Talk:Arimaa/Archive 1

Archive 1

Archive 2

Archive 3

Old talk

Malathion 07:58, 17 Apr 2005 (UTC) Would it do any good to mention the fact that Go is already a fine example of a board game where computers are still vastly worse than the best humans? Go seems to share many of the qualities that are described here, in particular extremely long range planning, a wide choice of moves for any given position, and a heavy weight placed on positional considerations rather than material gain.

Yes, I think it would make sense to mention Go as the prime example of a game computers play poorly relative to humans. But apparently there are others such as Havannah, Octi and Twixt. I almost want to create a page "List of two-player abstract strategy games at which the best humans beat the best programs". --Fritzlein 04:50, 18 Apr 2005 (UTC)

Even a comparatively mediocre player can currently beat the best Go software. Most advanced Go software is considered to be at 5th to 8th Kyu, amateur scale, which can be related to a human amateur who has played regularly for about two to three years at most. A professional Go player would beat the very best Go engine available on any day that he wasn't dead drunk, which makes those things useless except for beginner training. --84.186.219.85 11:24, 22 October 2006 (UTC)

Trap squares

I think more explanation is needed on the trap squares, and how these can be used to remove pieces from the board. Since the concept of a trap square is quite uncommon in board games.

From what I have seen, there only seems to be one paragraph explaining the trap square and even that is buried in the middle where it is not easily recognized. --82.3.32.72 13:36, 29 Nov 2004 (UTC)

That's a fair criticism. I recall that when I learned the rules myself, the trap square was explained late in the presentation, even though it was something I was very curious about. It might even pay to explain that there is no replacement capture, as in chess. If you can edit the article to make it clearer to the average reader, please go ahead. Otherwise I will see if I can think of a way to restructure things to cater to people who are totally unfamiliar with the game. --Fritzlein 23:05, 29 Nov 2004 (UTC)

AI & Chess

"The successful quest to build a world-championship-caliber chess program has contributed essentially nothing to the field of artificial intelligence": huh? sure it has, although people don't seem to listen. Computers don't think like humans do; a machine that can do millions of multiplications a second and store gigabytes of information [i]exactly[/i], but for which fuzzy pattern matching is a non-trivial operation does things differently from a machine that is excellent at fuzzy pattern matching, but has trouble storing more than a few bytes exactly or doing [i]any[/i] multiplication. Funny that. A winning Arimaa program will demonstrate just that, IMO. That's my POV, but it helps show that this statement is POV. --Prosfilaes 03:47, 3 Jan 2005 (UTC)

My own POV is that computers are already intelligent in many reasonable senses of the word. I watched on ICC when Kasparov was playing Deep Blue, and heard people objecting to comments like "Deep Blue thinks the bishop is worth more than the knight right now," objecting on the grounds that computers don't think, and to say that Deep Blue "thinks" anything is an abuse of language. I personally believe that it was and is an extremely natural use of language, and to insist that we not talk that way is pedantry. You and I probably agree more than we disagree about whether computers display intelligence.

That said, I believe that my original statement in the article was not as subjective as you are making it out to be. The phrase "artificial intelligence" is coming to have a technical meaning which is divergent from my understanding of "intelligence". There is a community of people who are interested in making computers do certain things well that they don't do well at present, and who are as interested in how it is done as in what is done. To call that set of goals and techniques "artificial intelligence" is in some ways at odds with a common notion of intelligence, but less so than (for example) what ecomonists call "efficiency" is at odds with a common notion of efficiency. For you to object to the way in which artificial intelligence is used in a technical sense seems to me quite as pedantic as objections to saying that computers think.

If there are more accurate words to use in these contexts than "think" and "artificial intelligence", then by all means substitute them for clarity. But to insert a "He believes" in front of the statement illuminates only that you had a difference of opinion, and doesn't illuminate your grounds for objecting. Am I right that you are not objecting to the statement of fact so much as taking exception to the way most people define artificial intelligence? Do you accept that strong chess programs are strong for reasons other than pattern recognition, learning from mistakes, using neural networks, self-modifying, or any of the hodge-podge of things that are lumped together under the name of artificial intelligence? If so, then let's get this issue straightened out in some other way than the current edit. Let's recognize it as a fact that chess was conquered by computers in a way that AI people didn't find useful or applicable to the problems that interested them. --Fritzlein 07:06, 3 Jan 2005 (UTC)

P.S. It is worth noting that Bomb, the best Arimaa-playing program at present, does not use any AI techniques (as commonly defined), but rather uses techniques which have worked well for chess. They don't work as well for Arimaa as they do for chess, but so far they work better than anything else that has been tried.

This is a quick response--your message needs much more thought to respond in full--but I had AI in college, and one of the sections was on alpha-beta trees. If AI has a technical meaning that doesn't include alpha-beta trees, I would say that that technical meaning is too esoteric for wikipedia. --Prosfilaes 23:05, 3 Jan 2005 (UTC)

Hmmm... maybe the definition of AI that I've learned isn't as widespread as I think. Just out of curiosity, do alpha-beta trees have applications outside of computer gaming? --Fritzlein 05:00, 4 Jan 2005 (UTC)

Well, anywhere stratagy is needed, of course. But beyond that, I suppose that the same concept, if not the exact implimentation, could be helpful in, say, public heath sims as a way to try and determine the best course of action. Of course, it could be argued that that is just a complex game with real applications... hmm... --Kinkoblast 19:13, 16 May 2006 (UTC)

I don't see how alpha-beta trees could have application out of a game; it's a pretty narrow technique. I'm going to try something on the page. --Prosfilaes 07:09, 6 Jan 2005 (UTC)

Fantastic, I love the current edit. --Fritzlein 01:41, 9 Jan 2005 (UTC)

Images need replacement or release under the GFDL

Jimmy Wales just posted the pronoucement on the WikiEN-l that "All images which are for non-commercial only use and by permission only are not acceptable for Wikipedia and _will be deleted_." Since there are a lot of images that fall under this category here, I figured I'd drop a note on this talk page to make sure frequent contributors here were aware of it. Perhaps Omar Syed could be convinved to GFDL these images, or some other substitute set created - it seems like a fairly straightforward thing to do once one has appropriate little animal icons to copy and paste onto a blank grid. Bryan 04:46, 19 May 2005 (UTC)

I have sent the following message using the contact form at http://arimaa.com/arimaa/contact/

Hello, this concerns the informative Arimaa article currently on the English version of Wikipedia:

http://en.wikipedia.org/wiki/Arimaa

Discussion on the article:

http://en.wikipedia.org/wiki/Talk:Arimaa

The images depicting game scenarios in the article are used under permission according to http://arimaa.com/arimaa/graphics/ ; however, Wikipedia is no longer be able to use non-commercial and by-permission-only graphics. Myself and other editors of the Arimaa article would like to know if you would be willing to license the graphics under the GFDL, which is the primary license which Wikipedia adheres to. More information can be found at:

http://en.wikipedia.org/wiki/GFDL

If it is not possible or unfavorable to do this, we will replace the graphics in the article with a different icon set. A reply to this issue via e-mail or via the article's discussion page would be helpful.

-Jason McCay

--Poiuyt Man ^talk 11:21, 21 May 2005 (UTC)

I doubt Omar will GFDL the icons. How much time do we have to replace them? If nothing else, I would use the chess position template to replace all images before they are deleted. How will I know when the termination date is approaching? Thanks in advance. --Fritzlein 02:57, 27 May 2005 (UTC)

I changed the copyright on Image:Arimaa_Setup.jpg. If this is incorrect, please let me know AND THE REASON WHY! The rest of the board shots fall under the same category.--None-of-the-Above 04:40, 1 September 2005 (UTC)

The two images in the main Arimaa article were taken from the arimaa.com server's Flash interface, so I believe the change to the copyright notice is technically incorrect. The images in the Arimaa Tactics and Arimaa Strategy articles were created using David Fotland's copyrighted software, so there the notice would be correct. The images look essentially indistinguishable because they each use piece and board graphics which (I believe) are copyrighted by Omar Syed, i.e. David obtained Omar's permission to use Omar's graphics in his software. I don't know what that does to the legal status of the images, but if it somehow is less legally problematic to have images from the Arimaa server rather than from Fotland's software, even if they are based on the same underlying piece graphics, then I guess I could recreate the images. --Fritzlein 13:56, 1 September 2005 (UTC)

My thoughts exactly. I wasn't sure, that is why I did the test run. Using the exe on my HD, I could reproduce the images. It would take forever, but it is technically very possible. I still don't know if this solves the problem, but at least I got your attention. ;) --None-of-the-Above 15:38, 1 September 2005 (UTC)

As of mid-November 2005, Omar released the piece and board graphics to the public domain. See the new text at http://arimaa.com/arimaa/graphics/. As I understand it, that clears up any licensing problems with the images in the Wikipedia articles. One project for me would be to change the license tags on all the images, but I think an even better project would be to make an Arimaa-diagram template based on the chess-diagram template. That would make all the current image files superfluous, i.e. candidates for deletion. --Fritzlein 03:46, 2 December 2005 (UTC)

Human Advantage

It is correct to attribute the difficulty fast computers running an AI program reasonably well-designed exclusively for playing Arimaa have mainly to the extremely high branching factor of this game. The four-move-per-turn move cycle and consideration of the vast number of additional, meaningfully-distinct orders for casting the four individual moves comprising a turn is a related factor worthy of specific mention as well.

In attempting to pinpoint the learning psychology involved and how a human advantage manifests, I believe the current description is somewhat incorrect based upon my own related experiences in playing various chess variants (which have positive transfer to Arimaa).

When I first look at the gameboard immediately after my opponent has moved where the positions are complex, I am rarely, totally disoriented even if the moves made were unexpected since I am following the game to that point intently. However, I am usually moderately disoriented at that moment by consideration of the numerous possibilities for response, esp. any competing, important offensive and defensive priorities which must be chosen between sacrificially. I would actually assess my initial state as similar to that of a computer opponent- overwhelmed by the possibilities.

Fortunately, I am usually able to quickly identify all of the most pressing offensive and defensive theatres, inerrantly pruning a vast number of bad and trivial moves from further consideration. Then, it is a matter of course to correctly identify the very most important ONE out of typically 2-6 options for response. It often becomes clear with certainty which move is the most important one to make after a thorough analysis of all of the candidate moves is made. Sometimes this requires only a little time; sometimes this requires a moderate amount of time ... but this never requires a huge amount of time such as is the case where a computer is trapped working thru a combinatorial explosion at a given ply of search depth.

I am interested in and value the experiences of others and their opinions.

--AceVentura

Interesting speculation, Ace. Now that you mention it, I have very little clue why humans can play Arimaa well, and what the elements of the psycholgoical process are. I can observe myself as I play, but that does not provide infallible information.

Nevertheless, no matter how inaccurate my wild guess on the subject may be, the issue needs to be addressed. It is not at all sufficient to say that computers play Arimaa badly due to a high branching factor, because there are games with a high branching factor that humans play awfully compared to computers. There must be something about the game that humans are able to quantify more quickly and accurately than computers can.

From my experience I would say that it is possible that I consider and reject many moves so quickly I am unaware of the pruning I am doing. However, it seems highly implausible that I am rejecting thousands or even hundreds of possibilities each move. I am convinced there are whole categories of moves I don't consider at all, for example because I judge a certain theater to be temporarily relatively unimportant. To my mind it is qualitiatively different to neglect to consider a move than it is to consider and reject that move. You may call each a sort of pruning, but that conceals more than it reveals.

Moreover, as I ponder a position, I routinely find new moves I hadn't considered before that I then judge to be better than any of my previous candidates. This happens even in postal games where I have studied a position for half an hour or more, and suddenly see something new that solves my problems more efficiently or effectively. It does not strike me as being at all similar to selecting from a small number of candidate moves if I can bring new moves into the mix very late in the thinking process. It seems like I say things to myself such as "There must be some move which threatens his camel without exposing my cat. Aha! I have found one!" Is this not more akin to building up a set of candidate moves, than it is akin to winnowing down the multitude to a few?

I don't know how others experience their own thought processes, but I feel that I orient myself by first identifying strategic elements of the position and only second considering relevant moves. I could perhaps orient myself a different way, namely by looking at a variety of possible moves I can make and using those samples to inform me what potential the position holds, but this seems hopelessly inefficient. If there are only a handful of important possibilities, what are the chances I would stumble upon even one of them, never mind several of them that are worth comparing? No, it is much better to try to grasp the position first, and then generate relevant moves. It is a method that seems to work for me, at any rate.

That said, I am aware that my guesses about human game psychology are tenuous, and I am open to the article being edited in any number of ways. In particular (although this has nothing to do with how humans play) you are quite right that generating repetitive moves from different step-orderings consumes a significant percentage of CPU cycles for Arimaa software, and it would be a technological breakthrough to find a way generate each unique move only once. Hash tables may be used to prevent duplicated searching further down the tree when a move is generated multiple times, but it would be nice to be able to avoid multiple generation in the first place. --Fritzlein 06:28, 16 October 2005 (UTC)

Just as the opposite methods by which chess supercomputers running sophisticated AI programs and top, human chessmasters play chess (for example) extremely well must both be respected, likewise should contrasting methods of evaluation used by skillful human players.

In light of your interesting, detailed description of how you "solve the board" when it is your turn to move (in Arimaa and other board games), I no longer have any confidence that I should replace your personal account, infused into the article, with mine. Unfortunately, I fear that adding my personal account (and others who have not yet spoken-up) alongside yours in describing the human thinking process would unavoidably render this encyclopediac article too vague, confusing or self-contradictory. Instead, I recommend that existing attempts to define and describe the human advantage in Arimaa, however it undoubtedly exists, should be made much more concise so as to confine ourselves to facts we are relatively sure of (if any). --AceVentura

I agree that my guesses sit awkwardly in an encyclopedia article. I should be more clear about how speculative I am being. One way to do this would be to include alternatives, but I fear you are right that it would make the section vague and rambling. Probably more appropriate would be for me to remove that paragraph entirely and replace it with "we don't know how we do it, but we do it", or words to that effect.

Still, I feel the need to say something about the remarkable human ability to pick a good move among thousands, because the current vogue is towards oversimplification along the lines of "Any game with a high branching factor will of course give an advantage to humans over computers". This is probably outright false, but at a minimum doesn't tell the whole story. There must be features in the game which are easy for humans to spot and hard for computers to spot, and a high branching factor doesn't insure this. --Fritzlein 17:08, 16 October 2005 (UTC)

I have removed the paragraph on how humans think about Arimaa, and replaced it with a new section which doesn't try so much to answer the question as to make clear that some question needs to be answered. --Fritzlein 19:01, 20 October 2005 (UTC)

Excellent work! The section you added "human competence" nicely counter-balances the pre-existing section "computer ineptitude". Anyway, there was only one paragraph that I previously took exception to. Describing the human advantage as you did in terms of adaptation or improvement with experience is verifiably factual and simple without taking a dangerous turn into unprovable theories of psychology. Yes, it is responsible to mention that the nature of the human advantage is not fully understood. --AceVentura

Reading these articles has led me to consider what humans don't do. For example our perceptual limitations might be an advantage. Human beings cannot visualize, let alone evaluate, more than a small number of moves in a board game like Arimaa. Yet they can defeat computers. In other words an approach based on limited perception is more effective than brute force analysis or large task methods.

Perhaps it is due to our physical and mental limitations that we are even able to make sense of games like Arimaa - or the world in general. Human beings tend to perceive the world as physical objects or conceptual groupings. The rest (math, causation) is ignored. This approach was necessary for our survival as a species - a useful evolutionary algorithm perhaps. I do hope Arimaa will help! Pendragon39 20:33, 25 January 2006 (UTC)

There's also something to be said for gathering information visually, while the poor computer merely gets blind input from its keyboard. Pattern recognition and object tracking are useful functions being developed for automated systems elsewhere... board games should be no exception.

My analysis of human psychology in playing Arimaa would be as follows: The human player examines the board position visually. During the course of this examination, certain areas of the board will attract their attention. The player may make a cursory examination of less interesting areas of the board, but his or her focus will return majorily to those areas deemed to be of greater significance. The player will then examine possible moves in more detail by visualizing the movement of pieces. In terms of what is evaluated as 'interesting' or 'significant', there are two categories: opportunity and threat. More effort is given to moves or situations that appear to fit these two concepts. Pendragon39 18:14, 27 January 2006 (UTC)

Moving rules

The rules section reads: "A move consists of one to four steps." Later in this paragraph, however, the Wikipedia page claims: "[...] one may not, for example, take one step forward and one step back with the same piece, effectively passing the turn." Now that is somewhat contradictory, isn't it? Some confusion might also arise due to the fact that the term "move" is unnecessarily replaced with "turn". The official rules at http://arimaa.com/arimaa/learn/rulesIntro.html are much clearer about this: "A player can pass some of the steps, but at least one step must be taken on each turn to change the game position." I think said paragraph should be rewritten accordingly. 145.254.33.193 18:52, 8 March 2006 (UTC)Def

I agree completely. The terms "step" and "move" are meant to be equivalent but are used inconsistently with the term "turn" used to distinguish that there are 4-moves-per-turn throughout the game.

"Step" is NOT a standard term equivalent to "move" in the chess variant community although generally, limited-range pieces with the ability to move only 1 geometrically-contiguous space at a time per move are called "steppers". I presume the inventor unnecessarily and presumptuously made-up this incorrect usage. Of course, he can do as he likes on his own web page but YES, we have standards here.

AceVentura

I agree that there is some confusion between "step" and "move". I tried to be consistent in the usage of "step" as being a constituent part of a "move". A step relocates a piece one square orthogonally, whereas a "move" consists of one to four steps. Unfortunately, "move" also has lots of generic meanings. Perhaps it would be clearer to use "step" exclusively to mean moving one space and "turn" exclusively to mean one to four steps, while letting "move" retain a general meaning which is sometimes synonymous with turn, sometimes with step, and sometimes with neither. When it is important to distinguish between a step and a turn we can use the appropriate term, and otherwise freely use "move". I have provisionally made this change in the article in order to see how it looks. Does it appear clearer this way?

I purposely didn't copy "at least one step must be taken on each turn to change the game position" from the official rule, because I found that clause to be confusing when I was learning the rules. The essential requirement is that the position must change, not that a step must be taken. If the requirement were merely that at least one step be taken, then one could take a step forward and a step back with the same piece, which would have the same effect as making zero steps. My personal feeling is there is no need to mention the requirement of taking at least one step at all, because it is subsumed by the requirement of changing the position.

I don't think the sentence, "A turn consists of making one to four steps," contradicts the sentence, "A turn must make a net change to the position." The latter merely restricts the former. If it is still confusing this way, I'll try again to think of better wording.

--Fritzlein 20:15, 9 March 2006 (UTC)

Although in popular gamers' language, "move" is sometimes used synonymously with "turn", this is probably due strictly to the commonplace truth of the assumption in most board games which allow 1 move per turn. In Arimaa, where such an assumption is false, "move" should never be used synonymously with "turn". By the way, "turn" has a sufficiently distinct definition as the limited duration, measured in "moves", that one player can exclusively control or operate upon the game following the rules.

Your work thusfar has improved the clarity of this article markedly, with respect to this issue.

AceVentura

Ratings

How are player ratings calculated? What does "RU" mean? Thanks. -- 86.120.225.27 10:43, 22 March 2006 (UTC)

RU means Rating Uncertainty. See this page at arimaa.com for the whole Arimaa Rating System. -- Jokes Free4Me 13:23, 29 March 2006 (UTC)

Chess variant?

I don't think this really should be categorized as a chess variant. It has almost no features in common with chess: movement, capturing, and even the goal are completely unrelated. It has about as much in common with checkers as it does with chess. It seems more like it was inspired by chess (and the strength of computer players thereof) than a variant of it. You can use a (slightly modified) chess set to play it, but that's it. — Gwalla | Talk 23:09, 5 November 2006 (UTC)

It is considered a chess variant because a) it was inspired by chess, b) it is played with a chess set, and c) it is a strategy game. Yet I agree, it about as much in common with checkers as it does with chess. --Sibahi_talk 19:01, 6 November 2006 (UTC)

To me, playing Arimaa doesn't feel like playing chess in the way that playing bughouse feels sort of like playing chess, so I also am uncomfortable classifying Arimaa as a chess variant. On the other hand, the equipment is the same, and that isn't accidental. Syed realized that a game could be made computationally more difficult than chess by enlarging the board and adding more pieces, but he wasn't out to "complexify" chess. So not only was Arimaa inspired by chess, it was designed with the constraint of using the same board and pieces. For example Arimaa has in common with chess (and not checkers) that each player has six kinds of pieces. I wouldn't introduce Arimaa as a chess variant, but it seems reasonable to classify it as such. --Fritzlein 04:02, 7 November 2006 (UTC)

Standard Chess Pieces Substitution

I have not researched to find-out when this practice started or who was responsible for it but this article, in its current state, contains a violation of standards that surely has a maddening, chaotic effect upon some chess and chess variant hobbyists.

[I know the inventor of this game likes to boast that it can be played using a standard chess set.]

The substitution of the standard chess pieces (the king, queen, rooks, bishops, knights and pawns) for Arimaa pieces (the elephant, camel, horses, dogs, cats and rabbits) is totally unacceptable for an article within the chess variants category esp. since the Arimaa pieces move and capture in a totally different manner than chess pieces.

To prevent frustrating and confusing the reader to the point of judging this article to be insane, I am going to terminate with prejudice ALL chess piece diagrams and mentions of chess pieces within it as being non-applicable and inappropriate.

Noone is so economically disadvantaged that they cannot afford to have the right piece graphics running on their computer. So, it will require a damned good reason (and quickly) to dissuade me.

AceVentura

No one is too economically disadvantaged to have a computer? See Homelessness and Malnutrition for a couple articles that disagree. Even among Wikipedia articles, there are many using from libraries and schools. If there is a standard, citable mapping from the chess pieces to the Arimaa pieces, we should mention it.--Prosfilaes 14:41, 25 January 2007 (UTC)

I only meant that there is no excuse for anyone who already owns a computer to not obtain and use the right Arimaa piece graphics. Yes, I realize most people in the world cannot afford to purchase a computer.

Don't you understand that there is no legitimate need to accommodate a mapping of chess pieces for Arimaa pieces at all? Why substitute incorrect, chess pieces when you can use original, correct, Arimaa pieces just as easily? After all, Arimaa was invented in 2002 for use with computers. --AceVentura

Who are you to say what's legitimate? A lot of people don't like to be attached to a computer all the time. Anyone who wants to play with another person face-to-face, or likes to study the board with a copy physically at hand (a kinesthetic learner) would need a Arimaa to chess board map or an actual physical board.--Prosfilaes 17:21, 25 January 2007 (UTC)

People do play Arimaa on physical chess boards. It is possible to indicate the standard mapping of Arimaa pieces to chess pieces without violating anyone's sensibilities by calling Arimaa a chess variant. --Fritzlein 05:51, 3 February 2007 (UTC)

Advertisement?

I don't understand the advertisement tag, and I have removed it. There are no Arimaa sets for sale. Playing on the arimaa.com web site is entirely free. How can commercialism enter in if no money changes hands?

If you have a specific frustration with the article, let's discuss what needs to be changed, rather than slapping on a generic tag that doesn't apply very well. In particular, if the article seems to praise Arimaa in a way that the game doesn't deserve to be praised, let's discuss that specific sentence or section. --Fritzlein 05:51, 3 February 2007 (UTC)

Technically, you are correct that the advertisement tag is inappropriate. [Note- I would have used a fraud tag, instead, if I could have found one.] So, I will leave the advertisement tag off. However, noone will be as fortunate at expunging my remarks-edits from this talk page. Moreover, any potential threats of a libel lawsuit, publicly or privately, will not achieve a cheap, easy victory as censorship or retraction (under insincere duress) that is intended.

The $10,000 prize offerred to any AI programmer who can beat an expert human player at Arimaa USING AN ORDINARY HOME COMPUTER is fraudulent by being virtually impossible. This feat is calculated correctly by the inventor as being orders of magnitude out-of-reach by state-of-the-art CPU's, regardless of the quality of the written program in terms of its efficiency, resourcefulness or tactical soundness, due to the extremely-high, estimated branching factor of Arimaa. [Of course, this offer is wisely withdrawn by 2020 in case of some unforeseen, quantum leap in computer technology.]

This situation is analogous to offerring a programmer a huge prize if he/she can develop an application that can play a normal, 2-hour DVD movie flawlessly on an IBM 386 running with a 12 Mhz CPU! [Frankly, I am surprised that any programmer has even attempted to win the $10,000 prize (precariously assuming that the money even exists).] Yes, it profoundly annoys me to hear about such nonsense. No, success is not humanly possible when working under intentionally impossible conditions and restrictions. Even the most objective, realistic presentation of this $10,000 prize offer within the article insults the readers' intelligence and violates the important standard whereby Wikipedia should not be used as an instrument to perpetrate fraud for the benefit of a commercial interest. It is by way of this article's silent implication that this prize offer can possibly be legitimate that is totally fails to meet the standard of trustworthiness to the reader.

Please do not hold any illusions to the contrary? Arimaa is definitely a commercial product with the long-term goal of generating profits for its inventor well above all expenses incurred for-

1. The US Patent for Arimaa rules.

2. The US Trademark for "Arimaa".

3. The development of a program that plays Arimaa.

4. Running a commercial web site to host, play and popularize Arimaa.

Incidentally, it is noteworthy that the free program that plays Arimaa is crippleware. The fully-functional program costs money ($19.95 US).

What I want specifically is for all mention of the quest for the questionable $10,000 prize to be removed from this page. Individuals can find-out all about it, if foolish enough to be interested, at other web sites linked to from here.

--AceVentura

Although you cannot imagine how the prize could be won, that does not mean that it cannot be won. How can you accurately estimate what algorithmic breakthroughs there might be in the next 13 years of computing? 99of9 14:41, 3 February 2007 (UTC)

Making an offer for something widely believed to be impossible is not fraudulent. There is no theoretical reason why a modern desktop can't beat an expert human at Arimaa; it just happens that the algorithms we know won't do the job. Part of the stated reason for the contest is to encourage the creation of better algorithms. It would be interesting to add some cited criticism of the game and the contest, but I see absolutely no reason to delete the information of the prize from the page.--Prosfilaes 14:53, 3 February 2007 (UTC)

So, I have 2 editors (thusfar) making the serious contention that the prize contest is NOT impossible to win at all.

The reason? [Prepare for a huge, run-on sentence] Modern-day AI programmers, despite many decades of collective experience in designing and testing their work using board games and chess variants, are all such extreme dumb-asses that they have totally overlooked simple algorithms [Note- Complex algorithms would consume too many resources as they cycle.] useful within evaluation functions to play Arimaa that could help speed-up the search for incisive moves by MANY orders of magnitude thereby making a victory for an ordinary PC realistically possible against a human expert.

From whom did this dubious propaganda originate- the inventor? Are not you two merely splitting hairs over the difference between "absolutely impossible" and "virtually impossible"?

--AceVentura

Why would complex algorithms consume too many resources? Optimal sorting and searching algorithms are considerably more complex than the naive simple solutions. What matters is the cycles used, not the code size. Your emphasis on evaluation functions misses the point, IMO. For one thing, a lot of work has been done on the code that calls the evaluation functions, like MTD-f and Negascout.

More importantly, I think you're looking way too small. The propaganda you complain about comes from AI researchers who have been swearing full AI is right around the corner since five minutes after the Enigma machine cracked its first code. I don't see any chance of getting good solutions to hard AI problems like translation and leave games that can be described completely in a few lines of code without good solutions. Maybe it will come out of neural nets or fuzzy logic or more likely something no one expects, but I expect that alpha-beta type searches are not the end-all and be-all for game playing.--Prosfilaes 13:50, 4 February 2007 (UTC)

If and when Omar Syed begins to generate revenue (never mind profit) from his holding of the Arimaa patent and copyright, additional scrutiny of the article will be appropriate. As it stands, however, he has not made a dime, has distributed over $7,000 in prizes to developers, players, and spectators, and has spent several thousands more on the infrastructure that would enable someone to win the $10,000 grand prize. A neutral observer might well infer from Syed's actions that he is more interested in spurring AI research and/or in proving a point than he is interested in making money. I am not neutral: I have personally begged Syed to mass-produce Arimaa sets for the purpose of making money and popularizing the game, and he has repeatedly refused.

The lack of commercialism aside, this article is as factual as those on for-profit games, for example, Axis and Allies and Stratego. Even if Syed were making money hand-over-fist from Arimaa (and I hope he does some day), the text would be appropriate.

The thrust of your objection appears to be that the $10,000 prize will never be paid out, either because it doesn't exist, or because the conditions of victory are impossible to meet. You present this as a fact. Quite apart from what the text of the article should read, you and I now have an opportunity to make a wager we both will consider profitable. Whether you believe the Arimaa Challenge is "absolutely impossible" to win, or only "virtually impossible" to win, you are quite certain that there isn't even, say, a 9% chance that the money will be paid out. Therefore you should be happy to give me ten to one odds against the event.

At 10:1 odds, I will wager any sum you care to name that Omar Syed will pay out the $10,000 prize by 2020 to someone who has met his Arimaa Challenge criteria. Thus if you wager $100, and Syed doesn't pay out the prize (because nobody has won or because the offer was not genuine or for any other reason), I will pay you $10, whereas if Syed does pay out the prize, I will win $100 from you. We can each forward our stake to a neutral third-party arbiter to hold and pay out to the eventual victor. (If we make the bet, I will naturally recuse myself from defending the Challenge, so as not to have a conflict of interest.) So, how much do you want to bet? --Fritzlein 20:04, 3 February 2007 (UTC)

Gambling is generally illegal in US-America (where we both reside) except in certain places and under restrictive conditions where it is permitted. Besides, I never gamble. Gambling either hurts yourself or the person you win against. Moreover, I do not want to be bothered by keeping an "unfinished business" post-it-note in my financial folder until 2020. I appreciate your brave offer, though, in the spirit it was intended. It is very sporting and it demonstrates confidence. Despite my strong doubts, I actually hope I am someday proven wrong.

In any event, I am outnumbered by other editors who disagree with my proposed changes.

--AceVentura

I respect your decision never to gamble. My mother would have been happier with me if I had felt the same way you do.

Interestingly, Syed himself rates his chance of paying out the $10,000 prize by 2020 near 50%. Indeed, he is quite convinced that if we develop sufficiently smart software fifty years from now, that same software running on a vintage 2002 home computer will be able to beat the best humans. In other words, he feels that hardware is already fast enough if only the program is smart enough.

I rate the chances of a payout by 2020 around 20% myself. Everyone I know considers the chances near zero without an algorithmic breakthrough, i.e. the current programs running 100 times faster wouldn't be enough. Indeed, it is possible the current programs running 10,000 times faster wouldn't be enough, so the uncertainty comes mostly from software approaches that haven't yet been tried.

If someone makes learning software to play Arimaa well, Syed will actually be happy to pay out the prize money. The only way he will think he has "lost" is if boring, brute-force techniques turn out to be sufficient. But in some sense he will have accidentally won simply by creating a game that is great fun to play quite apart from whether computers are good at it.

I actually hope that computers can never beat the top humans at Arimaa. If that happens, it will be just one more step towards my obsolescence. But as much as I hope the Arimaa Challenge is insurmountable, I sadly expect it will fall long before computers stop getting faster. --Fritzlein 00:15, 4 February 2007 (UTC)

Amusingly, under Bayesian predictions, 50% odds is what one should predict if one is utterly and completely ignorant of any useful information with regard to the subject matter. --Gwern (contribs) 08:44 11 August 2007 (GMT)

Challenges

The table lists four, the text says three... --belg4mit @ 2007-02-25 —The preceding unsigned comment was added by 71.192.58.23 (talk) 19:58, 25 February 2007 (UTC).

The fourth Challenge Match is in progress. When it is completed (hopefully Thursday) I will update the text to match. --Fritzlein 02:53, 27 February 2007 (UTC)

Prize Value

On the main page, the prize is listed in a chronological table which suggests that the prize value has increased by ~75%, to (currently) $17,100. However, I can find no documentation of this prize increase on the site. Can someone please explain the disparity in numbers?

The article has been edited to show the source of the funds. The additional pledges are short-term, i.e. only $10,000 has been pledged through 2020, so in that sense the original prize fund has not grown, but in all probability whoever designs a world-beating Arimaa program will win rather more than $10,000. —Preceding unsigned comment added by 63.250.222.254 (talk) 17:48, August 30, 2007 (UTC)

Material Handicaps

It is a very dodgy exercise to compare Arimaa material handicaps to chess material handicaps because Arimaa is much more of a blockade, breakthrough, and control while and chess is much more of a capture game. Anyway if you look at the material evaluation functions that the computers use, a camel handicap in Arimaa is more like a rook handicap in chess, so thinking of it as "queen odds" arises mostly from the fact that the queen is used to represent the camel when playing with a chess board. —Preceding unsigned comment added by 96.226.2.62 (talk) 17:01, 11 September 2007 (UTC)

Who cares?

For such a long article into which people have obviously put a fair bit of work, there's a really huge gaping hole: there is no outside context for this game whatsoever. As far as the article is concerned, it seems only about ten people play it (listed in the challenge table).

Some third-party reviews are in order. Or, if it's more of a AI construct than an actual game, then maybe some third-party mentions in the AI literature on its interest. Or something, anything. Regards. --192.75.48.150 (talk) 18:07, 25 June 2008 (UTC)

I play it. I like it. I actually first heard of it here on Wikipedia. Yes, it may deserve links from more places, but I am glad that it's listed. So I care, for what ever that's worth. 165.125.144.16 (talk) 19:01, 10 April 2009 (UTC)

It looked like an interesting game, until I got to "pulling pieces". That's an unintuitive, gratuitous complication that makes the game really retarded. —Preceding unsigned comment added by 212.71.114.220 (talk) 13:54, 24 August 2010 (UTC)

Seven reasons

"It has been argued that a computer has beaten the world chess champion but not beaten the human in the Arimaa challenge because of six reasons". I would think there would actually be seven reasons. 7: The low prize pool value of 10,000 dollars is not enough to attract serious programming talent and the relative unknownness of Arimaa means that many programmers simply don't know about the challenge or care to spend time on it. 122.49.138.156 (talk) 08:50, 16 February 2011 (UTC)

I think your reason of a low prize not attracting sufficient talent is included in the article’s reason 1. Reason 1 says it is a new and relatively unknown game therefore there is insufficient programming effort. Your reason of small prize also causes insufficient programming effort. A larger prize would attract more talent. When computer chess was starting in 1960s and 1970s there where no large prizes but there was many people working on computer chess. I do not think there will a large prize for Arimaa. It is a catch 22. A sponsor will offer a large prize when there is already a large audience. The sponsor wants to get publicity from the large audience. Ibm hired the chess programmers and sponsored a large prize for the computer vs. human world champion chess tournament because ibm could get publicity from the large number of people who know about chess. Ibm was not interested in chess. A company will offer a large prize for Arimaa when there is a large audience for Arimaa. On the other hand, there are classic games that do have large programming efforts without large prizes. Games with large programming effort are Shogi and Go. In spite of the large effort man beats the computer in these games. Shogi is more like chess than Arimaa. Mschribr (talk) 11:18, 23 February 2011 (UTC)

There are some definite issues with the speculation on the computer playing and some failure of NPOV. A modern computer chess program is not brute force but heuristic. The only issue I have about mentioning the prize is if it has not been mentioned in RS but the conditions are odd.Tetron76 (talk) 18:07, 1 March 2011 (UTC)

I do not understand what you are saying. Can you be more specific? What is RS? Modern chess computers do not use brute force or heuristic. They mostly use alpha-beta pruning. Mschribr (talk) 21:26, 2 March 2011 (UTC)

Anonmoos Removed Correct Statement

AnonMoos removed the statement that hydra it is a custom made 32 node Xeon cluster, with a total of 64 gigabytes of RAM. He said this is opinionated personal commentary. It is not. These are facts. If he cannot explain his actions, I will undo his changes. Mschribr (talk) 01:33, 23 March 2011 (UTC)

That's obviously not why I reverted your edit, so your remarks above are somewhat disingenuous...

I know that you feel very strongly that the why-Arimaa-is-harder-for-computers-than-chess arguments are all wrong, but (as emerged in the past discussions above), many people on and off Wikipedia disagree. For this reasons, unreferenced blunt statements about how things are allegedly "false" are not suitable for the article in that particular form. AnonMoos (talk) 09:01, 23 March 2011 (UTC)

Let us try to leave feeling out of this discussion. Ok? The article compared the Arimaa challenge match to the Man vs Machine World Team Championship. I stated the comparison is false because hydra is custom made. Does that belong in the Arimaa article? Mschribr (talk) 11:33, 23 March 2011 (UTC)

Why don't you rephrase it as a brief factual statement backed up by a source, instead of as a lengthy opinionated screed? AnonMoos (talk) 15:21, 23 March 2011 (UTC)

Reliable sources: Google Scholar

I just noticed that some scientific papers, such as Master's Thesis works, have been written about Arimaa. You can find them with Google Scholar: http://scholar.google.com/scholar?q=arimaa Someone might want to read them and introduce some additional references into this article. —ZeroOne (talk / @) 23:23, 26 March 2010 (UTC)

See also http://arimaa.com/arimaa/papers/ Fritzlein (talk) 02:11, 17 April 2011 (UTC)

Assessment comment

The comment(s) below were originally left at Talk:Arimaa/Comments, and are posted here for posterity. Following several discussions in past years, these subpages are now deprecated. The comments may be irrelevant or outdated; if so, please feel free to remove this section.

Needs more references, definitely not B class. Reassessing to start class. Airplaneman ✈ 23:25, 10 April 2010 (UTC)

Last edited at 23:25, 10 April 2010 (UTC). Substituted at 20:09, 2 May 2016 (UTC)