Project 4: Build an EPA-Component Robot Ranking Sheet

Goal

A single Google Sheet (or pandas notebook) that ranks every team at your event by capability, using your own scouting as the spine and public analytics as a cross-check. This is the artifact a picklist (covered in the existing strategy module) is built from, so here we focus only on constructing and validating it.

Columns to compute

Start from your aggregated scouting (averages per team from QRScout output), then add derived metrics that map directly to REEFSCAPE scoring and ranking points. Use the real point values so the number is in actual match points. Because coral is worth more in auto than teleop, weight the two phases separately:

• Avg points contributed = auto coral (3*avgAutoL1 + 4*avgAutoL2 + 6*avgAutoL3 + 7*avgAutoL4) + teleop coral (2*avgTelL1 + 3*avgTelL2 + 4*avgTelL3 + 5*avgTelL4) + algae (6*avgAlgaeProc + 4*avgAlgaeNet) + auto leave (3*leaveRate) + endgame average (Park=2, shallow Cage=6, deep Cage=12).
• Auto reliability = fraction of matches with Leave + at least 1 coral scored in auto (this maps to the Auto RP, which needs all robots to leave AND at least one coral in auto).
• L4 rate = avg L4 coral per match. High L4 teams drive the Coral RP, which needs 5 coral on each of 4 levels (or on 3 levels if the Coopertition bonus is earned via 2 algae in each alliance's processor) at regular-season events — FIRST raises that bar to 7 coral per level for the FIRST Championship only.
• Endgame value = expected barge points. The Barge RP needs 14 barge points at regular-season events (FIRST raises this to 16 for the FIRST Championship only), so one deep climb (12) plus a park (2) — or two deep climbs (12 each) — clears it.
• Cross-check columns: TBA OPR, Statbotics EPA total, and Statbotics endgame EPA pulled via Project 3.

Spreadsheet recipe

1. Tab raw: every scanned QRScout row.
2. Tab agg: =AVERAGEIF(raw!$C:$C, $A2, raw!D:D) style formulas to average each metric per team (column A = team list).
3. Tab rank: the derived columns above, plus a consistency column = STDEV of total points per match (lower is more reliable).
4. Sort rank by Avg points contributed descending.

Validate before you trust it

The whole sheet is worthless if the inputs are wrong, so run three checks:

• Match-sum check: for a sampled match, your three scouted robots' points should be within roughly 10-15% of the TBA score_breakdown alliance total. Large gaps mean a scout miscounted or a robot was mislabeled.
• OPR/EPA agreement: flag any team whose scouted average is far from both its TBA OPR and Statbotics EPA. That team gets a super-scout next match.
• Sample-size guard: hide ranks for teams with fewer than ~3 scouted matches; one bad partner alliance can wreck a small average.

Why component breakdown beats a single number

Two robots can have identical total EPA but win you very different matches: one is a pure-offense L4 coral machine, the other a reliable deep-climb + algae robot that secures the Barge RP. A component sheet lets a strategist pick the robot that fills a specific gap (e.g., "we already have offense, we need a guaranteed climb"), which is exactly the decision alliance selection turns on.