Topic: Abalone-complexity versus Chess-complexity


Greetings,
Funky-AbaloneTheory-JazzClub

P.S.: You can find the original topic-sujet-message in this forum, too.

We present here the topic-related sentences:

-------------------------------------------------------

Some time ago, I had thought about an abalone IA based on a neural network.

Actual abalone IAs are currently strong because of brute force against common players, ie their ability to calculate deeper moves.

However, I think this approach has some limitations. The first limitation is in the time needed to calculate depper moves, which grows exponentially. The second limitation is in the current level of players, which is weak when compared to some other games (like chess), but i am confident that this limitation will rise with the players becoming better and better.
Once players will be trained to such deep calculations, there is the risk that those programs become 1- too weak or 2- far too slow.  

--Last edited by Funky-AbaloneTheory-JazzClub on 2005-11-18 18:06:37 --

Hello chris, nice to see you here  ;)

Hi silver

Hi chris,

if you have a look at the End-game-practice-tasks, then you will see that there are some tasks of "Winning-in-5-move" and a task "Winning-in-6-moves" - many tasks will be created, I hope so. This level is at least equal to the Level 10 of Aba-Pro.
With these tasks, many players can practice the deep-thought.
And some players have alreay practiced those tasks.

Of my point of view, to play games with long thinking-time is a way to think deep, too. But you have to integrate strategical and positional thoughts here.
In these Endgame-tasks, you can practice pretty isolated the tactical thinking. Therefore I prefere these tasks to practice deep tactical thinking.

On the other hand tasks with 15 or more moves-solution are - of my point of view - not proper to practice the deep thought, because many or all players  have to take a board with marbles, the conditionas are too far away from game-condtions and if tasks are so far away from the win, then probably you have to include strategical and postional thinking, too.

Thats why you will find here practice-endgame tasks with a thought-depth of 1 to 10 moves.

I have practiced in chess my deep-thought only with those Mate-tasks, because there are special books with these tasks. After you have practiced those tasks for some months, it is not difficult to look 5-6 moves  deepb(5-6 moves of each colour).
I dont think, that it will be in Abalone very different to chess.


Greetings, Silver  

--Last edited by Funky-AbaloneTheory-JazzClub on 2005-10-26 13:39:23 --

I think there is a fondamental difference between abalone and chess : the combinatory explosion, that brings the horizon (the point beyond you are unable to see) far closer in abalone than in chess.

As soon as you get complex interactions (see my post about what i call complex interactions), you have thousands possibilities to think about just in order to predict the 3 or 4 next moves. The number of combinations grows far faster than it does in chess, and this number is usually reduced by staying completely compact, ie minimizing the interactions between the two players, which leads, and i'm pretty sure you'll agree on that, to poor and boring games, because it becomes then too easy to block and too difficult to attack

yes, I agree, but the fantastic and important thing is, that as more complex a game is as less moves are possible good moves.
Thats why in Belgium-Daisy in the opening there are always 3-4 very good moves. In Standard Aba-Pro is much longer calculatíng in block-situations.
I dont think, that there are so many differences between chess and Abalone, just in complex situations. Probably there are most of the time only 2-3 good moves in complex middle-game-situations - sometimes only one.

Quote :

that as more complex a game is as less moves are possible good moves

Hum, i think that this point is very debattable...

Quote :

Thats why in Belgium-Daisy in the opening there are always 3-4 very good moves

I won't argue here because, believe it or not, i cannot remember the belgium daisy configuration

However, my intuitive reformulation (you know how i like to disagree) would be "there are always 3-4 very good *well known* moves"

Hi Chris,

how can you explain that programmes think longer in block-situations ? If a situation is complex, then there are most o time less possible good moves. This seems only contra-intiuitive to our "common sense imaginations", I think so.

Silver  

--Last edited by Silversurfer on 2005-10-25 21:31:36 --

I guess this is really dependant on the algorithm, particulary when its priority is structure (compactness) and position (the center).

In such algorithms, once the computer's balls are extremely compacts, the "weigh" of each move becomes closer to the weigh of other moves, implying more computations are needed to be able to really distinguish one among the others.

This is just a theorical explanation, each algorithm needing to be deeply analysed in order to explain its specific behavior in this or that case.  

--Last edited by chriscool on 2005-10-25 21:28:06 --

Ok, I have some arguments:

1. I think good players in chess think 5 moves deep in usual games. And I dont think that this is not possible in Abalone, I can imagine that Abalone-players will think deeper than 5 moves: this is level 10 on Aba-Pro.

2. I think there are "patterns" in chess for example of the Pawns: double pawns, isolated-pawns, edge-pawns, or the bishops or the Knight (good night/bad Knight,Bishop related to the pawn-structure)
I notice here many similarities to isolated or caged marbles. More general: there are "structure-patterns", you have to know.

3. In chess you can take the centre, but since Nimzowitsch all chess players know, that this is not very important, because there are many openings (or example all "Indian"-Openings or Ben-Oni and so on, who are able to play against a large  centre.
In Abalone the centre is more important, I think so. This is an argument against the complexity of Abalone.

4. I have put a complexity-analysis into the Programming-Forum. Chees and Abalone are - compared with 19-19 Go - similar here.
But if you include the different move-patterns in chess of any figure, I dont think that Abalone is here more complex.

5. Endgame, opening and middlegame, I still think these periods are identical.  

--Last edited by Silversurfer on 2005-10-25 21:53:53 --

1- I don't mean to be harsh, but your intimate conviction is not enough to make an arguement
More seriously, i think the depth you can anticipate depends directly on the complexity of the interaction between the two players (ie the number of points of contact). You can forsee the 5 next moves only when you have few interaction (ie when you are very compact in structure and that you throw away almost all the moves compromizing your compactness).

2- I agree there are observable patterns, but only once most of your balls are compact : it becomes a little bit irrelevant, for instance, to speak about isolated balls as a pattern when most of your's and your opponent's balls are isolated (i exagerate on purpose).

3- The center is important, for sure, but i remember goha winning many games while not giving himself a sh*t about the center. Once again, the more compact you are, the more important the center is because possessing it increases then the sub-set of your "tactical environement" (the set of your possible moves, what you relates to tactical analysis) not compromizing your structural integrity (ie : when possessing the center, you have more possible moves that won't compromise your compactness).

4- I have to read this before argueing

5- hum- You lost me. A) in what are they identical ? B) In what is this relevant ?

If i except the 4- and 5- that are unclear to me so far, you didn't demonstrate that abalone was at the same level of complexity than chess, but that there were some ways to lead the game so that its complexity can reach the same level, mainly by focusing on compactness.

1. When I said "5 moves"-deepth than this is an "average". It is obious that this depends on the structure, I think so.
I relate this average to common complex games.

2. "isolated balls" is a very important positional-pattern in Abalone, I think so. I dont agree here with you, totaly.
Have you read the positional-level-chapter?

3. Its possible to win in Abalone without the centre, especially if the opponents are not good but I have no doubts that the centre is more important in Abalone. Only if you get some marbles on the edge, than this has  an exchange-value. But this is not what I mean, there are other cases.
The centre is obvious as an important value, you have to change something against it, if you want to have an equal situation. And the possibilities are reduced by this principle to less probable serious possibilities.
I think this is not "totaly" different to chess, but in the measure very different.    

4. I think that these periods are indicators of complexity, too.


I dont know, but I think - particular for the second point - you have a lack of Abalone-experience here.

I think my average is, if I play not too fast, but not long games, 3-4 moves per move. Sometimes there are mistakes, no doubt.
But my chess experience and my Abalone-experience are very similar in many developments.

But I dont think that I have to proof something. This is my opinion and your opinion is different. I can live with this situation  

--Last edited by Silversurfer on 2005-10-25 22:48:17 --

Quote :

"isolated balls" is a very important positional-pattern in Abalone, I think so. I dont agree here with you, totaly

I think you misunderstood me. I didn't mean it was not an important pattern, but that it couldn't be described as a pattern *if* you don't have a stable/compact structure relative to this isolated ball. This notion of "isolated", then, is not an absolute but a relative concept. If all your balls are spread, the concept of "isolated ball" becomes less and less significant.

Obviously, i agree with you that structure is important in this game, and you can assume without many risks that your balls will sooner or later aggregate into structures. At this moment, the notion of isolated ball will become relevant as a pattern.  

--Last edited by chriscool on 2005-10-25 22:55:15 --

ok, this was a missunderstanding. I thought this is an argument, not a remark.
I think in real games isolated balls are very often very important for several patterns, like I discrbed it in the positional-level-chapter. There are much more funtions for isolated balls. To me the "term" "isolated" only means that "there is no contact to any other marble of the same colour".
You can defend with one isolated marble if you put it on a key point, you can break your opponents' main-group-open with a isolated marble in his main-group, and so on.  

to cage a isolated marble:


attacking with the isolated black marble:


In (2) the white isolated ball is "possible forerunner", to break the black group open. But it is not simple to do this.
In (2 ') the white isolated ball is not a "possible forerunner", its rather a possible ejected marble.
 

and many other functions in more complex situations.  

--Last edited by Silversurfer on 2005-10-25 23:08:45 --

Quote :

It is obious that this depends on the structure, I think so

That was precisely my point : increasing the number of points of interaction increases considerably the complexity, in an exponential curve, when it grows more linearily in chess (just a guess from the rules, i'm not an expert in chess).

ok, I have to reflect about this last sentence. I am not sure, if I understand you well.
It seems to me that you regard with a  more theoretical point of view in this point than me. My point of view was more based on my impression of the structure of the real-games: very good abalone-players and very good chess-players in complex games.

If you regard the "pure possibility-curve" with related dimensions than I agree from a theoretical point of view, but not of practical point of view. I think the practical principles reduce the possibilities in Abalone very much.  

--Last edited by Silversurfer on 2005-10-25 23:25:40 --

Hum...

First, we have to admit that large theorical fields are still to be explored in abalone, mainly because there are so few abalone players, especially when compared to chess players. More, very few abalone players risk themselves in novative approaches (ghosts of elo hunters, i guess).

Second, my comments are not completely theoritical and are relying a lot on current eob and goha's experimentations.

In this regard, my comments are probably more applyable to variant than standard games (and this is the very reason why eob and goha are playing only variant currently).  

--Last edited by chriscool on 2005-10-25 23:28:31 --

I did not want to say that your comments are not based on practice, but I want to say "the point of view" to this object was different, sure with practic in the background. I think this are too different things.
Maybe I have to say "perspective" instead of "point of view"?
One can argue with a theoretical or a practice point of view, independent of practical or theorectical knowledge. Probably if there is no knowledge, than this is bullshit, but the perspective is still a perspective.
I mean this difference: "perspectice", not the "background".  

--Last edited by Silversurfer on 2005-10-25 23:33:17 --