LPL · measured by record gap late
Like Plays Like
Supports SORT Like Plays LikeERL Early Results Shape Late OpponentsFVA Fixed Venue, Adaptive OpponentPHS A Gumbo Game Has StakesSTK Something Stinks
Context
Nine games is a small sample, and when a small sample cannot grow, the way to make it better is to choose the observations. The game between two 6-1 teams in November tells you something. The game between a 6-1 team and a 1-6 team tells you what you already knew, and it costs the same Saturday.
Problem
Late schedules drawn by rotation waste games on mismatches. The contenders are not separated from each other by November, because they never played each other, and the strugglers never got a game they could win.
Forces
- Sorting pulls against rematches. Match the top by record often enough and they run out of fresh opponents.
- Sorting pulls against rivalry. A rival's record has nothing to do with whether you play them.
- Sorting pulls against venue. Matching by record while honoring a published venue template takes a solver, because a bracket cannot do both.
- Sorting pulls against the league's bids. The better the top sorts itself, the more losses the top absorbs.
- Coarse pulls against fine. Pools of four sort coarsely, and week-by-week pairing sorts finely but needs a tiebreak culture the sport does not have yet.
Solution
After an opening stage, pair schools whose results so far are as alike as possible, subject to no repeats, no reserved rivals, and the published venue template. Do it round by round so each pairing uses the newest information. Publish the pairing rule in words a fan could apply at the kitchen table.
Therefore
Let standing choose the opponent in every adaptive slot, under a published rule.
Measure
The mean absolute difference in conference win count between opponents in the last three adaptive Saturdays, compared with the same statistic under the published rotation.
Lore
Scotland's top league splits in two after thirty-three rounds and the top six play each other down the stretch. The rank-banded pool is the same idea for football: the four pool winners form a new pool, the four runners-up form another, and so on down. The good play the good and the bad play the bad, and every one of them has a game it can win.
Supported by
WPL Winners' Path, Losers' PathERL Early Results Shape Late Opponents
Pulls against RRP Reserved Rivalry PairFOP Fresh Opponent PreferredBID Protect the Bids
Seen in
| scenario | language | simulated | value |
|---|---|---|---|
| status-quo | × breaks | · n/a | 2.119 |
| pool-play | + honors | + honors | 1.289 |
| two-cups | + honors | + honors | 1.324 |
| two-cups-b | + honors | + honors | 1.363 |
| four-cups | + honors | + honors | 1.303 |
| pools-playoff | + honors | + honors | 1.320 |
| mini-playoff | + honors | + honors | 1.369 |
| chase | + honors | + honors | 1.510 |
| seven-settle | + honors | + honors | 1.449 |
| seven-final | + honors | + honors | 1.557 |
| seven-semis | + honors | + honors | 1.453 |
| eight-final | ~ strains | ~ strains | 2.055 |
- status-quo: breaks. Opponents are set years ahead.
- pool-play: honors. Rank bands sort by construction.
- two-cups: honors. Classification brackets pair equal records every round.
- four-cups: honors. The same, plus a consolation pool.
- pools-playoff: honors. The playoff and Gumbo games pair neighbors; November gaps are a fifth narrower than the rotation.
- mini-playoff: honors. November opponents are closer in record than the rotation gives, by a fifth.
- chase: honors. Swiss pairing on the standings.
- eight-final: strains. Every late opponent but the final is published, so the table chooses only the two finalists.
See it: status-quopool-playtwo-cupstwo-cups-bfour-cupspools-playoffmini-playoffchaseseven-settleseven-finalseven-semiseight-final