Longest streak of comps with a PR

Person

Walk a cuber's comps in order — as long as a comp produces at least one PR (any event's single or average), the streak ticks up; one comp without a PR resets it.

A "sustained improvement" metric — you don't need to be top in any event, but you need to extract a fresh PR from somewhere at every comp.

By the numbers

single + avg

PR types

Either resets +1

Top 100

Leaderboard depth

Sorted by streak desc

≤ 历史最佳

PR check

`val <= pb` — ties count

Data source

Full results joined to persons / competitions / round_types, sorted by start_date, round_type.rank. Each row has event_id, best (single), average. Any round breaking the per-event PB counts the whole comp as "had a PR".

sql

SELECT person_link, competition_link,
       event_id, best single, average
FROM results
JOIN persons person ON person.wca_id = person_id AND person.sub_id = 1
JOIN competitions competition ON competition.id = competition_id
JOIN round_types round_type ON round_type.id = round_type_id
ORDER BY competition.start_date, round_type.rank

Algorithm / pipeline

Bucket by person → comp

Outer bucket: person_link. Inner bucket: gather all event-rounds at one comp into a single "comp snapshot".

Track per-event PB

pbsByEvent holds single-PB and average-PB per event (both start at Infinity); iterate every result in the snapshot.

Decide "comp had a PR"

Any event × type where val > 0 && val <= pb fires the refresh and flips competitionWithPb. <= means ties count.

Advance / reset

PR found → count += 1, refresh longest; no PR → close current, open fresh on next comp.

Caveats & edges

Ties count (we use <=); strict-improvement would need < — looser here.
Any event × any type PR makes the whole comp a "PR comp"; 6-event PRs and 1-event PRs score the same.
best = -1 / 0 (DNF / not attempted) cannot trigger a PR — val > 0 gate.

Longest streak of comps with a PR

By the numbers

Data source

Algorithm / pipeline

Caveats & edges

Related stats & links