How to scout an FTC event

Stop rebuilding what already exists

Every year teams build a spreadsheet that tallies scores by team, and every year it produces a worse estimate than a least-squares fit of the same data. If a team played six qualification matches, their alliance scores contain enough overlap to separate their contribution from their partners' — that is what OPR does, and OPRc does it better where the season publishes score details.

Those numbers exist for every event on Robolyst, computed the moment results are posted, including for events you did not attend. Your scouting effort is better spent on what a score cannot show.

What only a human can see

Send scouts to record the things that never reach the scoring table. These are the observations that decide alliance selection when two teams look identical on paper.

• Cycle time — how long one complete scoring cycle takes, timed, not estimated
• Failure mode — what breaks, how often, and how fast they recover from it
• Driver skill — does the robot get where it is going, or fight the field the whole match?
• Autonomous reliability — how many of their auto runs actually completed, out of how many attempts
• Endgame — can they do it under pressure, or only when the field is clear?
• Defense — do they play it, and can they take it?
• Pit behaviour — is the robot maintained between matches, or held together with hope?

Run it with three people, not thirty

Elaborate scouting systems collapse because they need more people than a team has. A workable minimum: one person timing cycles for the robots you might pick, one taking notes in the pits, and one keeping the schedule and flagging which matches matter.

Give them a fixed sheet with the same fields every match. A free-text log is unreadable by Saturday afternoon, and unreadable notes are the same as no notes.

Read the numbers correctly

Three mistakes account for most bad picks made from data.

Confusing a record with a robot. A 5–1 team that drew strong partners and weak opponents is not better than a 3–3 team that drew the opposite. That difference is exactly what total luck measures — check it before you trust a record.

Comparing across events. OPR is fit inside one event. A 90 at a stacked regional and a 90 at an eight-team scrimmage are not the same robot. Compare within an event, or use season averages that span several.

Ignoring variance. Two teams averaging 60 are not equivalent if one scores 60 every match and the other alternates 20 and 100. In a three-match elimination series, the steady one is usually the better partner. Consistency and reliability are the columns for this.

Picking an alliance partner

By selection you are answering one question: which available robot most increases our chance of winning three matches in a row?

1. Rank the available teams by contribution — OPRc if the season has it, OPR if not.
2. Cross off anyone whose scouting notes show a failure you cannot afford. A high average with a robot that dies once every four matches is a trap.
3. Look at the section split. If your robot is weak in autonomous, a partner strong there is worth more than a higher raw number.
4. Prefer consistency for the elimination format. You need three good matches, not one great one.
5. Have a second and third choice ready and agreed before you are at the table. The team ahead of you will take your first pick more often than you expect.

Before the next event

Write down what your scouting got right and wrong while you still remember. The teams that scout well in their third year do it because they kept notes in their first, not because they bought better software.