i hate to complain

I used to compete in breakthrough...I have noticed that the robots are taking over.. To me this is like a guy riding a motorcycle in the Boston marathon... It kind of ruins the whole thing...that is my 2 bits worth...

Which bots do you mean, except of wanderer_c ?

if it's clear in the name of the account, that the user is a bot.
i think is acceptable.

I never used the blacklist, but what happens if a listed user qualifies for the same tournament?

When everything is clear with the bot it seems ok. I think it is interesting to find the moment, when bot beats men. Probably wanderer will withdraw from tournaments, when asked to do so. But still it would be nice to watch him for example every third ch.

The real pain is when some people start using programs and deny doing it. It can destroy the competition as it was in the case of LoA.

At the moment Wanderer_c is still below the very top.
When this changes, the following might be an interesting
setup:
Bots no longer in the Championships, but the winner
of the Championships plays an exhibition match against
the best bot.

Ingo.

I think that in games like breakthrough, bots should stay away from the championships. I think the bots are a neat addition to this site, but for most games, I would prefer if they stayed out of the championships and sticked to RTs and such. That way, it'd be easy to aviod playing against the bots if you so desire.

I agree with Marius, but would go further and say that bots should not play in all Championships (regardless of their standard). I think it would be a mistake to have a complex rule saying "bots allowed here but not there" - a clear, simple rule is better for everyone. Ingo's idea is great, but there would need to be a way of viewing the exhibition match afterwards otherwise it will be lost (YouTube video?) :o)

There are plenty of opportunities for programmers to test their engines, for example the MCs and RTs.

Anybody know the answer to Mirko's question? I think you are forced to play them, but if they write any comments they won't appear above the board, but I'm not 100% sure.

Yeah, I think you're right, Alain. You can't avoid meeting someone in tournaments just by blacklisting them. The only effect is that you won't see any in-game messages from them, I think.

As for the first part: Clear rules are simpler, but not necessarily better. Let's go in the other direction. What happens if the general rule allowed bots in RTs and MCs, but not in championships - for ALL games? I think that would spoil RTs and MCs in at least Chess and Reversi, for sure.

I think the opportunity to play a bot against the best humans is good for the development of AI algorithms by legitimate computer science researchers. LG graciously provides this opportunity. For better or worse, in at least some of the games, the best humans play only in the championships.

My biggest problem with robots is some players work there way up to the championship and dont get the glory ? because of a machine.....this could happen in breakthrough.....

Playing against top-notch bots is a great opportunity for us to hone our skills and be humble. If bots can consistently outsmart the best human players, we shall embrace the fact that many classic games are computationally shallow. Why don't we celebrate the amazing advancement of artificial intelligence and move on to other games?

I'm not suggesting that bots shouldn't be able to play on LG. I just agree with the OP that there should be at least one format (the Championship is the obvious example) which is bot-free.

@Marleys: I think your assertion that the "(in some games) the best humans only play in the championship" is only partly true. For one thing, I doubt whether the very best human mind in any game plays on LG at all (no disrespect intended). This is besides the point, though - no-one is suggesting that bot-writers cannot hone their engines or that humans cannot hone their skills by playing bots...just not in the championship.

@z: Even though Draughts is now solved and chess is effectively "dead" (by the definition of whether the top players can beat the top programs), these are still immensely popular games, bringing pleasure to millions. Neither game is computationally shallow. There is room to celebrate both the amazing advancements of AI *and* for there to be a format here on LG that is bot-free. The two are not mutually exclusive :o)

I speak as player, I have only moderate interesting in AI.
While I understand that developing bots can be interesting, some people don't understand(or worse don't care) that developing bots harm games. Most people have "old fashioned" thinking and consider AI inherent worse than human mind, while I think that this is reverse and domination of natural man over machine is only due to poor technology.

"Playing against top-notch bots is a great opportunity for us to hone our skills and be humble"
I think I'm humble and I don't need bot that beat me. If I would need then I just play against eg. houdini in chess.

"If bots can consistently outsmart the best human players, we shall embrace the fact that many classic games are computationally shallow."
Problem isn't that games are computationally shallow. Problem is that often in games creativity is overwhelmed by infallibility so bot which don't make mistakes just won against human.

"Why don’t we celebrate the amazing advancement of artificial intelligence and move on to other games?"
Advance of AI in games hurt only games. Advance AI in general is not that easy to evaluate. It can bring potentially very terrible thing. It's not sure that small group of rich having robots will be friendly for rest of humanity(now they have to be because crowd is physically strong). It's not sure that AI will be always friendly for people. Moving to other games isn't so easy. Can you give me at least 1 game that for sure bot will not won championship in next 5 years? Even "unprogrammable" several years ago 19x19 GO now have bots with 3d+ strength.

Making bots with moderate strength are good for games with little popularity, but making bot to beat top humans is bad thing. It almost kill games with no "support"(tournament) over the board. Because in only place when it can be played there is no difference between cheater and master.
Even in games with great popularity in "real life" like chess engines better than people injure game. Good article that I recommend is "http://www.chess.com/article/view/computers-in-chess-the-conclusion".
Only good thing I find with bots competing here(I find no good with developing bots) in championship is that this give people advice what games cannot be regarded seriously because it's clear that game soon will have cheaters in competition.

I play amazons in the top league where there is also a bot playing, Arrow2_bot. It has won the championship once, but did not win the latest one. After the bot had won its first championship the creator, Martin Mueller, asked peoples opinion about its continued participation in the championship. I was the only one replying and I was in favour of letting it participate unless it keeps winning all the championships. I have only won against the bot once, but lost several times. Some of the players in the top league seem to have less problems winning against it than I have.
I think, with people behaving like Martin Mueller it is not a problem with bots in the championship since I am sure he will redraw the bot from the championship if amazons players ask him to do so.

People cheating is an entirely different thing.

I think that robots like wanderer_c should be allowed to play in the championships. My most recent game in the top championship just ended, and I had a nice game with the robot, and lost for the first time against it. Even though there might be a point where the computer is so strong that it will end up beating everybody, I don't think that the day will come soon. In a game like breakthrough, there's so many combinations, (although fewer than a more complex game like chess), that I think top players will be able to give a good shot at the bot.

The bots should be a test, a proof of worthiness for the human who's hopefully going to win, but should not just dominate, as that proves nothing.

Incidentally, I'm not happy with RoRoRo and naive child *both* being in EWN#1 (it was unforseen this championship, but is unavoidable next one), so will retire RoRoRo from the next championship (as naive child won this one). I hope that naive child will accept a 50-game challenge to see which one gets to play in the championship after that, on the presumption that their ratings will let them inject themselves straight into the top division.

robots against humans...

i don't feel it like a "threat"
i don't care if bots rulz the first division.

I would care if a bot dominates in a game. But so far, I don't think that has happened. Even if it does, I think the bots are wonderful to have around, if just not in a championship. But I think it needs to be taken on a case by case basis.

To be clear, I don't play in the top league in any game and don't really know if any particular bot *is* actually dominating. That was just the feeling I got. The decision should be up to those who it actually effects and who play at the top of those games.

@Phil: There would be even three bots (with jano_wol) in the 1st league of the EWN championship. However, I already retired naive_child_c from all tournaments. Unfortunately, my private laptop broke down some months ago and the program could not play for weeks. I took that occasion to quit playing on LG. The bot itself played on a few days only in the last 3-4 months. My girlfriend moved instead to avoid timeouts. She's nearly as strong as the program and will probably complete the open games of the bot and my own games probably too.

It could be that naive_child_c would win a single 50-points-match against RoRo but that would be luck. RoRo's the stronger bot and I think that jano_wol would be the more interesting opponent. This program seems to be very strong even if the bot has not yet finished so many games.

Suppose a talented human player consistently beats others in a particular game, would anyone here propose that we ban him from future championship tournaments because he had ruined the game? That doesn't sound fair, does it?

@YHW @Fatphil

if u guys are going to retire your bots, they deserve (at least) a "goodbye tournament"
A tournament just for bots.

@mungo

> I think, with people behaving like Martin Mueller it is not a problem
> with bots in the championship since I am sure he will redraw the bot
> from the championship if amazons players ask him to do so.

I agree, and am sure that Richard Lorentz, the "father" of Wanderer_c,
would do the same in Breakthrough.

Ingo.

vstjrt: it doesn't hurt a game for me if there exist programs which play better than I do, any more than it hurts a game for me if there exist humans which play better than I do. The important question for me is whether I find the game interesting and fun, not whether some other human or program plays better than I do.

@kpato: the "retirement" wouldn't be a real retirement, I just think that the tournaments should be primarily for the humans to prove how good they are. They only need one gatekeeper bot to prove that they're worthy, not more.

I do like the idea of a for-the-bots championship though. Could be the quickest one ever...

You could easily have for-the-bots TOURNAMENTS. They don't have to formally be championships. Just host a "bots special" RT! :-)

Sorry to be a bit off topic, but I was asking myself if we can learn something of the bots?

What does make Wanderer_c so strong, what part is skill (heuristics, evaluation function,...) and what part is brute force (how deep can he see in the end game) ?

Did he change the way people should play the opening, is he aware of zugzwang situation, is the trade-off between position and speed programmed or learned,... ?

Is it a strategical or a tactical game, or does the bot blur these territories?

Thanks for any answer from the specialists.

_"Is it a strategical or a tactical game, or does the bot blur these territories?"_

Maybe, in this particular area of AI development, the bots will rather _emphasize_ the difference. Bots will easily dominate in any tactical game nowadays. Only games that allow a significant degree of effective long time planning may yet outsmart them with players able to plan beyond a bot's horizon. But bots are continuously expanding their horizon. In the case of new games, bots will usually have an advantage to begin with for lack of sufficient human insight.

There have been some very interesting comments in this thread and I've been hanging back, planning a somewhat long post after this (Breakthrough) Championship was decided. But I think Carroll asked some key questions and I'd like to respond to them as best I can now, while the subject is still hot.

What makes Wanderer so strong? Heuristics? Yes. Evaluation function? Yes. Both are still rather primitive but evolving thanks to the many interesting games people are willing to play against Wanderer. I believe quite a bit more work is needed before Wanderer will land in the top 3. If ever....

Since Wanderer is MCTS based, evaluation and heuristics are implemented a bit differently from what many are used to in the traditional mini-max framework. But in summary, the evaluation is pretty much done by weighting pieces according to where they are on the board. Advanced pieces get higher values than those further back -- except pieces on the "traditional" defensive squares on the first row. These are given higher weights to discourage them from leaving those excellent posts. The evaluation also gives a bonus if a piece is advanced and there are few defensive pieces in front of it, but it is not clear if this really helps very much.

As for heuristics, we discourage captures in the beginning of the game (again, it is not clear if this really helps, but it is an attempt to prevent repetition of some humiliating losses suffered as a result of Wanderer's misguided desire to trade) and, of course, discourage moves that "hang" pieces. We also discourage moves to the edge of the board.

Did he change the way people should play the opening? I doubt it, but I leave that to the strong players that routinely beat Wanderer.

Is he aware of zugzwang? The short answer is "no". There is no code specifically dealing with zugzwang. But the way MCTS trees are built, such positions seem to be fairly easy to recognize.

I'm not sure what you mean by the trade-off between position and speed but as was pointed out to me by Stop Sign (if I remember correctly), Breakthrough is very much a game of speed and Wanderer tries to deal with this by keeping track of the game progress, shifting more towards tactics and less towards strategy as the game proceeds.

Speaking of strategy vs. tactics (I will discuss this in more detail when the Championship is over), humans are still much better at the strategic side of Breakthrough (look at the last 3 losses to edbonnet for 3 very good examples -- and for hints as to how to defeat Wanderer!) but are more likely to fall victim to tactical shots than Wanderer. I believe it is a game of both strategy and tactics (which, in my opinion, is why it is such an interesting AI project) and Wanderer very much distinguishes between the two -- though not all that successfully with strategy -- yet. :)

call for Sigourney Weaver, she'll know what to do!

bots suck big time,,,take the fun out of it...cheating is such a hollow victory...accomplishes nothing...waste of creative coding

Bots rock big time! With their fast play they make it much *more* fun.

(As long as they are parametrised such that they don't (a) play stupidly; or (b) play perfectly.)