Bowling league schedules: keeping a team off a lane pair without breaking the rotation

The free bowling schedule maker now keeps a team off a lane pair every week as a hard rule. Measured on 142 seasons against the usual hand fix.

Plenty of bowling leagues have one team that cannot use one pair of lanes. A bowler with a walker has no room on the approach of the end pair by the wall. A team asks to stay away from the pair by the bar. The bowling league schedule page already rotates every team across the lane pairs as evenly as the season allows. Today it also takes lane limits: one line per team, such as Pin Pals: 7-8, Spare Me: last or Gutter Gang: ends for both end pairs. That team never lands on that pair, in any week.

The usual fix, and why it drifts

The way an organiser does this by hand is to build the even schedule first. Then, each week the team lands on its pair, they swap its match with the one next door. That keeps the limit, but every swap moves two matches, four teams, onto different pairs, and nothing puts the counts back. We ran that fix on 142 seasons (6 to 24 teams, 26 to 34 weeks, four kinds of limit). It left some team off its fair share in every one of them, 3,345 counts off in total. In 3 weeks it found no swap that kept the limit at all.

Limits inside the search

The page's lane search has three stages: a week-by-week assignment (the Hungarian method), a season-wide re-solve of each week with the others fixed, and a focused random walk for the last few counts. A limit is now a hard rule in all three. A banned match-on-pair cell costs more than every balance price combined, so the exact week step never takes one when the week has any legal deal. Because that step is exact, a banned cell in its answer is also a proof that the week has no legal deal. Then the page refuses the season and names the week, rather than quietly breaking a limit. (In a 4-team league, three teams avoiding the same pair is enough.)

"Even" also needs a new meaning. The avoided pair still hosts a match every week, so the other teams have to bowl on it a little more often. The page works out each team's fair share by scaling a matrix: every team gets its number of weeks, every pair gets its number of slots, and banned cells get zero. The target is each share rounded down or up. With no limits that reduces exactly to the old rule, and the 48 seasons we compared came out identical to before.

How it measured

every team in its fair sharecounts offback-to-back repeatslimits broken
hand fix0 of 142 seasons3,3453,4383 weeks
page133 of 142242,3380

It takes about 0.8 s a season in the browser. A SAT solver checked the misses. Four (6 teams on 3 pairs with two teams off the end pair, or one team off both ends) cannot reach the target at all, so the page's result is as close as any deal gets. Four more, all 20 teams over 30 weeks, have a deal the search did not find. That is the same season length that needed a stored solver answer before limits existed, and it is the next thing to close.

One gap stays manual. Position rounds (1st v 2nd by the standings that night) put places on pairs in order, and their teams are known only on the night. The page says to check a limited team's pair on those weeks.

The scripts and full tables are in the project's eval/bowling-lane-avoid.txt, and the page is free and runs entirely in your browser.

All writing · More in Tools and projects · More on League Schedule Generator · All tools and research