The OpenTafl Computer Tafl Open - An Abstract Strategy AI Tournament

[Fishbreath]

About two weeks to go! Deadlines and submission guidelines are available at the tournament website.

I also took some time to spin up an Amazon EC2 instance of c3.large size, on which all the AIs will be running. Here are OpenTafl's benchmark mode results for v0.4.4.7b:

1   2   3   4   5   6   7

./OpenTafl.sh --benchmark OpenTafl v0.4.4.7b log from Sat Dec 03 02:20:31 UTC 2016 on ip-172-31-5-180 Java version: 1.8.0_111 Brandub 7: finished in 27.429 sec, searching 1133609 nodes to depth 6 Tablut 9: finished in 36.718 sec, searching 1127203 nodes to depth 5 Copenhagen 11: finished in 6.562 sec, searching 23562 nodes to depth 4 Tablut 15: finished in 11.343 sec, searching 113556 nodes to depth 4

These results are directly comparable to the results of any other benchmark run using OpenTafl v0.4.4.7b, and are a good indicator of both single-core performance and memory performance. (For comparison, my laptop, with a recent Core i7m, is about twice as fast on Brandub through Copenhagen, and about 33% faster for Tablut.) Plan accordingly!

I'll be running the OpenTafl server on another EC2 instance in the same datacenter, to reduce the influence of latency on AI time usage planning. When tournament games are ongoing, I'll post the server address here so people can spectate.

[Fishbreath]

One week to go before the official deadline! As stated at the tournament website, I'll probably accept late entries up until the 22nd or 23rd when tournament play will start. I will test every entry on Amazon's hardware as soon as the entry comes in, though, so if I have trouble running it, earlier entries will have more time to get things figured out.

[Jono]

I'm having some trouble submitting by email -- gmail blocks runnable jars in attachments. Any ideas on how to get around this?

Games published by our own members! Check 'em out!

[Fishbreath]

Either a shared Google Drive or Dropbox item would be fine. If neither of those are possible, shoot me an email and I'll set up some other submission mechanism.

[Jono]

Righto, done via dropbox. Let me know if you don't get a message from dropbox (or have any issues running the AI).

[Fishbreath]

Looks to be working fine. Your evaluation function is very clever. In the interests of suspense, I haven't played it against OpenTafl yet, and I don't think I will until I'm doing all the tournament round robin stuff, but I played a quick game against it, and even with only 10 seconds, it seems to play pretty smart.

[Jono]

Cool, good to hear. Right now I'm pretty sure it is hard-coded to use 10s regardless of how much time it is told it has, so it won't get any better than that

[Fishbreath]

You can follow this tag at my blog for tafl tournament coverage.

[Fishbreath]

Today is the last day for late entries.

[Fishbreath]

Long story short: the only entry to arrive functional and on-time came from our very own Jono. He graciously agreed to a one-match playoff with OpenTafl for the title, rather than taking it outright as is his right. There's a liveblog up at soapbox.manywords.press, and if you have the OpenTafl client, you can spectate by pointing your client's server setting to taflopen.manywords.press.

Games published by our own members! Check 'em out!

[Jono]

I'd love to see the game records if you could make them available. Any comments on the games would be nice too since I don't play well and so don't pick up on a lot of the nuance (both good and bsd) in the moves.

[Fishbreath]

They're available in OpenTafl replay form here. Along with more in-depth overall analysis, I hope to do some annotations on them, too, possibly at the end of this week. (I have a friend with whom I play a regular tafl game, and it'll be a fun exercise to do a joint commentary.)

[Jono]

Wow, just saw the one where J.A.R.L. was attacking. Looks like it seriously broke, I'm not sure it should ever choose those last couple of moves even if it is given almost no time to think. I guess the pruning somehow blew up or something.

[Fishbreath]

It must have done—it played one or two moves without using anything like all of its time.

[Jono]

Oh. It does that when it can see the end of the game without search. It is deterministic so I'll should be able to recreate that game. Shame it failed to play prpperly on the day...

[Fishbreath]

I'm up for an exhibition game at a later date if you get it sorted out.

[Jono]

It's reproducable but definitely a bug of some sort. It plays fast after f6-i6 because it can see that state is a certain win for white. For some reason during the search, even with no pruning, it fails to see that this is a winning move for white.

I'll get the source code released first then see if I can find time to sort this out. Will keep you posted.

[Fishbreath]

I plan on running another tournament in 2018, provided there's sufficient interest.

[ddyer]

Tafl is an easy target.  Are you planning tafl again, or looking for a more
intrinsically difficult game to write an AI for?   

I've written a pile of AIs for Boardspace.net, but the games I haven't
succeeded in writing a good enough AI for are Volo, Plateau, and Twixt.
...Also Go, but that's another story.

[Fishbreath]

I'm planning tafl again; it's less an interest in abstract strategy AI generally and more an interest in the intersection of AI with tafl. Also, I've already written all the tournament tools I need for tafl.

I don't see it as an intrinsically easy target—certainly, it's not go hard, but it's mathematically harder than chess. (Assuming 11x11 boards, in terms of branching factor and state space, by my calculations.) It's much less well-studied, and also very unlikely to attract attention from the big machine-learning shops given its lack of popularity.