Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> General Discussion >> Cyborg vs Human Arimaa
        
(Message started by: rbarreira on Apr 18th, 2012, 8:14am)
Title: Cyborg vs Human Arimaa
Post by rbarreira on Apr 18th, 2012, 8:14am
Cyborg = Bot-assisted human Arimaa player

Has there ever been an Arimaa game involving Cyborgs? I sometimes think it would be interesting because despite the fact that my bot is much stronger than me (mainly due to tactics and overlooking responses by the bot), I can still often see that it's going in a wrong strategic direction, and I'm a weak player myself...

We can see a cyborg in two different ways - a bot with someone to correct some of its strategic flaws, or a human with a tool to correct their tactics, check for blunders and avoiding blind spots. One or the other perspective would be more accurate depending on the situation (on a very tactical endgame the human might resort to simply playing the move suggested by the bot).

At a live time control it might be hard to play in this way (at least until we start having neural-computer interfaces), as it requires some back and forth checking of different lines in order for the cyborg to show their power. But at a postal time control (1-7 days per move) it looks very feasible.

Does anyone else think this sounds interesting? Just another random idea that has been kicking around in my head for some time...
Title: Re: Cyborg vs Human Arimaa
Post by Fritzlein on Apr 18th, 2012, 10:09am
I am also interested in this question, because I believe chess cyborgs had some success in live "freestyle" tournaments.  And hasn't computer analysis already proved fruitful for Arimaa in the postal Mob games?
Title: Re: Cyborg vs Human Arimaa
Post by mistre on Apr 18th, 2012, 10:12am
Very interesting idea.  Yes, this would definitely work best in a postal setting.

I envision something like the bot supplying its top 3 move options and then the player has the option to choose one of them or choose their own move. At the end of the game, we could show the percentage of time the player chose their own move/vs a bot move and also see the choices on a move by move basis.

Ideally the bot should have about an equal playing strength to the player.  My hunch is that a player will decide to choose a bot move increasingly more often towards the end game.

I would volunteer for such an experiment!
Title: Re: Cyborg vs Human Arimaa
Post by Katsunami on Apr 18th, 2012, 12:24pm

on 04/18/12 at 10:09:49, Fritzlein wrote:
I am also interested in this question, because I believe chess cyborgs had some success in live "freestyle" tournaments.


Yes, indeed. Even while the best chess engines play at supergrandmaster strength, the strength still improved when assisted by humans. The improvement is mostly strategic. The other way is also true. Humans improve tactically when assisted by a chess engine. The combination of a top grandmaster and a top engine is consistently stronger than either alone. It's estimated that such a combination of a 2800 ELO player assisted by a top engine will have a 3000+ rating.

I don't expect it to be different for Arimaa; therefore I think rbarreira's line of thinking is correct.
Title: Re: Cyborg vs Human Arimaa
Post by Adanac on Apr 18th, 2012, 12:57pm
This article highlighted some interesting results from a cyborg chess match.  Apparently knowing how to use your bot is more important than having the best bot or being the best player.

http://www.collisiondetection.net/mt/archives/2010/02/why_cyborgs_are.php

Lured by the substantial prize money, several groups of strong grandmasters working with several computers at the same time entered the competition. At first, the results seemed predictable. The teams of human plus machine dominated even the strongest computers. The chess machine Hydra, which is a chess-specific supercomputer like Deep Blue, was no match for a strong human player using a relatively weak laptop. Human strategic guidance combined with the tactical acuity of a computer was overwhelming.

The surprise came at the conclusion of the event. The winner was revealed to be not a grandmaster with a state-of-the-art PC but a pair of amateur American chess players using three computers at the same time. Their skill at manipulating and “coaching” their computers to look very deeply into positions effectively counteracted the superior chess understanding of their grandmaster opponents and the greater computational power of other participants. Weak human + machine + better process was superior to a strong computer alone and, more remarkably, superior to a strong human + machine + inferior process.
Title: Re: Cyborg vs Human Arimaa
Post by Katsunami on Apr 18th, 2012, 2:10pm
So.... Strong player + powerful computer + optimized process = epic playing. That means that both humans and computers can still learn a lot, even in Chess.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Apr 18th, 2012, 3:12pm
I could be wrong, but I feel that in Arimaa even a weak player can make a good combination with a bot. I probably would not be able to avoid all of my bot's weak strategic moves, but still many of them. Would that be enough to give a strong player a hard time, I don't know...

If someone wants to challenge the combination of me + my bot some time, we are available.
Title: Re: Cyborg vs Human Arimaa
Post by Katsunami on Apr 18th, 2012, 4:51pm
You saw that Briareus was able to give chessandgo a hard time on it's own, because the bot is such a great defender.

I think that a combination of a strong bot such as Briarius and a human, to supplement the bot's strategic play, is already stronger than the best humans of this moment.
Title: Re: Cyborg vs Human Arimaa
Post by hyperpape on Apr 19th, 2012, 8:51am
I suspect a lot of the trick would be interface related: creating a tree of variations, figuring out which moves the bot thinks are strong resistance, being able to easily say "look deeper at this line", etc.

Of course you can manually do this from logs and inputting positions as strings, but I imagine that it would be slower and more difficult than it would be with a good interface.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Apr 19th, 2012, 9:16am

on 04/19/12 at 08:51:53, hyperpape wrote:
I suspect a lot of the trick would be interface related: creating a tree of variations, figuring out which moves the bot thinks are strong resistance, being able to easily say "look deeper at this line", etc.

Of course you can manually do this from logs and inputting positions as strings, but I imagine that it would be slower and more difficult than it would be with a good interface.


So far I have found that the plan window suffices for doing this kind of analysis with the bot.

The only thing I feel is missing is a way to exclude certain moves from analysis, which is useful both for finding the "second best", "third best" moves etc. and also to stop the bot from sticking to bad moves that it happens to like. If someone wanted to play a game like this against me I would definitely implement this and use it in my analysis.
Title: Re: Cyborg vs Human Arimaa
Post by Fritzlein on Apr 19th, 2012, 10:59am

on 04/19/12 at 09:16:00, rbarreira wrote:
So far I have found that the plan window suffices for doing this kind of analysis with the bot.

The only thing I feel is missing is a way to exclude certain moves from analysis, which is useful both for finding the "second best", "third best" moves etc. and also to stop the bot from sticking to bad moves that it happens to like. If someone wanted to play a game like this against me I would definitely implement this and use it in my analysis.

Would you be interested in assisting briareus in our Postal Mixer game?  I am curious what it would feel like, and I don't mind the hit to my WHRP if I lose.
Title: Re: Cyborg vs Human Arimaa
Post by Dolus on Apr 19th, 2012, 11:35am
We should try a Cyborg vs. The Mob game. :)

One of our top players teamed up with a top bot, against the community.
Title: Re: Cyborg vs Human Arimaa
Post by Hippo on Apr 19th, 2012, 11:48am

on 04/18/12 at 08:14:17, rbarreira wrote:
Cyborg = Bot-assisted human Arimaa player

Has there ever been an Arimaa game involving Cyborgs? I sometimes think it would be interesting because despite the fact that my bot is much stronger than me (mainly due to tactics and overlooking responses by the bot), I can still often see that it's going in a wrong strategic direction, and I'm a weak player myself...

We can see a cyborg in two different ways - a bot with someone to correct some of its strategic flaws, or a human with a tool to correct their tactics, check for blunders and avoiding blind spots. One or the other perspective would be more accurate depending on the situation (on a very tactical endgame the human might resort to simply playing the move suggested by the bot).

At a live time control it might be hard to play in this way (at least until we start having neural-computer interfaces), as it requires some back and forth checking of different lines in order for the cyborg to show their power. But at a postal time control (1-7 days per move) it looks very feasible.

Does anyone else think this sounds interesting? Just another random idea that has been kicking around in my head for some time...


Nice idea ... Tize could dominate this field.
I would still concentrate on developping standalone bot.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Apr 19th, 2012, 1:16pm

on 04/19/12 at 10:59:28, Fritzlein wrote:
Would you be interested in assisting briareus in our Postal Mixer game?  I am curious what it would feel like, and I don't mind the hit to my WHRP if I lose.


That would be a bit messy as I would have to hack the bot to let me input a move for a game, and the postal controller to avoid thinking for your game by itself. It would be error prone too, any mistake on my part and the bot would lose by illegal move.

I agree it would be interesting though. Maybe we can set up a new postal game for it? I can even create an account starting with cyborg_ instead of bot_. The regular postal time control would be fine but a slightly slower one would be a bit more relaxed for me. What would you prefer?
Title: Re: Cyborg vs Human Arimaa
Post by Fritzlein on Apr 20th, 2012, 7:45am

on 04/19/12 at 13:16:08, rbarreira wrote:
What would you prefer?

Perhaps I should wait until I demonstrate that I can handle the postal games I already have before starting anything new.  At the moment I really shouldn't take on new commitments.  I hope you get a chance to try out the experiment somehow, though.
Title: Re: Cyborg vs Human Arimaa
Post by arimaa_master on Apr 20th, 2012, 8:12am

on 04/19/12 at 13:16:08, rbarreira wrote:
That would be a bit messy as I would have to hack the bot to let me input a move for a game, and the postal controller to avoid thinking for your game by itself. It would be error prone too, any mistake on my part and the bot would lose by illegal move.

I agree it would be interesting though. Maybe we can set up a new postal game for it? I can even create an account starting with cyborg_ instead of bot_. The regular postal time control would be fine but a slightly slower one would be a bit more relaxed for me. What would you prefer?


Since Fritzlein is on his full capacity already I would be honored to play such postal game against cyborg_briareus.

However only if you think that I am a decent match for such game otherwise I understand that you don´t want to waste your time.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Apr 20th, 2012, 8:37am

on 04/20/12 at 08:12:26, arimaa_master wrote:
Since Fritzlein is on his full capacity already I would be honored to play such postal game against cyborg_briareus.

However only if you think that I am a decent match for such game otherwise I understand that you don´t want to waste your time.


That would be great. You beat briareus postally so you're definitely strong enough. And for all I know I could make briareus weaker by interfering too much :P.

What time control do you prefer? I would be OK with anything from 1 day to 2 days per move.
Title: Re: Cyborg vs Human Arimaa
Post by Adanac on Apr 20th, 2012, 8:46am

on 04/20/12 at 08:37:38, rbarreira wrote:
That would be great. You beat briareus postally so you're definitely strong enough. And for all I know I could make briareus weaker by interfering too much :P.

What time control do you prefer? I would be OK with anything from 1 day to 2 days per move.


If you want to try two games at the same time, I'll give it a try also.  Two days per move is fine with me.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Apr 20th, 2012, 8:50am

on 04/20/12 at 08:46:37, Adanac wrote:
If you want to try two games at the same time, I'll give it a try also.  Two days per move is fine with me.


I'm not sure that I have the energy/time/computer time to do two games at the same time! Let me see how it goes in the beginning against arimaa_master and I'll let you know in maybe a few weeks!
Title: Re: Cyborg vs Human Arimaa
Post by arimaa_master on Apr 20th, 2012, 8:53am

on 04/20/12 at 08:37:38, rbarreira wrote:
That would be great. You beat briareus postally so you're definitely strong enough. And for all I know I could make briareus weaker by interfering too much :P.

What time control do you prefer? I would be OK with anything from 1 day to 2 days per move.


Thanks, two days per move would be best for me.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Apr 20th, 2012, 9:09am

on 04/20/12 at 08:53:50, arimaa_master wrote:
Thanks, two days per move would be best for me.


OK, I just sent the invite with time control 2d/60d/100/0/0d/21d

I flipped a coin at www.random.org and it came out with you playing Gold.

I had to play two games against ArimaaScoreP1 first or it wouldn't let me invite... (fortunately bot assistance was not required for those two games :P)
Title: Re: Cyborg vs Human Arimaa
Post by Dolus on May 1st, 2012, 10:26am
Let us know the game number when it starts so we can watch. :D
Title: Re: Cyborg vs Human Arimaa
Post by arimaa_master on May 2nd, 2012, 6:50am

on 05/01/12 at 10:26:19, Dolus wrote:
Let us know the game number when it starts so we can watch. :D


It is already started!
It is move four for cyborg_briareus.
Game Number 285052
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on May 2nd, 2012, 9:30am
This is interesting. Now I want to try too! If any bot developers have the time and find me worthy then I'd love to give it a shot.
Title: Re: Cyborg vs Human Arimaa
Post by minderbinder on May 2nd, 2012, 10:28am
Just as a question, I'm wondering if significant improvement to the "cyborg" team can be had by expanding the definition of what a cyborg can be (not that cyborgs would really need the help, since even in chess cyborgs still far outmatch everything else).

The idea is that the cyborg would not only be human + bot, but human + bot + other digital tools. For example, imagine a program designed to indicate, for each region ('region' can have many definitions), which player has the most control - perhaps producing some sort of map of the board. Or a program to identify vulnerable pieces on both sides, or vulnerable traps, or possible goal routes, etc. Or maybe even just a function to identify the most important elements in a position (i.e. "goal attack in east" or "potential camel hostage" or something like that).

Maybe all these functions can be handled by bots as they are - I'm not familiar enough with bot development to know for sure (I believe that most bots are an evaluation function combined with a search function - is this accurate?). I was just curious what someone with more familiarity with bot development thinks of this.
Title: Re: Cyborg vs Human Arimaa
Post by supersamu on May 2nd, 2012, 2:47pm
@rbarreira : could you also please write down on what moves you interfered with the bot and what moves the bot preferred at that time? It would also be interesting to know what your intentions were when you made another move as the one the bot proposed.
This will certainly be interesting to watch.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on May 3rd, 2012, 7:42am

on 05/02/12 at 14:47:34, supersamu wrote:
@rbarreira : could you also please write down on what moves you interfered with the bot and what moves the bot preferred at that time? It would also be interesting to know what your intentions were when you made another move as the one the bot proposed.
This will certainly be interesting to watch.


For the sake of posterity I will try to write my thoughts on at least some of the moves and perhaps a description of how I got to each move, even if many of my comments about the game turn out to be cringe/facepalm-worthy.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on May 3rd, 2012, 7:47am

on 05/02/12 at 10:28:41, minderbinder wrote:
Just as a question, I'm wondering if significant improvement to the "cyborg" team can be had by expanding the definition of what a cyborg can be (not that cyborgs would really need the help, since even in chess cyborgs still far outmatch everything else).

The idea is that the cyborg would not only be human + bot, but human + bot + other digital tools. For example, imagine a program designed to indicate, for each region ('region' can have many definitions), which player has the most control - perhaps producing some sort of map of the board. Or a program to identify vulnerable pieces on both sides, or vulnerable traps, or possible goal routes, etc. Or maybe even just a function to identify the most important elements in a position (i.e. "goal attack in east" or "potential camel hostage" or something like that).

Maybe all these functions can be handled by bots as they are - I'm not familiar enough with bot development to know for sure (I believe that most bots are an evaluation function combined with a search function - is this accurate?). I was just curious what someone with more familiarity with bot development thinks of this.


I suppose that for any decent cyborg team, the human would be in the best position to identify those strategic markers. I believe most bot developers would agree that making tools for identifying them with good reliability is not an easy task.

To a certain extent, bots that have some strategic knowledge should help to identify such things, by providing suggestions for moves which highlight the dangers/opportunities in a position. But at the level bots are at now, the human cannot rely on the bot to always sniff out those things...
Title: Re: Cyborg vs Human Arimaa
Post by hyperpape on May 3rd, 2012, 11:03am
Posting the link for our convenience: http://arimaa.com/arimaa/java/ys/ms4/v5/js_sit.cgi?sid=6005324419&grid=3&rand=54147169
Title: Re: Cyborg vs Human Arimaa
Post by omar on May 4th, 2012, 7:16pm
Interesting experiment. I've added a link to this game in the announcements.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on May 4th, 2012, 7:41pm

on 05/04/12 at 19:16:30, omar wrote:
Interesting experiment. I've added a link to this game in the announcements.

Check the spelling, it currently says "arimaa_mater".
I suppose this is a good time to bring up the error on the game record pages http://arimaa.com/arimaa/gameroom/pastrecord.cgi?id=XXXX, the losses column says "Loses".

Edit: Hey, that actually links to a real player!
Title: Re: Cyborg vs Human Arimaa
Post by omar on May 4th, 2012, 8:46pm
Thanks, fixed it.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Aug 11th, 2012, 10:36pm
Since the game against arimaa_master is almost over as of 25g ;), it seems to be a good time to revive this thread.

I'm very curious to know about your process and how often you overrode the bot, and how often you let it decide, or did you pretty much act as a team every move with each part giving approximately equal input?

Also, I'd be honored to be the next to take the challenge if you're ready. I think I've done enough to prove myself worthy.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Aug 12th, 2012, 4:57pm

on 08/11/12 at 22:36:19, browni3141 wrote:
Since the game against arimaa_master is almost over as of 25g ;), it seems to be a good time to revive this thread.

I'm very curious to know about your process and how often you overrode the bot, and how often you let it decide, or did you pretty much act as a team every move with each part giving approximately equal input?

Also, I'd be honored to be the next to take the challenge if you're ready. I think I've done enough to prove myself worthy.


I am happy to do at least 1-2 more games like this, but I think Adanac has his turn now if he wants it, as he was the first to propose a game after arimaa_master. But after Adanac it seems you'd be next.

I will paste the notes I wrote during the game in a few days after the game is over, although I ended up not writing notes for many of the moves.

In the opening I interfered a lot, in some particular moves the bot was having mostly stupid ideas, so in those moves it just checked my ideas for mistakes the bot can see quickly. But there were many moves where those roles were reversed, and also many moves where I played the bot's suggestion with none or almost no analysis by me. In terms of computation time, I often let the bot analyze overnight or even longer (on a computer which is about 2x slower than the one bot_briareus uses for the postal).
Title: Re: Cyborg vs Human Arimaa
Post by Adanac on Aug 13th, 2012, 8:01am
You should go ahead.  Browni3141 is one of the fastest improving players in Arimaa and it'll be interesting to see how he does in a cyborg match.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Aug 15th, 2012, 2:04pm

on 08/13/12 at 08:01:44, Adanac wrote:
You should go ahead.  Browni3141 is one of the fastest improving players in Arimaa and it'll be interesting to see how he does in a cyborg match.

Thank you, Adanac. I hope I'm not taking anything away from you though. If you were looking forward to the game then I can wait. Either way I think it will be one worth watching.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Aug 16th, 2012, 8:14pm
Adanac has given the clear go-ahead in a PM, so go ahead and challenge me when you're ready.

You can choose the time control, I don't really care what it is as long as it is within reason.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Aug 17th, 2012, 1:07pm
Thanks arimaa_master for the game, you definitely made me/us sweat during the opening.

Here is the little I have written of my thoughts during the game. Due to laziness and lack of organization/time It only goes up to move 10s, and even then not all moves have comments.

From move 11s on the bot did almost all of the analysis.


Quote:
Pre-game thoughts:

For the opening phases, the bot will help me avoid lots of tactical blunders and some strategic ones by refuting candidate moves, but I can't rely on it to always do it. The bot does not always quickly spot dangerous attacking patterns, so I'll have to keep in mind the bot's weaknesses and compensate by looking for those myself.

If I can get an even position out of the opening / early-middlegame I hope that the bot's tactics and somewhat deep searches at postal time will enable me to win the game. But a single mistake in the opening (and me or the bot are certainly not able to avoid all of them) against a strong player like arimaa_master can make it rather difficult. My optimistic side tells me that as long as the opening mistakes are not too bad, there's still a good chance to do well due to the bot's good play in the later phases of the game.

1s:

Already in the setup move I'm feeling how little practice I have with serious Arimaa play. I know that as Silver, I should be able to find an unbalanced setup that gains some sort of advantage against Gold's balanced setup.

But it is certainly not trivial: I don't want to put my elephant behind a trap as I'm not sure what is a good answer to a 2g that goes E -> C5 / F5. As a response to that 2g, my elephant either has to go back to the center or it has to decentralize itself, both alternatives taking quite a few steps to accomplish when starting behind the trap. However I've seen some players doing it, maybe they know a good answer to it. I'm not planning to analyze specific moves from other games in detail as copying moves from other games is not my aim here.

Putting the elephant too off-center is also out of the question, and so is putting opposite to the Gold elephant. So that leaves me with the e file for the elephant... am I truly that constrained?

I'll put the camel opposite Gold's b file horse with the intention of attacking it [ after making the setup move I now wonder if I should have put it on the g file instead, where it is farther from the Gold elephant... am I making mistakes already? this teaches me to ponder a bit more before making a move, maybe I should have written my thoughts here before making it, since apparently writing things down helps me reflect on them. hopefully this lesson does not cost me too much. ]

The horses should go on the c and g files in order to get to attacking positions quickly if necessary (although I don't expect to do an immediate elephant-horse attack in the East). I'll put a dog on the front of d file as it's probably good to have it there instead of the weaker cat. Behind the f trap I'll leave a cat in case there's a capture race. The front rabbits go on the extremes of the board where they're the hardest to pull.

5s:

I think the bot is not being very useful in terms of proposing moves for 5s. The only half(?)-decent move it has suggested somewhere along its analysis is 5s e->b3, which I don't like that much as it decentralizes. Other than that it ocasionally starts prefering 3-for-1 moves such as pulling the M to e2. But of course I'm still using it to try out variations from candidate moves.

I'm not very happy with the situation on 5s (I feel I went wrong somewhere, perhaps immediately in the setup as mentioned earlier?). Obviously gold is trying to pull the a5 rabbit, so I have to be prepared for 6g moves where the rabbit gets dislodged to b4, either with the E staying where it is or, perhaps more annoyingly, moving to c4. For a while I couldn't see any good 5s moves to counter this, but I then saw the possibility of 5s ee3w dd6s cd7s hc7s. It quickly became the candidate move, even though I don't think it's a crushing move. This move threatens a (not very immediate) elephant-horse attack, and by starting a mini-swarm on the c3 trap may allow me to advance with the camel on this wing if necessary to counter a rabbit frame or the silver horse getting hostaged by the gold camel. The gold camel seems to have a fairly short path to the west, so I'll have to look out for hostage possibilities.

I checked a lot of variations from this move, making extensive use of the bot to get suggested moves or find refutations all along 6g-7s or even deeper. But even though I went fairly deep, I can't be sure I've checked enough alternatives at 6g or 7g, so arimaa_master might surprise me with some crushing move that will make me feel even worse about the position. If I recall correctly, some of those variations were starting with, for example:

5s ee3w dd6s cd7s hc7s
6g Eb5e ra5e rb5s Ec5w
6s hc6s hc5s rb4w dd5w

5s ee3w dd6s cd7s hc7s
6g Eb5s ra5e Eb4e rb5s
6s mb7s mb6s rb4w mb5s / 6s hc6w hb6s rb4w hb5s / 6s mb7s mb6s rb4w hc6s

... and tons more which I don't remember (maybe I should start being more methodical and writing them down in a tree for better analysis... my analysis mechanics are not that good yet).

I'll give it a few more hours to see if I think of any alternative to 5s ee3w dd6s cd7s hc7s.



7s:

arimaa_master played a 7g that had been considered in analysis. The planned reply was something like 7s ec4w eb4e Hb3n mb7s, but after analysis with the bot I can see some problems with this move. In particular it allows 8g Ra2n Ra3n Hb4s Hg3w which cunningly prevents the cat from escaping while substantially delaying the harassment of the gold horse in the west.



8s:

I feel relieved that gold did not play the move I feared the most: 8g Ra2n Ra3n Hb4s Hg3w (freeing the horse for a while and preventing the cat from escaping through the east)

My first idea: 8s ra5s db5w rc8s mb6s

Some of the bot's suggestions: 8s cd4n ec4e ed4w Cd3n, 8s hc6n mb6w db5n cd4n, 8s cd4e ce4n ce5n rc8s, 8s ra5s de7w dd7s rf8w, 8s de7w dd7w ra5s cd4n

Another idea of mine: 8s ra5s db5w Hb4n ec4w

The bot does not seem to dislike this last at idea, it scores it around 0 which is an improvement from the situation at 7s. The best reply it sees as Gold is 9g da5n Hb5w Ec5s hc6s, which looks scary for Silver at first but allows the nice-looking 9s eb4n hc5e eb5e Ha5e .

So far I prefer my move to the bot's suggestions, it's more aggressive and "messy" which may be a good thing when playing against humans. A drawback I see is that the silver elephant gets decentralized, perhaps something like 9g Dc2n Dc3n Rc1n Rb1n can cause trouble? I'm not surprised the bot does not see this 9g, since its blockade eval is asymmetric and so it does not care much about elephant mobility. Maybe this 8s is not that good after all? Time to write down some lines:

8s ra5s db5w Hb4n ec4w
9g Dc2n Dc3n Rc1n Rb1n
9s Dc4s eb4e rc8s cd4n (does this make my 8s look good again?)

Maybe an even better choice for Gold is 9g Dc2n Dc3n Cd3w Rc1n or 9g Dc2n Dc3n Rc1n Rc2n

... Much later: (due to uncertainty on the previous analysis and a busy week IRL)

The bot has been running for a long time on the position and it settled on 8s cd4n ec4e ed4w Cd3n after more than 100 hours of computation. Here's a possible line given by the bot which illustrates how this move might be good:

8s cd4n ec4e ed4w Cd3n
9g Dc2e Dd2n Rc1n Ra3n
9s hc6e cd5e Cd4n ec4e (that's rather tricky as the cat may get forked between two traps soon)

I worry about 9g cd5e Ec5e Cd4s Ed5w though.

... A few days later:

I'm still not sure about the lines above, in the meantime the bot has continued to run on the position and it currently prefers 8s cd4e ce4n ce5n rc8s which looks safer than my ideas above, so I'm going with it...

9s:

I was on vacation in the beginning of this move, so just like 8s some reserve will be spent here...

I left the bot analyzing for a bit more than a day and it doesn't seem unhappy with the position. After finishing the depth 5 search it is choosing 9s mb6w db5n Hb4n ec4w. It looks good, as it threatens a potential horse hostage and despite decentralizing the elephant, it looks like that's not a big problem (I don't see it getting blockaded at least, worst case it should be able to escape through the c6 trap).

Here are some possible variations:


9s mb6w db5n Hb4n ec4w
10g Ec5s hc6s Rb1n Db3w
10s db6e hc5e dc6w hd5n


9s mb6w db5n Hb4n ec4w
10g Ec5e hc6s hc5s Ed5w
10s hc4e hd4n ce6s hd5n

I can't see anything wrong with this 9s so I think I'm going to make it right away and not use up any more reserve.

PS: A close second in the bot's analysis (score differing only 2 centi-rabbits) was 9s mb6n db5n Hb4n ec4w. I think I prefer the bot's first choice (the horse hostage looks quicker to secure, but that's just a quick eye-balling, not any detailed analysis that I've done on it...).

10s:

The bot's suggestion after finishing depth 5 was 10s eb4e Ra4e ra5s ma6s with score 0.35. Gold's west looks clogged which might make it hard to stop silver from hostaging the horse rather soon. If I understand correctly, Fritzlein always says that a camel hostaging a horse should have a supporting piece ahead of it to help save it from getting pulled out by the enemy elephant. It looks like the dog can be the supporting piece if a horse hostage does arise. The rabbit advance looks bold, but maybe it's OK as there might be some prospects to generate goal threats there in the medium/long term.

Some variations I checked out with the bot, both looking OK for silver:

10s eb4e Ra4e ra5s ma6s
11g db6n Hb5n Rb4n Db3n
11s ma5n ma6s Hb6w hc6e

10s eb4e Ra4e ra5s ma6s
11g db6w Hb5n Rb4n Db3n
11s hc6e da6n ma5n Rb5w

I also checked out a variation which turns out to lose the horse for Gold:

11g db6w Hb5n Ec5w
11s ra8s ec4n rc7w hc6n

I think I like the bot's suggestion and I will make it now, gaining a little bit of reserve (and more importantly not severely draining it as happened in the last two moves).
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Aug 17th, 2012, 8:16pm
A link to the game: http://arimaa.com/arimaa/gameroom/opengamewin.cgi?client=1&gameid=306132&role=v&side=w

I'm pretty sure I did it right this time.
Title: Re: Cyborg vs Human Arimaa
Post by arimaa_master on Aug 18th, 2012, 6:04pm

on 08/17/12 at 13:07:55, rbarreira wrote:
Thanks arimaa_master for the game, you definitely made me/us sweat during the opening.

Here is the little I have written of my thoughts during the game. Due to laziness and lack of organization/time It only goes up to move 10s, and even then not all moves have comments.

From move 11s on the bot did almost all of the analysis.


Thanks for the game too. Especially thanks for the comments - because I didn´t know when I went wrong - now I have some clues.
Title: Re: Cyborg vs Human Arimaa
Post by Fritzlein on Aug 18th, 2012, 9:26pm
Thanks for the commentary, Ricardo.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Oct 13th, 2012, 12:19pm
rbarreira brought this up in the game comments, but it was never addressed.
Why is this game (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=241423) not rated? While it shouldn't be WHRH rated, there is no reason for it not to be gameroom rated.
Title: Re: Cyborg vs Human Arimaa
Post by aaaa on Oct 13th, 2012, 1:20pm
I unrated the game because the "cyborg_" prefix isn't an official prefix and I believe that there should be transparency in how rating lists are calculated. Being used for official purposes, the WHRH ratings should, in my opinion, cleanly follow from the database of games and not depend on further arcane exceptions, thus making it easily reproducible by those not aware of them.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Oct 13th, 2012, 2:40pm

on 10/13/12 at 13:20:26, aaaa wrote:
I unrated the game because the "cyborg_" prefix isn't an official prefix and I believe that there should be transparency in how rating lists are calculated. Being used for official purposes, the WHRH ratings should, in my opinion, cleanly follow from the database of games and not depend on further arcane exceptions, thus making it easily reproducible by those not aware of them.

I'm talking about the site's gameroom rating system though. The game is not gameroom rated. Aren't they two different systems? Can't the game be gameroom rated but not WHRH rated? The type of player doesn't matter for gameroom ratings.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Oct 13th, 2012, 2:44pm

on 10/13/12 at 13:20:26, aaaa wrote:
I unrated the game because the "cyborg_" prefix isn't an official prefix and I believe that there should be transparency in how rating lists are calculated. Being used for official purposes, the WHRH ratings should, in my opinion, cleanly follow from the database of games and not depend on further arcane exceptions, thus making it easily reproducible by those not aware of them.


It would have been nice if some sort of comment was posted to the game explaining this (or did I miss it?). Then we could have had a normal discussion about what to do instead of a stealth unrating. It was strange to see the game turn into an unrated one after woh said that he added the exception for cyborg_ players, as I understand it his WHRH calculation now ignores those players.

For example I might propose that cyborg_ be made into an "official" prefix, if Omar thinks that's a good idea.

BTW thanks browni for re-asking this question.
Title: Re: Cyborg vs Human Arimaa
Post by aaaa on Oct 13th, 2012, 3:18pm
OK, that was a misjudgment on my part, for which I apologize.
Title: Re: Cyborg vs Human Arimaa
Post by Nombril on Oct 13th, 2012, 9:33pm
Oh, I am surprised that it wasn't played as a U game in the first place.  If two human players collaborated on a game, we would expect it to be U.  When we played the simul games, they were U.  My expectation is that all rated games are a single player against another player.

I caught some flack for not playing U games when I was playing multiple bots at one time.

I think the reverse question would be more appropriate, why should a game be rated when one player was computer aided and the other was not?
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Oct 13th, 2012, 10:34pm

on 10/13/12 at 21:33:26, Nombril wrote:
Oh, I am surprised that it wasn't played as a U game in the first place.  If two human players collaborated on a game, we would expect it to be U.  When we played the simul games, they were U.  My expectation is that all rated games are a single player against another player.

I caught some flack for not playing U games when I was playing multiple bots at one time.

I think the reverse question would be more appropriate, why should a game be rated when one player was computer aided and the other was not?

It's interesting to me that you ask that question, because before I thought it was "obvious" that the games should be rated, and that everyone else would think so too.
The circumstances you describe are all different from cyborg_briareus' games in the same way. In all of those scenarios the player's output is different somehow from their normal output. For cyborg_briareus games, a human assisted by computer IS the norm. I think that the human+computer should be treated as a single entity if all games are played that way, not a human being assisted by a bot. cyborg_briareus is a single player.
Title: Re: Cyborg vs Human Arimaa
Post by rbarreira on Oct 14th, 2012, 2:32pm

on 10/13/12 at 21:33:26, Nombril wrote:
If two human players collaborated on a game, we would expect it to be U.


Not necessarily, I think it could be OK to have usernames for a team and play rated games with them. As long as it's fully disclosed what a certain username is, anything's fair game IMO. The only issue is that it affects the WHRH, but that can be handled (and in the cyborg_ case it was handled after someone pointed out this problem, which I hadn't thought of when I made the game rated).


Quote:
When we played the simul games, they were U.



Quote:
I caught some flack for not playing U games when I was playing multiple bots at one time.


There's a good argument for simultaneous games to be unrated, the player who's doing the simuls is playing at a much weaker strength than usual, due to having his mind on many games and not using the whole time control. If those games were rated it would cause a distortion in the gameroom ratings.

A cyborg player or a team of players does not have the same issue, as long as they always play with the same team composition.


Quote:
I think the reverse question would be more appropriate, why should a game be rated when one player was computer aided and the other was not?


I think we should allow rated play as long as we're playing Arimaa (and not a variation like the EEE games or handicap games) and the entities playing under both usernames are reasonably constant and fully disclosed.

I say reasonably constant because it wouldn't be reasonable for a bot to be playing on a home computer during even months and on a supercomputer in odd months. Same thing for players doing simuls on one week and single games on the next week. It puts too much stress (and misinformation) into the ratings.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Oct 14th, 2012, 3:28pm
I would also like to suggest that this account's games be included in WHRP. I don't see a reason for it not to be.
Title: Re: Cyborg vs Human Arimaa
Post by browni3141 on Oct 15th, 2012, 10:53pm
Hey, omar. Could you update the gameroom page if you get the chance?
Title: Re: Cyborg vs Human Arimaa
Post by Fritzlein on Oct 16th, 2012, 10:06am

on 10/15/12 at 22:53:31, browni3141 wrote:
Hey, omar. Could you update the gameroom page if you get the chance?

Do you mean the announcements?  I think I have permissions to edit that.  But I can't edit the link to the game in the upper right.  Since Omar isn't keeping up with the forum, send site requests via http://arimaa.com/arimaa/contact/
Title: Re: Cyborg vs Human Arimaa
Post by woh on Oct 16th, 2012, 4:49pm

on 10/14/12 at 15:28:52, browni3141 wrote:
I would also like to suggest that this account's games be included in WHRP. I don't see a reason for it not to be.


The game was included in WHRP of August 21st and 28th. Since the WHRP of September 4th the game was no longer included because it had been unrated.

To include the games of cyborg_briareus in the WHRP they should stay rated. They will not be included in WHRH because I have added cyborg_briareus to the list of players to exclude. I added this option some time ago for calculating a specific bot ranking.

on 10/30/09 at 15:22:46, aaaa wrote:
Would you be willing to apply your rating system to rated games involving only developer bots?


Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1!
YaBB © 2000-2003. All Rights Reserved.