Talk:Chess engine

old talk

I have disambiguated many references on this page that formerly linked to unrelated articles because the chess engines in question had the names of common items. However, as I was in the P's (I was working alphabetically) I began to wonder if this is in the Wikipedia's best interest. Certainly, to add "(chess)" to the links on this page is productive, as it eliminates references to unrelated articles, even if it only creates a redlink in its place. But, as I disambiguated these dozens of chess engines on other article pages, I found that I was just adding redlink after redlink to the bottoms of articles and disambiguation pages. AFAIK, dab pages aren't really supposed to link to nonexistent articles if possible. I also felt really bad tacking on this barely related redlink notice to some wonderfully written pieces of brilliant prose. What should be done here? I know I've left the job half-finished...should I revert or continue? —Ed Cormany 22:15, 13 May 2004 (UTC)[reply]

Adding red links to disambiguation pages is absolutely fine. I'd say it's quite useful, in fact, because there may be several reasonable ways to disambiguate something, and if you don't add them, people won't know know what other articles are pointing to (for instance, there may be a risk of somebody linking to Adam (chess engine) rather than Adam (chess) unless there's a pointer to the latter from Adam). And putting a link at the bottom of an article (even a really good one) doesn't really take away from the article itself, IMO, so I wouldn't say there's anything wrong with what you're doing.

That said, I wouldn't worry too much about carrying on with it if you find it boring or have pangs of guilt or whatever. Most of the engines listed on this page are rather obscure and I doubt we'll have articles on them any time soon. I doubt we'll even have links to them from any article other than this. So although it's a job that will probably need to be done at some point, it's not a pressing issue, and nothing much will be hurt if it isn't done right now. --Camembert

Thanks. Your words are encouraging. I'm not sure if I'll carry through to the end of the alphabet or not, but I may well now. =) —Ed Cormany 01:38, 14 May 2004 (UTC)[reply]

The other thought is that having red links eventually adds up if a topic *is* referenced from multiple places--might appear on a most-wanted list or, when someone is considering whether to create a new page and is staring at a blank "Editing" page, they can click What Links Here and see whether there are articles that reference it. Sometimes helps me decide whether to create an article that will fill in a few red links automatically. Anyway, nice job (as I also said on your user page). Elf | Talk 04:15, 14 May 2004 (UTC)[reply]

Most of the programs on this list should not have encyclopedia articles. There just is not enough to say about them to ever get a verifiable article that will ever be greater than stub-length. I am going to unlink all but the ones which already do have an article. Please do not re-create the link until you also create the detailed article about the engine itself. (And as a caution to new users, people who create many self-evident articles with the sole content of "XYZ is a chess engine" as a way around this principle have been accused of vandalism for that behavior.) Rossami (talk) 03:08, 11 Jun 2005 (UTC)

Shouldn't there be a category for computer chess players too? --Malathion 5 July 2005 02:46 (UTC)

Yes, it happened. See Category:Chess engines. --IanOsgood 17:06, 26 September 2007 (UTC)[reply]

Licensing

Is it possible to add information about the license of each engine, such as GNU, BSD, closed-source, or whatever? Nicholasink 15:11, 4 March 2006 (UTC)[reply]

I think that the title of the page should be more explicit, and that a phrase in the beginning of the document should present the target of the list and what is allowed. Is the target of the list presenting what is a chess engine and how it works? In this case the list should only contain the most representative engines. Is the target to present the best engine available to the public in order to chose one that fit your need? In this case the list would contain only the best one. I would perhaps let this page in this actual state - open to all chess engines - and create a new one with a link to it. So there would be less problems. Outsiders with new engines could put there engine one the wikipedia and have less excuse to pollute the page you have in mind.

List is of dubious value

This list is of dubious value. To the layperson, it is just random names. It does not distinguish between software and dedicated computers. It does not distinguish programs of historical interest from those of current interest. Half these engines are non-notable personal projects. Any static list will become obsolete almost immediately. I recommend removing the list entirely, perhaps merging the notable programs into computer chess, perhaps introducing categories instead. --IanOsgood 20:37, 14 August 2006 (UTC)[reply]

"Wikipedia:Listcruft is exactly what I was talking about. I'd propose this page for deletion, but it seems to have a lot of history, therefore interest. Could we at least categorize the links on this page a little? I propose at least deleting every item that does not have an internal or external link. --IanOsgood 00:26, 11 October 2006 (UTC)[reply]

Note: around this time the page was renamed from List of chess engines to Chess engines.

Free chess engines

List of non-notable engines removed, so there is no temptation to add them back to the article. --IanOsgood 16:57, 4 November 2007 (UTC)[reply]

Took out those engines that now have links Laetoli

Could you please give your reference for the ELO ratings? Peferably list both the testing agency, date of testing, and the reference hardware used. Without a reference, the ratings could be removed at any time. Also think hard about the notability of these engines. We don't wan't this page to turn back into a link farm. --IanOsgood 21:38, 2 November 2007 (UTC)[reply]

I took the ELO rating from three sites and normalised the data because the ELOs depend on CPU speed. If you want I can link to those sites or I can show my data and how I derived it. What do you think?

As regards notability, I see your point but a lot of people like to run compative tournaments or play against an engine with a specific strength, i.e. no one can beat Rybka, so I think they're all useful. Writing a chess program is not trivial with scarcely any prospect of financial gain, so I doubt anyone is going to do it in order to harvest links.

What I had intended to do once I'd gathered all the engines plus data is list them according to ELO strength and include the author's name and country they come from. If you prefer we can dump those for which I cannot find ELO rankings.

I noticed someone removed the approximate sign (~) I'd put that in to show that the rating were just that (they're rounded to the nearest 5). Laetoli —Preceding comment was added at 23:00, 2 November 2007 (UTC)[reply]

Merging data from multiple sources constitutes original research, and so does not belong on Wikipedia. Please reference a single source for ratings, or remove them altogether. We already give links to the major rating lists if folks want to go look up ratings themselves.

If you want to work on the page, you could cull the list only to those few dozen free engines that are near the top of one of the rating lists, as a measure of notability. --IanOsgood 04:35, 4 November 2007 (UTC)[reply]

Okay, I concede defeat. Took out Ikarus (private engine & not freely available) and put in authors and country codes. I hope this is alright by you. Also gave link to ratings page. Laetoli —Preceding comment was added at 12:16, 5 November 2007 (UTC)[reply]

The list of "Freely available chess engines" includes Ruffian 1.0.1 as "(latest free version)". In fact the latest free version is Ruffian 1.0.5, available from the Arena download page. Version 1.0.5 is usually regarded as the latest free version of Ruffian in computer chess community - it is tested in CCRL 40/40, CCRL 40/4, CEGT 40/20, CEGT 40/4, FACE. Also WB/UCI chronology lists Ruffian 1.0.5 as the last free version. Ruffian 1.0.5 was first released on 2003-03-19, AFAIK. 333ES (talk) 08:56, 6 August 2009 (UTC)[reply]

The official web page for Ruffian, however, does not agree with any of the aforementioned sources, which only has Ruffian 1.0.1 available to download. —Preceding unsigned comment added by 202.170.162.99 (talk) 21:21, 11 May 2010 (UTC)[reply]

Please elaborate on the comment for reversing changeset 14:20, 11 May 2010. The comment makes no sense. —Preceding unsigned comment added by Corrie.engelbrecht (talk • contribs)

Sure. I was suspicious when I saw several of the engine ratings drop by over a hundred points (for example, Booot, crafty). Looking closer, I saw many version regressions (for example, Crafty from 23.0 down to 22.1). I thought I saw that you were not using any of the open source versions of engines in your update. (Open source engines are colored orange in the CCRL list, free but closed source engines are colored green.) For the purposes of this list, both types are considered free engines. Maybe that wasn't the problem; I see that some of your ratings are for open source engines. Did you check that you had the top rated free version of the program? Many of the engines have been updated since the last edit. Maybe the "best versions only" list would be more useful: http://www.computerchess.org.uk/ccrl/4040/ --IanOsgood (talk) 22:34, 11 May 2010 (UTC)[reply]

The change you reverted included exactly the information you say you want. The header of the ratings column was changed to say that it was the best version according to the ratings. It is possible that there are omission, I will add those later, but I am going to reinstate the changes I made because you have presented no evidence that your reversal was warranted. Please take more care before doing things like this, because you did not do due dilligence before taking your actions, and this is wasting my time, and yours. —Preceding unsigned comment added by Corrie.engelbrecht (talk • contribs) 00:37, 12 May 2010 (UTC)[reply]

OK. I'll help. --IanOsgood (talk) 02:46, 12 May 2010 (UTC)[reply]

Thanks for being so friendly. Now I sound like a bully:) And thanks for the help! --Corrie.engelbrecht (talk) 07:05, 12 May 2010 (UTC)[reply]

Proposed page move

I propose that this page, currently at Chess engines should be move to Chess engine to conform with WP naming conventions regarding plurals of nouns. The page cannot be moved by a non-administrative user because the redirect page Chess engine has a history. BlueValour 23:25, 15 November 2006 (UTC)[reply]

NOTE: move was made. --Guy Macon (talk) 19:47, 30 March 2012 (UTC)[reply]

Source Code

It would be nice to list which engines are provided with source code. I know the following do: Amundsen, Amy, Arasan, BBChess, Crafty, Faile, Fruit, Glaurung & Scatha, GNU Chess, Knight Cap, Nero, Pepito, SJeng, Toga II, Witz. (add more if you know)

This article should merge with Computer chess

Subject says it all Rocksong 23:16, 15 February 2007 (UTC)[reply]

Oppose - Computer chess is already long and a merge of the two articles would push it over the top. I agree that some rationalisation is needed but length indicates two articles though not necessarily split as now. BlueValour 05:45, 16 February 2007 (UTC)[reply]
Oppose - As per Bluevalour. —The preceding unsigned comment was added by Aarontay (talk • contribs) 05:47, 16 February 2007 (UTC).[reply]
Oppose - This page started out as a "list" article, with only minimal explanatory paragraphs. Such a (large) article should remain separate from the (large) main topic article. I would instead suggest refactoring by moving content related to computer chess theory back into the main computer chess article. --IanOsgood 17:12, 26 September 2007 (UTC)[reply]

Then what is the difference between the subject matter of the two articles? "Computer chess" and "Chess engine" are almost synonyms. Rocksong 06:20, 16 February 2007 (UTC)[reply]

Not quite. Of course any chess program by definition would have an engine (the component that actually runs the algorithms), but in the Computer chess circles (as you probably already know), chess engine typically refers to the relatively new idea (1990s?) of splitting a chess program into two parts so the engine or brain can be ported to run on any other interface that runs the same protocol. It is allied to the concepts of Xboard Communication Protocol, UCI, etc and there has being a trend in the past 2 decades for modular engines to be released even by commercial packages. It is also a more specialized term than computer chess which is devoted generally to computer chess programming techniques (and history of). You can argue whether this concept is notable, but it is clearly distinguishable from computer chess topic.Aarontay 07:10, 16 February 2007 (UTC)[reply]

But that's not what this article is about. All the "chess engines" concepts are (AFAICT) already there in Computer chess, and don't need to be duplicated here. Apart from the section "Increasing strength" (which I think is out of place and belongs in Computer chess), this article more about listing/comparing the different engines, so it should be renamed to reflect that. So I guess I've changed my mind and I'm proposing a rename (to keep the article focused) rather than a merge. To put it another way, "Computer chess" is the main computer chess article, and this is one of the sub-articles, and should be named to reflect that. Rocksong 10:20, 16 February 2007 (UTC)[reply]

Actually that's what the article is about as can be seen by the first two sections , though it seems that it was duplicated in the main computer chess article as well. I agree "The increasing strength portion" is not quite relevant here. As you know, the main effect of this chess engine portability factor is that it makes engine versus engine contests a lot easier, and as a result, many hobbyist focus exclusively on chess engine strength that is perhaps why it was added here. If you give me a week, I can expand the first two sections with more meat (history etc). A seperate issue is what to do with the lists of chess engines as discussed above, that one I'm not too concerned. I propose keeping the chess engine article, and moving the lists to another article. Aarontay 02:38, 17 February 2007 (UTC)[reply]

But in any case, this article is not about "chess engines" as you have defined. It even lists dedicated hardware. If you look through the history of the article, you will see that until Oct-Nov 2006 it was almost nothing but a list of chess engines/computers. Since then well meaning editors have added extra comment, but honestly I think it is unnecessary and distracting, duplicating what is already in the very thorough Chess computers article. Now in my opinion, the fact that engines can be separate to the interface is pretty uninteresting, and really only deserves a sentence or two in Chess computers. (The real revolution has been that computers have become so powerful that almost all chess computers are now software - back in the 80s / early 90s you could buy a "chess computer", now these things I think are extinct.) Let's just focus on improving the main computer chess article, rather than doing the same (or very similar) work in two different articles. Rocksong 06:58, 18 February 2007 (UTC)[reply]

IMHO This sounds a lot like "I'm not interested hence it is not notable". I submit that the majority of editors including the originator of this article actually recognize this concept, though many are newcomers so they are not able to place it in context, but enough "well meaning editors" do. You are a well meaning editor as well, but correct me if I'm wrong but you are not quite familiar with computer chess as opposed to other well meaning editors in this article? Aarontay 14:19, 18 February 2007 (UTC)[reply]

Heh, above I accidentally wrote Chess computers instead of Computer chess. Ironically it turned out not to matter because Chess computers (and for that matter Chess computer) is a redirect to Computer chess). It really doesn't make sense to me that typing "Chess computers" brings up one article, while typing "Chess engines" brings up a quite different article, especially since the "Chess engines" article lists not only chess "engines", but also computers. I still think it makes sense to rename this article to "List of chess engines", and incorporate any discussion portions into Chess computers. Rocksong 07:03, 18 February 2007 (UTC)[reply]

I see that people have being overly generous with the use of redirects. But I suspect if it were up to you, there would only be one article - Computer chess. As I said, I don't particularly care for lists so if you want to move the lists to Lists of chess engines great, But Chess computers would imply hardware or even specialised hardware, so I'm not sure why you think the discussion here is suitable to be placed there. Aarontay 14:19, 18 February 2007 (UTC)[reply]

If it were up to me, there would be one main article - Computer chess - with lots of sub-articles. The reality is that chess computers have evolved gradually, from specialised-hardware-only 20 years ago, to almost invariably software-only now. Because the change was gradual, and the concepts overlap so much, the chess hardware and software articles need to be tightly coupled. In any case, there is no division at the moment: Computer chess has a lot on software-only, while Chess engine has some stuff on specialised hardware. I don't want people to waste time on two parallel articles. Rocksong 23:10, 18 February 2007 (UTC)[reply]

And I don't want to dumb Wikipedia down. No offense but to say that software and hardware needs to be tightly coupled, shows a studied ignorance of the origins of the field. Computer chess as a field has always focused on AI and algorithmic advances (Turing and Shannon) independent of the exact medium used as typified by Turing's historic paper/hand simulation of a chess program. That has nothing at all to do with specialized hardware vs software. Aarontay 07:55, 19 February 2007 (UTC)[reply]

You don't get my point, do you? That hardware and software can do the same thing. Look I'm not for dumbing down, I'm for avoiding duplication. I offered the suggestion to help avoid duplication. If you don't want to take that suggestion, than I guess I can't stop you. Also I strongly advise against calling other editors ignorant. Rocksong 08:46, 19 February 2007 (UTC)[reply]

As a compromise, I've changed the chess computer redirect to point to chess engine. This seems more appropriate, looking at the backlinks. --IanOsgood 17:12, 26 September 2007 (UTC)[reply]

Personal programs

What is the purpose of these without qualifying why they are here? (i.e. defeated a nother notable program) People can't buy or even download these and it doesn't seem much more than a list of names, less useful than the freely downloadable programs that we are not allowed to add to the page. Laetoli

Good point. This section is perhaps a relic of the initial categorization of the flat list of names. The only two notable ones are Ikarus and Ferret (winners of the World Computer Speed Chess Championship). I would not mind deleting the section entirely. --IanOsgood 18:16, 6 November 2007 (UTC)[reply]

I've removed the section and added a reference line to Ikarus & Ferret at the end of the list of free engines.Laetoli —Preceding comment was added at 15:51, 12 November 2007 (UTC)[reply]

Chess engine rating lists

I have removed the text "results are statistically significant" because the engines are rarely differentiated from those a few places above and below them to a 95% statistical significance. For example, if you look at http://www.computerchess.org.uk/ccrl/4040/ you will see that most of the engines could rise or fall several places within the given margins of error. On this list, only the top engine, Rybka, is safe in its position within the given margins of error. The 21st placed engine, Delfi, could fall to 28th place within the given margins of error even if those below 28th failed to rise (which they probably would if Delfi lost games to them, as may well be the case in this scenario). Laetoli

There was a similar wording at Chess engine rating lists:

[...] list uses statistically significant measurements of chess engine strength, [...]

I changed it to:

[...] list uses confidence interval estimates of chess engine strength, [...]

which is hopefully more to the point here. GregorB (talk) 09:54, 27 April 2008 (UTC)[reply]

Merge proposal

I think it would make sense to merge List of chess engine rating sources into this article, section "Chess engine rating lists". Please leave your comments about the proposed merge here. GregorB (talk) 21:41, 18 January 2009 (UTC)[reply]

Agree It seems like a good idea to me. Bubba73 (talk), 21:51, 18 January 2009 (UTC)[reply]

Since there was no objections, I merged it - for better or for worse.

Sadly, references from the List of chess engine rating sources are now the only references this article has. I hope this will improve in the future.

There are only four engine rating lists in the table. For more about the criteria used to include them (and exclude others), see Talk:List of chess engine rating sources. I think these four are probably the only notable rating lists (other than some specialized lists, covering e.g. palm devices and the like, that could still be added to the article). GregorB (talk) 15:23, 25 January 2009 (UTC)[reply]

Chess Computer should redirect to Computer Chess, not here

I'm sorry if I not some semantic genius, but when I type in "Chess Computer" I would expect to go to the article describing the history of chess-playing computers, not this article. At the least, the top of this article should contain a link to Computer Chess. 129.210.217.63 (talk) 22:16, 3 February 2009 (UTC)[reply]

NOTE: Chess Computer was changed to redirect to Computer Chess --Guy Macon (talk) 19:51, 30 March 2012 (UTC)[reply]

Reads well

The article is pretty good in my opinion, thought its coverage just right...--Billymac00 (talk) 04:19, 11 May 2009 (UTC)[reply]

Weid doesn't have any downloads

Weid chess engine doesn't have any engine downloads. Just an opening book that's all... So is that really an engine if there is no engine? :) 72.245.213.213 (talk) 20:45, 2 August 2010 (UTC)[reply]

I'm not familiar with Weid, but if it is as you say, then it is not an engine. Bubba73 ^{(You talkin' to me?)}, 21:57, 2 August 2010 (UTC)[reply]

I've seen it play in a few tournaments years ago and the author posts on talkchess.com from time to time. —Preceding unsigned comment added by 41.3.90.237 (talk) 18:18, 15 December 2010 (UTC)[reply]

Ratings table

Could anyone tech savvy enough center the text in most of the columns? I think it will look much better. Especially the column for Engine/platform entries. Year started and Rating could also use centering. Nghtwlkr (talk) 09:41, 24 May 2011 (UTC)[reply]

I centered the text in the ratings table. The other tables look best left-justified. --Guy Macon (talk) 20:06, 30 March 2012 (UTC)[reply]

revamping the lead paragraph

Ok, right now the grammar in the lead paragraph is so horrible I can barely make heads or tails of the thing nor do I find the content informative in the least regard. I'm thinking we can just restart the thing fresh from scratch again. How about something more like: “A chess engine, in computer architecture, is the core and most important chess component distinct from all the other components, such as the user interface and move history, encompassing both hardware and software capable of searching and/or analyzing chess positions striving to find the best move for a particular side in a given situation.”? I know, its one massive run-on but I think it does the article justice -- in contrast with what we have now. Any other options? --Thelazyleo (talk) 10:18, 2 March 2012 (UTC)[reply]

I just rewrote the lead paragraph. Please check it and make further improvement if you can. Thanks! --Guy Macon (talk) 20:33, 30 March 2012 (UTC)[reply]

x64 in Rating list table

Is the x64 really needed? Each and every entry has a x64 at the end - it is standard nowadays! It might be better to enter 32b in case a 32 bit engine apears in the top 3. — Preceding unsigned comment added by 80.171.10.41 (talk) 22:26, 30 December 2011 (UTC)[reply]

OK, noone is reacting. As EVERY entry has "x64" as a supplement I propose again to remove that as it is completly redundant nowadays! — Preceding unsigned comment added by 213.39.170.129 (talk) 19:36, 28 January 2012 (UTC)[reply]

I changed it and removed the 'x64'! It is simply standard and doesnt need to be mentioned at all. In case a 32bit engines will be in the TOP3 somewhere it can be refered to as w32/b32 or something similar! — Preceding unsigned comment added by 80.171.41.233 (talk) 20:01, 4 May 2012 (UTC)[reply]

I introduced a question about x64 and waited for more than 3 monthes as I consider it as irrelevant as EVERY engine is x64. Nghtwlkr introduced it again with the remark "it is important". Much more important woudl be that every engine is a windows engine but this is not mentioned as everyone expects this. No he changed it back to x64 again without any reasoning. I will remove it again as I give a reasoning. (And I repaet, 32 bit is irrelevant and becomes more irrelevant every day!) — Preceding unsigned comment added by 80.171.41.131 (talk) 17:26, 22 May 2012 (UTC)[reply]

Sorry I missed your original post. There is no law of nature that says the top-rated programs all have to be x64, even though they currently are. The author of, say, Redqueen could come up with a clever improvement next week that catapults the strength of his 32-bit chess engine to the top of all the ratings lists. Or it could happen that in a few years Intel will bring out a 128-bit CPU and Stockfish will be immediately recompiled for it, gaining strength and putting it at the top of the list. If you knew for a fact that all future top chess engines will be x64 forever, I would agree with the deletion, but you don't know that. --Guy Macon (talk) 17:49, 22 May 2012 (UTC)[reply]

Sure, there is no law for that, but there is the general rule of avoiding redundances! Right now ALL Engines ARE 64bit and that is well know. In case a (one) 32bit engine or in the future a (one) 128bit engine might come up one can name that in the beginning and when it becomes standard it can be removed again. I will remove the X64 again! — Preceding unsigned comment added by 213.209.110.45 (talk) 12:36, 25 May 2012 (UTC)[reply]

I just reverted your attempt to shove your version down our throats. Please read WP:CONSENSUS. --Guy Macon (talk) 13:15, 25 May 2012 (UTC)[reply]

Impractical tournament volumes?

In fact, the number of games that need to be played between fairly evenly matched engines, in order to achieve significance, runs into the thousands and is, therefore, impractical within the framework of a tournament.

The link supporting that statement is broken. More to the point, why would tournaments with thousands of games be impractical? Computer processing is fast, so that a reasonable tournament could have short time controls. The same two engines can play as many simulated games as you have CPUs, right? And they can certainly play one game after the next without ever resting.

I'm not an engine developer so I don't know that anything I just wrote is accurate, but it should be clear that this point is non-trivial and therefore that the article needs to explain why thousands of games per tournament is impractical. - Thanks, PhilipR (talk) 00:37, 25 October 2012 (UTC)[reply]

The problem is in your assumption: "Computer processing is fast, so that a reasonable tournament could have short time controls." yes, Computer processing is fast, but chess engines are designed to use that processing power to best advantage at lightning chess to tournament time controls. A chess engine could be designed to play best at thousands of moves per minute, but real programs are optimized to use the time they have well.

The reference, which you can find at

http://web.archive.org/web/20110714104652/http://www.mizarchessengine.com/columns/testing-a-chess-engine/

(I will update the page if the server doesn't come up soon)

says "So after 4500 games we can say that A is better than B with 95% of confidence". Even if you play speed chess at five minutes per side, that's over a month to complete one pairing, and about a year for a 10-engine round robin, assuming you own ten fast computers. By which time at least three of the engines will have a new version. If you only have one computer, that's ten years. --Guy Macon (talk) 01:33, 25 October 2012 (UTC)[reply]

Thanks for the info. I would actually not be surprised if one person or group of people (1) found it sufficiently important to know which engine performs best (2) had that many computers to tie up for that long. This is even truer if we relax the assumption of fast computers -- tech geeks often have a bunch of outdated ones that still have reasonable processing power.

But I'm also not understanding your assumptions. As you stated, I calculate that one match of 4500 games would take 31.25 days. I don't understand why a 10-engine round robin would take more than 5 * 9 times that, or 1406 machine*days. Five computers can host ten engines, and each engine would play the other nine.

But if you need 4500 for a two-engine match, you certainly don't need a round-robin of 4500 games each to declare a winner with 95% accuracy. For example, wasting time on the #1 engine versus the #10 engine adds no predictive value to identifying the best of the bunch, once you've identified them as "pretty good" and "relatively bad", respectively, with some confidence. Intuitively I'd guess this scales much better than you think, like maybe with the square root of number of engines, so sqrt(10)*4500 = 14230 games total might be enough. But that's just an intuitive guess. The point is, it's a lot less than 45000.

It's an interesting theoretical problem to envision how to design a tournament in minimum number of games to declare a winner with 95% confidence. Surely you'd want to run a bunch of round robins in sequence (rather than pairing up the same two for 4500 games) and then use your growing evidence to start eliminating contenders.

Finally, I get that speed chess conditions aren't optimized to display the true power of each engine, but they seem a lot more satisfying than sticking everyone in a tiny-sample tournament declaring a champion. -- Regards, PhilipR (talk) 04:44, 25 October 2012 (UTC)[reply]

Looking over your comments, I believe that you are correct and that I overestimated the number of round robins needed. I am going to run this past the fine folks at Wikiproject Statistics to see if someone has analyzed round robins and published. --Guy Macon (talk) 01:04, 26 October 2012 (UTC)[reply]

Guy, that reference is a blog/self-published which does not qualify as a reliable source for the purposes of Wikipedia. Quale (talk) 05:22, 25 October 2012 (UTC)[reply]

That reference (one URL change, same reference) has been in the article since 14 November 2007[1] with no objections and was written by Nicola Rizzuti, author of the Mizar chess engine, so it could be argued that he fits the description (WP:SPS) of an established expert on the topic of chess engines. That being said, the basic math behind sample size and confidence levels is well documented in the statistics literature, so if you seriously doubt Rizzuti's conclusions, I can dig up a cite from the academic literature that says pretty much the same thing -- it takes a lot of samples to be sure with 95% confidence which of two closely ranked chess engines is really better. Or I can bring it up at WP:RSN and get a second opinion about reliability, but it would be best to see if the site being down is temporary before doing that. --Guy Macon (talk) 07:19, 25 October 2012 (UTC)[reply]

I actually think the statistics are correct, but they aren't particular to computer chess. The same situation arises in human chess, or in fact any other two-player competition. In fact, it's essentially the same as Checking whether a coin is fair. As you say, there are sources unconnected to computer chess that will support the claim if computer chess sources aren't available. The author of a chess engine might be considered a subject matter expert, but it isn't apparent that writing a chess engine makes one an expert in statistics. All that said, I'm not especially concerned, but I did want to point out that this is a self-published source. Quale (talk) 23:53, 25 October 2012 (UTC)[reply]

Point well taken about expertise in chess engines and statistics. --Guy Macon (talk) 01:04, 26 October 2012 (UTC)[reply]

Private

What are the criteria for inclusion in the list of "private" engines? Should this list even be in the article? - dcljr (talk) 23:55, 28 January 2013 (UTC)[reply]

I don't think it belongs. I wrote more at Wikipedia:External links/Noticeboard#Chess engine, but that discussion hasn't garnered much participation yet. Quale (talk) 02:52, 29 January 2013 (UTC)[reply]

I have deleted that entire section per discussion at the external links noticeboard. ~~Of course the question of criteria for classifying chess engines still exists -- it has just moved to list of chess engines.~~^{Oops. Didn't follow the link. It redirects here. Perhaps we should talk about having a "List of chess engines" page.} --Guy Macon (talk) 20:12, 29 January 2013 (UTC)[reply]

Chess engine torture tests

If I can find enough well-sourced material, I would like to add a section to the article about test positions that are constructed to test / confuse chess engines. I found a couple at
http://blog.chess.com/zaifrun/cool-endgame-study---test-your-engines
http://blog.chess.com/zaifrun/test-your-chess-engine
but I have a feeling that there might be a collection of them somewhere. Does anybody know of one? --Guy Macon (talk) 20:16, 29 January 2013 (UTC)[reply]

Is this a WP:LINKFARM?

Please see discussion at Wikipedia:External links/Noticeboard#Chess engine --Guy Macon (talk) 14:01, 27 January 2013 (UTC)[reply]

No response here, so I am implementing the recommendation of the external links noticeboard --Guy Macon (talk) 02:59, 29 January 2013 (UTC)[reply]

What have you done? Before you came along and messed it up the page categorized the different chess engines. Now it doesn't. How is that progress? Where would one look on Wikipedia to see which chess engines fit into which category? Do you expect people to literally search out each chess engine to find out whether it's pedagogical or open source or proprietary? Or do you intend to create separate pages for each of those lists? No? I didn't think so. Why would you remove useful information from a page? It doesn't make any sense. Don't hide behind policies - use some common sense. Either put the page back the way it was or create the necessary separate pages to replace what you've destroyed. Don't just go around wiping out useful information: that is just anathema to the spirit of wikipedia. 'Recommendations of the external links noticeboard'?! If you don't like lists of external links then just internalize them all. Many of those links were already internal. Where they weren't and the page exists in wikipedia, link to it. Where it doesn't, create the page, or don't link it at all and just put a phrase next to the name of the engine with a few pieces of key information. Destroying it without ensuring the info is somewhere else on wikipedia in an easy to access format is both morally reprehensible and lazy. Play well, Alex. Chessbloke (talk) 21:30, 3 February 2013 (UTC)[reply]

I am open to discussing this, which until you posted the above appeared to be uncontested (contested edits should get more scrutiny). This will go a lot smoother if you dial the aggression back a notch and present your case in a calm. logical manner. Before we go any further in this discussion, please read Wikipedia:What Wikipedia is not#Wikipedia is not a mirror or a repository of links, images, or media files and Wikipedia:What Wikipedia is not#Wikipedia is not a directory

Re: "Don't hide behind policies - use some common sense", where do you think those policies came from? They came from the common sense of the Wikipedia community -- a bunch of editors just like us. If you believe that one of the policies does not make sense, you are free to suggest changes to the policies and see if the community agrees. --Guy Macon (talk) 01:35, 4 February 2013 (UTC)[reply]

The possibility of putting the links in a separate "List of chess engines" page was mentioned. I think that could be OK, and if you can be a bit calmer about this you might be able to find something that works. I have to say that I think that this article was improved by removing the linkfarm. Quale (talk) 03:06, 4 February 2013 (UTC)[reply]

I have no problem with creating a "List of chess engines" page and putting the deleted material there. The only question is whether anyone objects to that, in which case we need to discuss it further.

If you look at these two links Link 1 Link 2 you will see that on 00:47, 17 October 2006 (UTC) this page was renamed / moved from "List of chess engines" to "Chess engine" with the edit comment "More than a list" (that page was created on 15-16 March 2004: Link 3) --Guy Macon (talk) 09:49, 4 February 2013 (UTC)[reply]

Like Quale and Guy Macon I think the article is greatly improved by removing the link farm. As it stood it was something more suited to Wikia than an encyclopaedia like Wikipedia. A move to improve is in no way morally reprehensible and lazy. --Biker Biker (talk) 22:25, 4 February 2013 (UTC)[reply]

Not aggression; righteous indignation. Here's my case: the information was valuable, instead of putting in the effort of converting the information into a format compatible with the policies of wikipedia you simply deleted it. That is either laziness or vandalism, both of which are morally reprehensible. The policy is fine and sensible, my criticism was of your use of the policy to justify lazy editing a.k.a. deleting things which don't comply with policies instead of converting the things into a format which is suitable for an encyclopedia. If you lack the skill (as I do) or or the energy to improve this page rather than delete bits of it then do nothing to it. Play well, Alex. Chessbloke (talk) 19:31, 5 February 2013 (UTC)[reply]

I placed a standard warning on the Chessbloke's talk page[2] using Template:Uw-npa2. I encourage him to review WP:NPA, WP:CIVIL, WP:OWN and WP:CONSENSUS before responding. --Guy Macon (talk) 19:57, 5 February 2013 (UTC)[reply]

I think it would make more sense to create a new page for listing the chess ratings and dump the "linkfarm" there where people would be more interested in that sort of thing. --Thelazyleo (talk) 08:02, 6 February 2013 (UTC)[reply]

Or we could put them in List of chess software. That article would be improved if we split the list of software into engines and engine-plus-GUI (excluding engines that just tack on XBoard or WinBoard). --Guy Macon (talk) 08:21, 6 February 2013 (UTC)[reply]

I'm in favor of dumping it there and revamping that article with the added data. --Thelazyleo (talk) 08:27, 6 February 2013 (UTC)[reply]

Any objections? If not, Chessbloke, would you like to take the lead on improving List of chess software as discussed above? --Guy Macon (talk) 11:46, 6 February 2013 (UTC)[reply]

There are more kinds of chess software than just engines and GUIs. See Category:Chess software for some examples. This of course is fixable by adding more sections, but section on engines will dwarf the others. Quale (talk) 20:01, 6 February 2013 (UTC)[reply]

Chessbloke has indicated elsewhere that he will not be updating List of chess software as discussed here. I can do it, but I am a bit swamped on a hot project so I won't be able to get to it right away. If anyone is interested, please jump in. It's certainly worth having such a list. --Guy Macon (talk) 15:28, 7 February 2013 (UTC)[reply]

Hash Size?

What about hash sizes for engines, doesn't that also determine the playing strength of the engine by changing the speed it calculates? Houdini 3 came with a manual that states a formula to calculate the ideal hash size given the average search speed of the engine on one's hardware and the average time per move but my own testing has showed Houdini gets faster node searches with basically no hash. Is this a situation like Houdart described for hyperthreading, where hyperthreading increases the search speed but makes the whole process less efficient therefore is not recommended? 108.223.240.62 (talk) 22:33, 29 August 2013 (UTC)Dark_wizzie[reply]

It's actually a lot more complex than that. See [ http://rjlipton.wordpress.com/2012/05/04/digital-butterflies-and-prgs/ ]. --Guy Macon (talk) 01:21, 30 August 2013 (UTC)[reply]

takeback

Can a person beat a program if the person has no time controls and can use other programs and can take back as many moves as they want? I came here hoping to find answer — Preceding unsigned comment added by 47.72.78.79 (talk) 07:05, 22 October 2017 (UTC)[reply]

Alpha Zero

I'm new to wikipedia so forgive me. But shouldn't the AI engine that can consistently BEAT Stockfish 9 be mentioned? Should an AI Engine section be added? What about Leela AI? — Preceding unsigned comment added by JBunco (talk • contribs) 05:17, 26 July 2018 (UTC)[reply]

AlphaZero would probably be fine for the "dedicated hardware" section, or a new "AI engine" section; feel free to add either. But (a) the latest information I can find says that it was tested against Stockfish 8, not 9, and (b) since Stockfish was on significantly inferior hardware compared to AlphaZero - and the hardware wasn't even optimized for Stockfish's approach whereas AlphaZero's was - saying that it can "consistently beat Stockfish" is a bit unfair. Jtrevor99 (talk) 15:01, 26 July 2018 (UTC)[reply]

cheating

Section needed on use of chess engine in cheating 217.155.25.156 (talk) 16:43, 22 September 2022 (UTC)[reply]