search

axelahmer / lzcup-scheduler

This repository contains an Answer Set Programming (ASP) implementation using Clingo for scheduling the LZV Cup, an amateur indoor football league in Belgium.
0 stars 0 forks source link

Performance without closed games #5

Open ssardina opened 3 months ago

ssardina commented 3 months ago

With regard to "BTW, we never tested what happens if we totally remove the "close games"optimization. Do we do well without it?"

Have not tested in a proper capacity. It would just be a matter of commenting out the soft constraints and the optimize. I assume it would be very slightly better - intuitively because there is just less to solve per model. Let me know if you really care and I can write up a results table.

Originally posted by @axelahmer in https://github.com/axelahmer/lzcup-scheduler/issues/3#issuecomment-2160632275

ssardina commented 3 months ago

Correct, I have commented this in my code:

image

Now I am running instance 1, which has a cost of 2080, so 2 games not scheduled. It found the optional right away in seconds, but it is finding it hard to prove optimality (that is, prove that we cannot go lower than 2)

I don't understand your claim that it would be just slightly better; this is not just about "amount" of constraints right?

I thought that maybe planning for just unscheduled games would be much much faster, because it is not a numeric constraint and no-goods could get you to the optimal quick. In fact, it does, it gets immediately to 2, but then it gets stuck proving optimality, typical..