Human and Analytical Traps in Strategy Decisions

The last mile is human

Clean data and good models still produce bad decisions when human bias and sloppy analysis intervene. These traps are subtle because they feel like good judgment.

Reputation bias

Teams favor famous numbers and powerhouse reputations over what the data says at this event. A historically great team can have an off year or a broken mechanism; a no-name team can be quietly dominant. Fix: rank from this event's scouting and analytics, and require a data reason for any pick that contradicts the numbers.

Recency bias

The last match you watched dominates your memory. One spectacular climb or one embarrassing tip-over warps your read of a team you have seen ten times. Fix: decide from the full sample (averages plus best/worst observed), not the freshest memory. This is exactly why you keep records instead of relying on the stands' gut feel.

Anchoring on a single number

Ranking only by total EPA or total OPR hides fit. The best pick is often not the highest-rated robot; it is the robot that completes your alliance, e.g., a guaranteed deep climber when you already have offense, or a defender to neutralize the opponent's L4 machine. Fix: rank by component capability and explicitly ask "what gap does this fill?" before every pick.

Ignoring reliability and variance

A robot averaging 40 points but with one zero-point breakdown is riskier in a best-of-three elim than a steady 35-point robot. Averages hide variance. Fix: carry a consistency metric (standard deviation, or count of failed climbs) next to every average, and weight reliability higher as you go deeper in elims.

Overfitting to noise

Reading deep meaning into a 2-3 match sample, or chasing a tiny EPA difference between two teams, is false precision. Small differences are within the noise. Fix: treat near-ties as ties and break them on reliability, defense, or driver skill, not on the third decimal of a rating.

Confirmation bias in scouting itself

If strategists believe a team is great, scouts may unconsciously record generously. Fix: scouts record raw actions (taps), not judgments; keep "how good is this robot?" opinions in a separate subjective/super-scout field, clearly labeled, so objective counts stay clean.

A decision-review habit

After each event, review your picklist against what actually happened: which picks over- or under-performed, and why. Was it bad data, a bias, or genuine variance? Over a season this turns your strategy group into a measurably better decision-making instrument, the same way per-scout accuracy review improves your data. The teams that win consistently are not the ones with the fanciest model; they are the ones who systematically debug both their data and their own judgment.