Keeping Data Accurate and Trustworthy

Garbage in, garbage out

The fanciest app produces useless picklists if the data going in is wrong. Most scouting errors are human, not technical: a tired scout, a confusing form, watching the wrong robot, or typos during entry. Accuracy is mostly a people-and-process problem.

Common error sources and fixes

• Wrong robot watched. Scouts confuse which robot they are assigned to, especially across alliance colors. Fix: assign each scout a fixed field position (e.g., "red 2") for a whole session and pre-load the schedule so the app/sheet tells them the team number for that slot each match.
• Scout fatigue. Concentration falls off after many matches. Fix: rotate scouts every few matches, give breaks, and never run with too few people.
• Inconsistent definitions. Two scouts count a "cycle" differently. Fix: write one-sentence definitions for every metric and train the whole team to the same standard before the event.
• Transfer typos. Mostly a paper problem. Fix: two-person entry (one reads, one types) or a QR system that eliminates retyping.
• Missing matches. Gaps make averages unreliable. Fix: track coverage so you know which robot-matches you actually captured.

Validate the numbers

Build simple sanity checks into your spreadsheet:

• Range checks. Flag impossible values (more pieces than physically possible in a match).
• Cross-checks. The sum of the three robots' scouted contributions should be in the neighborhood of the alliance's official score from The Blue Alliance. Big mismatches mean a scouting error or a robot you mis-tracked.
• Reliability vs averages. A high average with many breakdown flags is a warning, not a green light.

Don't trust tiny samples

Early in an event you have only one or two matches per team. Two matches is not enough to judge a robot; a single great or terrible match swings the average wildly. Treat early data as provisional and let it stabilize. This is also why public metrics like OPR and EPA improve as more matches are played.

Measure your own accuracy

Strong programs audit themselves. After an event, compare your scouted alliance contributions to actual scores, or have two scouts independently track the same robot for a few matches and compare. Teams like Citrus Circuits (1678) publish whitepapers analyzing their own data accuracy, which is a great model to study. Knowing your error rate tells you how much to trust your own picklist versus public analytics.

Culture matters

Scouting is often seen as boring, so the data quality follows team culture. Make scouting respected: rotate everyone through it (including drivers and leads), explain how the data changed a real decision, and celebrate the scouts whose notes won a pick. A team that values scouting collects better data, and better data wins matches.