Wanderer can now play Breakthrough. None of us are strong Breakthrough players so we are not sure how well it really plays. Thanks to klaashaas it is automated and will accept all Breakthrough challenges. If there are no objections we will also have Wanderer play some rated games (monthly cups and waiting room) to get a better sense of its abilities. It is being hosted on a fairly slow machine so it is currently set to think for as long as 10 minutes per move but will move faster as the game progresses. We look forward to these games so that we can get an idea about where to focus our attention for future improvements. Thanks for playing. Here's hoping for some interesting games!

Great stuff!

yes, sounds cool. Sign it up for some tournaments. I like how noone really knows how to play well :)

Yes, Thnikkaman, we agree that strategic concepts are hard to come by. The basic program strategy for the first 20 or so moves is: try not to do anything really stupid.

I've challenged it to a couple of games, it's played a few unusual moves early on but I can see how the early stage would be tricky for a bot to understand. I'm expecting to have to calculate hard later on though.

I will have NataliaBot play some games with you (as soon as I remember the password!).

It looks like our mini-match will finish 1-1, Wanderer has won one but I think I'm about to win the other.

It certainly played better later on, meeting all of my direct threats and playing very sensible looking moves. I still think its opening play could be improved, but a very impressive program indeed!

Thanks for playing and thank you for the kind comments. It's a work in progress and I believe it now plays the opening quite a bit better. The games you (and others) are playing against Wanderer are a huge help for us, pointing out program flaws and teaching *us* how to play a good game which we then try to pass on to Wanderer. Feel free to play more when the spirit moves you!

I just finished a game against wanderer, and have a question: What does the “Eval” number mean? At first, I assumed it was wanderers evalutated win probability, but seeing as it was reported to be over 70 just before my winning move, I guess that is not the case. Knowing what this number means, I may be able to provide some ideas as to how to improve the algorithm.

Yes, eval is supposed to roughly represent how well it feels it is doing. But you are right that it is often disturbingly off target. I've just uploaded the latest version and it seems to be a bit more accurate. (Rematch? :) ) Also, if it discovers it is sure to lose then the eval value can be overly optimistic (i.e., not 0). The fact that it often reports values in the 70's in such cases means that while it is calculating it thinks it's doing fine, and then it suddenly discovers that the opponent can force a win. Something is probably still wrong if it can be so happy just before discovering its certain doom! (Yes, I can fix the code so that it reports 0 when it knows it's doomed, but leaving it the way it is tells me more about what's going on when I replay the games. And I suppose there is no need to report to the opponent that information!)

I seems to me like wanderer was not very conserned about me getting a few stones quite far. I think it paid much less attention to those stones than it should. Also, I believe it should consider which stones I can use to reinforce such a position, how fast I can do it, and if such stones can be stopped.

After I got my forward position on the left, I paid almost no attention to the right hand side of the board. I focused on what I needed to get through wanderers defence in that area, and how I could reinforce my position to achieve that. In addition, I needed to calculate if, and how fast, wanderers most forward stones could achieve a victory, to make sure I would win a race, if it came to that. I think if wanderer were to put more emphasis on its and its opponents most forward stones, it could probably improve its play.

Just finished another mini-match with a 1-1 result. Marius's comments about the evaluation seem to apply to those games, in the game I won the evaluation never dropped to zero, even near the end of the game when the result was obvious. The evaluation did reach 100 in Wanderer's win though, quite a bit before the end and before I had spotted the win myself.

I suppose once Wanderer starts playing rated games there will be no need to display the evaluation, and I can see the point about not informing the opponent about their forced win. It doesn't really matter though, the important thing is to get it playing good moves, which it generally seems to do.

Wanderer is playing some rated games now, but I'm not sure which ones. Tournament games are easy to spot but when looking at a game in progress how does one distinguish an unrated invitation game from a rated “waiting room” game?

I don't think you can. Consider signing it up for the championship :)

It's currently playing some monthly cups. Maybe after we've had a bit more experience there (and have it playing better!) we'll be brave enough to try the championship!

If Wanderer plays against itself, does it win more with Black or White?

Interesting question, so I ran a test. I played 500 games at 10 secs / move. Though this is about 100 times faster than Wanderer plays on LG, it still seemed like it would be a good indicator.

Results: White - 254 Black - 246