[phpBB Debug] PHP Warning: in file /includes/bbcode.php on line 112: preg_replace(): The /e modifier is no longer supported, use preg_replace_callback instead
AI Challenge Forums • View topic - Final Tournament Rules

It is currently Mon Jul 16, 2018 10:29 pm Advanced search

Final Tournament Rules

Random stuff about the contest, posts that don't fit in the other forums.

Final Tournament Rules

Postby jeff.cameron » Sun Feb 14, 2010 7:31 am

Here is a first draft of the final tournament rules: http://csclub.uwaterloo.ca/contest/final_tournament.php

I would like to invite feedback from anyone on these rules. Do they make sense to you? If not, you can change them! Post your feedback.
jeff.cameron
Contest Organizer
 
Posts: 91
Joined: Sun Jan 31, 2010 4:06 am

Re: Final Tournament Rules

Postby ebrahim » Sun Feb 14, 2010 6:27 pm

Two suggestions:
1. Run both on-going and final tournaments on random maps. Because:
* I think our goal is to design and implement AI. If it is an AI, it should be able to run his strategy for a class of maps, rather than a single hard-coded one.
* It is really not fun to program (or program against) some bot that contains "if map is X, perform this unbeatable hardcoded strategy".
2. Play each map both sides, regardless of being symmetric or asymmetric. I've explained the reason in this post: viewtopic.php?f=10&t=177
ebrahim
Lieutenant-Colonel
 
Posts: 49
Joined: Mon Feb 08, 2010 7:03 pm

Re: Final Tournament Rules

Postby oldherl » Mon Feb 15, 2010 3:50 am

I suggest that some random maps should be used not only in the final, but also in the everyday matches.
Some bots hard-code strategy against the map :If it is map X, then do Y. That is not real AI!
Random maps can lead them to failure.
oldherl
Cadet
 
Posts: 3
Joined: Wed Feb 10, 2010 1:28 am

Re: Final Tournament Rules

Postby amstan » Mon Feb 15, 2010 4:43 am

Yeah, that would be awesome, i'm trying to convince Jeff to use them.
Alexandru M. Stan
Contest Organizer
User avatar
amstan
Contest Organizer
 
Posts: 691
Joined: Sun Jan 31, 2010 4:02 am
Location: Stoney Creek, Ontario

Re: Final Tournament Rules

Postby txandi » Mon Feb 15, 2010 1:03 pm

I agree with that as long as *everyone* plays on the same randomly generated maps.
txandi
Cadet
 
Posts: 9
Joined: Sat Feb 06, 2010 3:11 pm

Re: Final Tournament Rules

Postby kduleba » Mon Feb 15, 2010 8:43 pm

You might want to look into swiss-system tournaments to work around infeasibility of scheduling O(n^2) games. It converges quite quickly and will probably be better than ad-hoc solutions of suspending lowest ranked entries.
kduleba
Captain
 
Posts: 27
Joined: Fri Feb 05, 2010 12:35 am

Re: Final Tournament Rules

Postby nickponline » Tue Feb 16, 2010 3:55 am

nickponline
Cadet
 
Posts: 1
Joined: Tue Feb 16, 2010 3:44 am

Re: Final Tournament Rules

Postby bruudruuster » Tue Feb 16, 2010 9:08 am

Hi All,

My remarks on the proposal for rules:

1) I would also insist to consider swiss tournament systems. It may be some hassle for you to set it up, but this is really the best shot at "precise rankings within 30 minutes", as you promise. Note that you can run the swiss tournament initially at N rounds and if you find you have more time left increase N at will. It also lets all bots play, which should make it interesting for all participants.

2) I suggest selecting a fixed set of random maps (e.g. 10) for the final and let a pair of players play on all those 10 maps. That gives you the most precise rankings, because you don't know in advance which random map is harder to play and playing hard maps against weaker bots gives you an advantage over playing easy maps against weaker bots.

3) I suggest selecting more large maps for the simple reason that all symmetric maps are fundamentally a draw. Weak bots may do well on reaching a draw on small maps but they will have a hard time on the larger maps. The small maps used so far are trivial for the strong bots. I am sure that larger maps are needed to distinguish between the top 20 or so bots; I have the impression that the ranking produced until recently is kind of random at the top.

Let's hope for an interesting final ;-)

Cheers!
bruudruuster
Cadet
 
Posts: 6
Joined: Fri Feb 12, 2010 10:13 pm

Re: Final Tournament Rules

Postby ruibarbosa » Tue Feb 16, 2010 12:33 pm

I think that once the general rankings of the tournament settle, you should make the top 10 bots play a final winner-takes-all tournament, where each bot plays all the other top bots several times in different maps... This would eliminate the randomness associated with the bigger tournament (where a bot might beat another bot in a given map, even though it would lose against it in other maps).
ruibarbosa
Cadet
 
Posts: 5
Joined: Wed Feb 10, 2010 1:31 am

Re: Final Tournament Rules

Postby Savaron » Tue Feb 16, 2010 1:09 pm

Why is the delivery of results in 30min so important to you? In my opinion generating correct results, especially of the top X should be your highest goal, whatever time it takes. Most of us spent a month on this challenge and I certainly don't need the results 30min after the deadline.

Use bigger maps and play those several time for each bot and take your day off to compute the results...
Savaron
Captain
 
Posts: 22
Joined: Tue Feb 09, 2010 10:32 pm

Next

Return to Misc

Who is online

Users browsing this forum: No registered users and 2 guests

cron