ratings and team metrics.
Between runs, pages show the last completed sync, so results can trail live scoring by a day or more.
Which matches count
A played match counts toward Elo and win–loss records when both alliances have two distinct teams and a reported final total. Qualification and playoff matches count equally, and a match the source lists more than once counts once.
Surrogates, no-shows, and disqualifications
- A flagged team is skipped for its own rating update. The other teams keep the match result.
- The flagged team’s rating still goes into its alliance’s average when the expected result is calculated, so the match is not exactly zero-sum.
- The match is also left out of that team’s record, consistency, highest alliance score, and opponent strength. Everyone else in the match keeps it.
- A no-show is not counted as a match played. Surrogate and disqualified appearances are.
- OPR works on whole alliances, so a match with any flagged team is dropped from that event’s OPR.
How results are decided
Wins, losses, and ties come from the official final alliance totals, which include penalty points. Non-penalty scores measure scoring ability and feed OPR, consistency, and highest alliance score, but they never decide a result. A match with no final totals is left out rather than decided from non-penalty scores.
Elo reads only the outcome: win, loss, or tie. Margin of victory has no effect.
What is excluded
- Matches that have not been played, or whose scores are not reported.
- Matches without two distinct teams on each alliance.
- Matches without a reported final total. They are omitted from Elo and records, never estimated.
- Flagged participants, as described above: per team for most statistics, per match for OPR.
- OPR for events whose schedule cannot separate every team.
- Score fields the source has not verified for a season. We never borrow another game’s fields.
- Results from earlier seasons. Nothing carries over.
Known limitations
- Elo is relative and local to a season. It compares teams only through opponents they share, so regions that rarely play each other can drift apart. Early-season ratings are noisy, and no confidence interval is published.
- Alliance-level statistics (Elo, OPR, consistency, highest alliance score) cannot separate a team from its partners within a match.
- Win probability is not calibrated. In a backtest on 2025–26 matches the model was under-confident, and the public predictor is labeled not validated. See the calibration table.
- OPR estimates shared alliance scoring, not one robot’s points, and values from different events cannot be compared.
- Opponent strength uses end-of-season ratings. Field strength and event difficulty use current ratings, not ratings on event day.
- Match data passes through a third-party source. A correction made upstream reaches this site on the next daily sync.
Season handling
Every season starts from scratch. All teams begin at 1500, nothing carries over, and ratings from different seasons are not comparable.
Score fields for 2026–27 BIOBUZZ are verified, so score-based metrics such as OPR, ranking point rates, and consistency are available for the current season.
2025–26 DECODE has verified score fields, so those metrics are available for it. Pages that need them offer it as an alternative, for example the 2025–26 DECODE leaderboard.