Reduced board sizes

[thisismyrobot]

The bot I submitted was state-less and based on the sort of tactics that I would use playing Tron.

This worked really really well on the large maps as my challenger would be still building a game tree whilst I was blocking them off. These tactics allowed me to peak as high as 10th overall.

Then, as the map sizes got reduced, I lost more and more as my challengers would have filled out their tree before I got close to a winning position.

Wouldn't it have been better to stick with larger board sizes so that competitors could be graded on how well they handle imperfect information. Otherwise, it seems that the solution to the game is based on the level of available resources (for the problem size), not clever, innovative and possibly new solutions. This could have possibly been achieved by reducing the available time per move, increasing the board size or dividing the Elo rating by the average time taken per move.

Anyway, I'd be really interested in other people's thoughts on this, please feel free to agree or disagree vigorously

[amstan]

I'm sorry about that. I wanted bigger maps as well(I have used 100x50 maps in those youtube videos).

Problem with bigger maps is the amount of time that games take. The maps that we used took 2 minutes on average to fill(17x17 average), if we would of doubled the size(to let's say 35x35) we would of got 8 minutes per game, which was not acceptable.

[thisismyrobot]

Hey amstan,

Thank you for your reply, I certainly understand the catch-22 you were faced with - it must have been hard trying to juggle the practicalities of having 800 competitors with the need to return results within two months of Friday night...

Anyway, you guys hosted a great contest, thank you

[Fritzlein]

please feel free to agree or disagree vigorously

I'll agree that large boards are a better measure of AI, but disagree vigorously with with part of what you wrote.

This worked really really well on the large maps as my challenger would be still building a game tree whilst I was blocking them off. These tactics allowed me to peak as high as 10th overall.

Then, as the map sizes got reduced, I lost more and more as my challengers would have filled out their tree before I got close to a winning position.

I would hypothesize that your rank dropped almost entirely due to improved competition, not due to changing maps that made the static, unchanging competition perform better relative to you. People who submitted a searcher and then didn't improve it experienced a decline of rank over time as well. There was a tremendous increase in the general level of play over the course of the contest.

This is not to say that you need to give in and do what everyone else is doing. It's good to have a variety of approaches. We need rebels and iconclasts. I also think it is reasonable to guess that you would have ranked higher if the finals had been entirely 30x30 maps. I encourage you to enter wtarle's ongoing contest to get a chance to test your theories on larger boards. Nevertheless I highly doubt that any bot which isn't a searcher could compete for the top ranks. It is useful to look ahead even when the search is slowed down by (say) a factor of ten due to a larger board.

I can't prove that you couldn't have ranked among the top bots on larger boards without search, but I think the preponderance of evidence is with my skepticism, so the onus is on you to prove that you can can beat searchers on large boards. The contest continues, so go get 'em!

[thisismyrobot]

Hey Fritzlein,

Thanks for the honest feedback, this has to be the first forum I've actually enjoyed posting on (No childish off-topic rubbish)

I guess it is really "put your money where your mouth is" time, so where do I go for wtarle's ongoing contest? I searched the boards for the username and only found your mention here and as the author of a deleted post.

[Fritzlein]

Hey Fritzlein,

Thanks for the honest feedback, this has to be the first forum I've actually enjoyed posting on (No childish off-topic rubbish)

It is generous of you to compliment my personality defect.

I guess it is really "put your money where your mouth is" time, so where do I go for wtarle's ongoing contest? I searched the boards for the username and only found your mention here and as the author of a deleted post.

Sorry. Instead see this page and this thread for ongoing competition. Dhartmei's larger boards truly do change the nature of the competition. Best of luck!