Arimaa Forum - Print Page


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

        Arimaa >> Game Analysis >> clojure - OpFor (15335) and thoughts in general
        
(Message started by: clojure on Sep 3^rd, 2010, 4:55am)

Title: clojure - OpFor (15335) and thoughts in general
Post by clojure on Sep 3^rd, 2010, 4:55am

Hello all, I have been playing bots some time now, and would like some external help.

I feel that when I play games with more than 2 minutes per move, I can play quite stably, and don't make immediate blunders too much. I even think that sometimes I get a position that has advantage for me, a situation where the structure is somewhat fixed, so I can think what strategy is needed.

Somehow I almost always lose, which means that either my positional analysis incorrect, and/or that I am very bad at rabbit goaling. It feels like anything I do, the opponent has already calculated and has counter-attack. Often the fixed position shatters, and my understanding of the situation becomes superficial. Since the nature of the game, often this makes carefully calculating opponent easier to avoid blunders. And mentally for me it's quite hard to get my pieces to fit to some sort of solid combination again.

So here's a game from yesterday. Near the very end I make simple mistakes, so forgive me for those, but I would appreciate still as much comments as possible.

I think the most crucial mistake is about move 33, where I probably should start advancing my right wing horse and dog before trying anything with my camel. That's because it seems I cannot do much harm in the opponent's home area because of material overwhelm of my opponent there.

My understanding of the position at that point
a) Opponent's camel is tied nicely?
b) Opponent's elephant is under pressure to protect his cat, or lose time if saving it. Elephant is even situated badly, since my camel is roaming freely.
c) I need to beware that opponent's elephant doesn't mess up the structure so that his rabbits can advance on the right wing
d) Even though quickly glancing it seems that in the northwest I have a close winning possibility, it's hard to move there, since F3 area blocks my elephant well.
e) My camel is in good position to fight goal winning

Because of d and e, I tried to help my elephant with camel, but it failed miserably. After that I came(l) back but made blunders.

I'm especially interested in hearing how I played the opening, the middle game, and what should I try to understand better. Is there clear misconceptions in my thinking or play?

I ordered Beginning Arimaa (Amazon just informed of its postage), but it will take a while before it gets here. Is there enough material for a sequel? I would like to encourage enthusiastic (e.g. Fritzlein) by supporting his considerable amount of work to help others.

Btw. is there a possibility to review the game like in KGS (go server)? So that one can take existing game and show a friend in live what should have been done, and having different alternative tree structures saved with comments. I think this is very nice way of getting new players excited, and also to discuss the game with opponent.

Some background info: I get to know Arimaa some years ago but played maybe one game against bot only. But to me it's clear this has lots of potential, and is nice game. But one unfortunate property educational-wise is its dynamics. It's similar to chess in how one can discuss the game in a book. Compare this to go where the position changes mostly with additive stones. This makes it visually much better suited for problems etc. That's why I hope that those that understand Arimaa could make material that is interactive.

Enough of ranting, nice to meet ya all, and hope I get a review for this game. Thanks!

The game: http://arimaa.com/arimaa/games/jsShowGame.cgi?gid=153354&s=b

Title: Re: clojure - OpFor (15335) and thoughts in genera
Post by speek on Sep 3^rd, 2010, 11:12am

I'm going through the game now. One thing that looked promising to me is on move 17s, you should have done:

hh6s ef5e Mg6e eg5n

Ie, take the camel hostage. Instead, what ends up happening is his camel escapes and takes your horse hostage.

Title: your plan at move 33 had some legs
Post by speek on Sep 3^rd, 2010, 11:27am

To my eyes, on move 33, you came up with the right plan - ie, move the camel over to the other side and use it to make scoring/capture threats. Opfor's reaction seems to justify this, as it was completely defensive.

However, on move 34, you use 4 steps to pull his horse down and advance the rabbit. Opfor needs only 2 steps to completely undo your move. In my mind, that means it cannot be a good move. Perhaps a better plan is to move the camel behind the horse to prepare a pull that can't be undone, ie pull the horse, advance the rabbt and move the elephant behind, or advance the next rabbit up too.

Of course, moving the camel behind is only 1 step. With the other steps you could put your dog in front of his A-file rabbit with the idea of pulling it and capturing it rather than pushing it back.

But, in the game, you gave up on your plan, moved back after accomplishing nothing other than losing a rabbit. And then Opfor blocks you from going back! I think the plan had legs :-)

Title: Re: your plan at move 33 had some legs
Post by clojure on Sep 3^rd, 2010, 11:39am

Hi speek, thanks for commenting!

I avoided the camel hostage because I've had bad luck with hostaging earlier. In this particular case, there was already a rabbit to make it more dangerous and elephant to make it impossible to make serious threats. It would have taken a lot of small shuffling to get anything done, and it is risky for my skills, I think.

In this case I also wanted to try more calm approach. I didn't mind that camel would take the horse, since it would deadlock himself and my elephant would be quickly there and on the outerside with influence.

And you're probably right about that I should have done after 33 differently but with the same strategy.

Thanks for the pointers!

Title: Re: clojure - OpFor (15335) and thoughts in genera
Post by speek on Sep 3^rd, 2010, 11:46am

Hi clojure, I still disagree about the camel hostage. The rabbit was already framed there, but as you saw, when it mattered, the gold elephant felt free to sacrifice it for mobility. If you had the camel hostage, that would have been a much harder choice.

Having the camel hostage is a static situation that just means your own camel can run around unopposed. Instead, you lost the use of a horse, and I don't think you had any way of getting it back easily. Certainly you never did during the game.

Title: Re: clojure - OpFor (15335) and thoughts in genera
Post by clojure on Sep 3^rd, 2010, 11:55am

I agree in general what you say. In this situation, however, I think I managed to create enough attack force with rabbits and the horse so that only by themselves they could make moving the camel very restrained. At least I would have time to do with the elephant something else. So I couldn't move the horse but neither could he move the camel. And in the end, I could move my camel but he couldn't. And as you pointed out previously, I probably could have finally freed my horse with my camel and the situation would have been very serious for silver.

In contrast, if I had taken the camel hostage, the opponent could have dragged his pieces near it and freed his elephant to attack my camel, and things could have become much more complicated.

This is how I feel about it. And it's exactly what I wish that people criticize this thinking, so I could get rid of my wrong viewpoint :)

Title: Re: clojure - OpFor (15335) and thoughts in genera
Post by Fritzlein on Sep 3^rd, 2010, 11:56am

on 09/03/10 at 04:55:08, clojure wrote:

Hello all, I have been playing bots some time now, and would like some external help.

You set yourself up for help when you give extensive comments as you have done in this post. It is much easier to know where to start when you give your own evaluation.

Quote:

Often the fixed position shatters, and my understanding of the situation becomes superficial. Since the nature of the game, often this makes carefully calculating opponent easier to avoid blunders. And mentally for me it's quite hard to get my pieces to fit to some sort of solid combination again.

This is a very good observation. When I play bots, I can do much better when the position is strategic, i.e. elephants tied down and no immediate captures for either side. When there start to be capture threats at more than one trap, I start to blunder more often. This means bots have a relative advantage in "messy" positions.

Nevertheless, I think it is a mistake to avoid messy positions. You were correct to attack with your camel on 29s, even though contesting opposing traps can make it harder to keep everything under control. When I learned chess, I was instructed to play open games starting with 1. e4, because even closed games will break open eventually, so you might as well start by learning to play open games. Similarly for Arimaa, you might as well learn how to play sharp positions, even if you blunder a lot, because the usual way for a strategic position to resolve is for the disadvantaged player to sacrifice material in order to make the position sharp again.

Quote:

My understanding of the position at that point
a) Opponent's camel is tied nicely?

It is optimistic for you to count this only as an asset. Opfor's camel is holding your horse hostage, so your elephant is tied to the defense of the f3 trap. Yes, there is no way for OpFor's camel to get free, and your swarm provides some pressure, but the pressure is mutual. It should be in the back of your mind that if your elephant must leave, you will start hemorrhaging in f3.

Quote:

b) Opponent's elephant is under pressure to protect his cat, or lose time if saving it. Elephant is even situated badly, since my camel is roaming freely.

This is also true, but again not as advantageous as you are presenting it. Consider what happens if OpFor gets his dog to e6 to relieve his elephant. Then OpFor would have a free elephant while yours is still tied down. Normally if the gold elephant rotates out, you would want to use your elephant to regain control of f6 and capture the invading small pieces, but remember that while you are fighting for f6, you could be losing pieces in f3.

This is not to say your position is bad, only to say that your opponent can make it messy fairly easily, so it is not to be viewed as powerful control.

Quote:

d) Even though quickly glancing it seems that in the northwest I have a close winning possibility, it's hard to move there, since F3 area blocks my elephant well.

Yes, it is good to see that you are not on the verge of goal.

Quote:

e) My camel is in good position to fight goal winning

I think this was your fatal strategic misconception. Gold has a thick goal defense east, west, and center. It was not the time to think of forcing goal on either side if you are going to be losing pieces in the process. For example, suppose you lose your three advanced rabbits around c3 while your camel is ushering forward a rabbit from g4 to g2. If then the gold elephant is forced to abandon his hostage cat to come home and help on the defense, what have you achieved? You will have lost three rabbits to gain a cat, with a resulting huge stalemate around the f3 trap. If you can't break through, you are a net loser.

When the defense is so thick, you need to be thinking about capturing pieces to cut a path to goal. On move 32s, if you use your camel to push the gold horse east with an eye to taking the b3 square with your dog, OpFor will be in dire straits. Let's say 32s MH>Rb4<v. You are set to win the trap control fight, and just a couple of captures will give you a clear path to goal.

Quote:

I think the most crucial mistake is about move 33, where I probably should start advancing my right wing horse and dog before trying anything with my camel.

So, you already spotted what I was talking about. Sorry for belaboring the point. I guess my contribution is that you are better off to think of captures being necessary before you can force goal, and to point out that 32s was already a mistake in light of the correct strategy of fighting for control of c3.

Quote:

I'm especially interested in hearing how I played the opening, the middle game, and what should I try to understand better. Is there clear misconceptions in my thinking or play?

I see no large issues. 10s/11g leaves you wide open for an elephant-horse attack against the c6 trap, so I think you got a touch lucky there. The general strategic point is that if you set up unbalanced with camel on one corner and no horse with it, moving the camel away from that wing leaves the wing very weak.

I like OpFor's move 15g, after which OpFor has the makings of a dangerous swarm. You were correct to push the gold cat back instead of sideways on 15s, so as to delay the wave of pieces coming forward. OpFor's 16g and following let you off the hook by not following up with a charge up the g and h files. OpFor had good chances to win the battle to free its elephant, had OpFor played for that outcome. The possibility of a swarm is just something you have to keep in mind when you pull a rabbit on the wing where your opponent's camel is and your camel isn't.

On 26s, you should be thinking about attacking the c3 trap starting from b3, because that side is furthest from the gold elephant. Centralizing your camel allowed OpFor's strong tactic on 27g. You got lucky that OpFor didn't take the rabbit on 28g, because it would have been worth it for OpFor to free its elephant and leave you in a bad strategic situation around f3. I would have rated you only slightly ahead in that scenario despite your winning a cat. Probably Opfor can frame and eventually win your g4 rabbit while keeping you horse hostage on h3. So 26s can be considered a tactical mistake, but there is a strategic principle there too about attacking traps from the outside in most cases.

Overall, you played purposefully and well; Opfor was in hot water prior to your 32s/33s mistakes.

Quote:

I ordered Beginning Arimaa (Amazon just informed of its postage), but it will take a while before it gets here. Is there enough material for a sequel?

Chessandgo is writing a book which will make an excellent sequel. I have read some of his drafts, and I can enthusiastically recommend his book to any intermediate players. It might not be the best introduction for someone just learning the game, but it would be perfect for you.

Title: Re: clojure - OpFor (15335) and thoughts in genera
Post by Fritzlein on Sep 3^rd, 2010, 12:06pm

on 09/03/10 at 11:12:41, speek wrote:

This is a very close call. You both make good points. However, to me taking the camel hostage looks too dangerous. What tips the balance in my mind is that the silver horse can be fixed on h5 by 18g Ra^^^Cb^. Silver needs that horse to fight on the rest of the board, and particularly to help fight for control of the f3 trap. With the horse stuck, when Gold swarms forward to free his elephant, suddenly Silver has to be super-cautious about swarming right back, for fear of losing a forward piece in f3. Thus the fight for elephant mobility tips slightly back in favor Gold.

The fact that Silver can't automatically take the Gold camel hostage is an argument for why the Gold camel charge was good in the first place. And the fact that the Gold camel charge was good is an argument for why it is dicey to pull a rabbit on the side where the opposing camel is and your camel isn't.

Title: Re: clojure - OpFor (15335) and thoughts in genera
Post by clojure on Sep 3^rd, 2010, 12:12pm

Hi Fritzlein, I have quickly looked what you said, and you provided exactly the kind of words I was hoping for. Thank you! And I'm definitely buying chessandgo's book whether or not it's beyond my scope of understanding ;) The little I've read and heard from both of you are very important for the Arimaa community to grow.

I have to ponder the game over more a bit later. The two main issues for me seems still to be that I evaluete fixed positions too optimistically for me and that I chase goal positions too soon. I was aware of them before, but it seems those are still the biggest problems (after silly blunders).

Title: Re: your plan at move 33 had some legs
Post by Fritzlein on Sep 3^rd, 2010, 3:36pm

on 09/03/10 at 11:27:44, speek wrote:

That could be correct, but it does involve sacrificing material around c3. Usually when someone sacrifices material for a goal attack like this, it will either be spectacularly right or spectacularly wrong. :) I have been on the wrong side of this judgment call often enough to doubt my intuition.

Title: Re: clojure - OpFor (15335) and thoughts in genera
Post by speek on Sep 3^rd, 2010, 8:31pm

Yes, well, advancing the rabbits against c3 was entirely premature. Given that though, letting the bot spend tempos capturing those rabbits while the camel does damage by f3 seems like the best use of those rabbits, to me :-)