Home-and-away seasons for 11 and 13 teams: less carry-over and fewer back-to-back games

Home-and-away seasons for 11 and 13 teams now start from a capped shift starter: lower carry-over and fewer breaks. A 15/31-team idea changed nothing.

The league schedule generator can build a home-and-away season (a double round robin) that does two things. It spreads out carry-over: if you play the strongest team, you don't want to keep meeting the same side the week after. It also keeps breaks down. A break is two home games in a row, or two away games in a row. The search scores a season as carry-over + 10 × breaks and keeps the lowest.

Earlier today the single round robin gained "capped shift candidates". The search takes a built-in starter timetable whose home/away pattern rotates with the rounds, then lowers its carry-over without letting its breaks go up. The home-and-away search didn't use these yet. Now it does: each capped candidate can be leg one, and leg two is searched against it as usual.

What changed

We ran every size from 9 to 40 teams on two seeds. Each season was checked as valid, and every number was recounted from the rounds:

TeamsSeedBefore (carry-over / breaks)NowScore
111497 / 23481 / 22727 → 701
112509 / 21505 / 19719 → 695
131733 / 35721 / 321,083 → 1,041
132739 / 32721 / 291,059 → 1,011

In all four rows both numbers went down, so nothing was traded away. At every other size the score did not change, and the extra searches added 30 to 40% to the run time. So the shipped default only tries capped leg ones up to 16 teams. From 17 teams up the search runs exactly as fast as before.

What didn't work

Leagues of 15 and 31 teams are "field sizes": with a bye they make 16 and 32, powers of two, where the scheduler starts from a perfectly balanced field timetable. We tried the 16- and 32-team shift starters as leg ones there. Nothing changed on either seed, and the run took twice as long. Two things explain it. Leg one of a field-size season already comes from the single search, which picks the 31-team shift starter when it wins. And a perfectly balanced leg one, played again with grounds swapped, is very hard to beat. The option stays in the code as fieldShift and is off.

The method, the full table and the script are in the venture's README ("Shift candidates in a home-and-away season, part two") and eval/shift-double-capped.mjs.

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