Best potential FMC mean

Potential

333fm mean-of-3 is just the arithmetic mean of three hand-written move counts. This stat flips the question: for one round (competition × round), if you took the best attempt-1 from anyone in that round, the best attempt-2 from anyone, and the best attempt-3 from anyone, what would the resulting mean be?

Concretely: aggregate attempt 1 across all competitors in the round to its MIN, same for attempts 2 and 3, then (min1 + min2 + min3) / 3. It is a measure of how easy the scrambles were, not how good any one solver is.

By the numbers

mean-of-3

Aggregation

Per-attempt minimum across the round, then averaged

`333fm`

Only event

FMC is the only hand-written move-count event in WCA

100

Rows returned

Top 100 rounds by mean ascending

整数

Integer values

FMC values stored as integer moves, no ×100 like timed events

Data source

Joins results and result_attempts. Groups by (competition_id, round_type_id); per attempt slot (1, 2, 3) it takes MIN(value) filtered to value > 0 (drops DNF/DNS/unattempted). Then (min1 + min2 + min3) / 3. 333fm move counts are stored as integers — no ×100 like timed events.

sql

SELECT (best1 + best2 + best3) / 3 AS mean,
       best1, best2, best3, competition_id, round_type_id
FROM (
  SELECT
    MIN(CASE WHEN ra.attempt_number = 1 AND ra.value > 0 THEN ra.value END) best1,
    MIN(CASE WHEN ra.attempt_number = 2 AND ra.value > 0 THEN ra.value END) best2,
    MIN(CASE WHEN ra.attempt_number = 3 AND ra.value > 0 THEN ra.value END) best3,
    r.competition_id, r.round_type_id
  FROM results r
  JOIN result_attempts ra ON ra.result_id = r.id
  WHERE r.event_id = '333fm'
  GROUP BY r.competition_id, r.round_type_id
) t
WHERE LEAST(best1, best2, best3) IS NOT NULL
ORDER BY mean
LIMIT 100;

Algorithm / pipeline

Bucket by (competition, round)

results joined with result_attempts, filtered to event_id = 333fm. Each attempt row is dropped into the bucket keyed by (competition_id, round_type_id).

Per-slot minimum

Within each bucket, run MIN(value WHERE value > 0) three times — once per attempt slot. This pulls out "the best attempt-1 by anyone in this round", and same for slots 2 and 3.

Keep only complete buckets

WHERE LEAST(best1, best2, best3) IS NOT NULL — all three slots must have at least one valid attempt. Rounds where some attempt slot is all-DNF / all-unattempted are dropped.

Compute potential mean, top 100

mean = (best1 + best2 + best3) / 3, ordered ascending, LIMIT 100. Note: no single solver may ever have posted this mean — it is a virtual best assembled from three different people, one per slot.

Key formulae

Formula

potentialMean(c, r) = (mₐ + m_b + m_c) / 3, mₖ = min{vₖ(p) : p ∈ round(c, r), vₖ(p) > 0}

c = competition, r = round, p = competitor, vₖ(p) = solver p's move count on attempt slot k. mₖ is the per-slot minimum across the whole round.

Caveats & edges

"Potential" is the key word — the three per-slot best results may come from three different people, so this mean is not necessarily one any individual ever achieved.
Only value > 0 counts: DNF (-1), DNS (-2), unattempted (0) are dropped. A round where some slot is entirely DNF cannot appear.
For FMC the value column stores integer move counts directly — there is no ×100 encoding. Do not multiply by 100 or the displayed mean will be off by two orders of magnitude.
MIN, not MAX: FMC is lower-is-better, unlike BLD points which are higher-is-better.