Champions Showdown: Improving the format

Advertising

Books, boards, sets: Chess Niggemann

Macauley Peterson Macauley served as the Editor in Chief of ChessBase News from July 2017 to March 2020. He is the producer of The Full English Breakfast chess podcast, and was an Associate Producer of the 2016 feature documentary, Magnus.

Discuss

Rules for reader comments

Petrarlsen 10/14/2018 06:08

@ celeje:

"No, I'm not consciously or subconsciously wanting everything to be delayed many years for no reason."

The fact is nonetheless that you still didn't answer the question I already asked you two times : "How would you explain that Chess960 - which hasn't even existed for a single quarter of a century - is quite generally considered by its proponents to be fixed for good in its smallest details, while traditional chess evolved through many centuries before reaching its modern form?"

celeje 10/14/2018 05:53

@ Petrarlsen:

celeje: "The best evidence is from professional competitive tournament games by the best players."

Petrarlsen: "How would it be better than a computer playing against himself, if we agree that computers play Chess960 at a better level that the best human players? All the more if, as lajosarpad suggested, you add "human insight" by having GMs participating to the evaluation. "

I think the computer testing is a good idea. I was going to make a separate comment & still will. BUT I'm not sure it is it the best way on its own. lajosarpad is worried that chess engines have their inbuilt traditional-chess biases and they will therefore play advanced but flawed chess960, which will skew the results. If that's true, that's possibly a serious problem. The other problem is that right now, if you just have computers playing lots and lots of traditional-chess games to see what happens with traditional chess, they will not reproduce all the current human opening theory. They'll play the same narrow openings over and over again. You can add a randomness factor into their opening moves, so that their openings are more varied. That's better, but they still will not reproduce humans' biggest and best opening books. Then the chess engines will be testing not really the chess960 initial position but just the small opening repertoire they like in that chess960 initial position.

I'm less worried than lajosarpad about that, i.e. that the engine's may have untrustworthy chess960 evaluation, but it should be thought about.
e.g. lajosarpad: "I gave you an argument that the engines are most probably having legacy logic and I gave you the doubt of all the starting positions being thoroughly tested. I find it strange to doubt these highly probable issues, since this should be the null hypothesis".

Your "if we agree that computers play Chess960 at a better level that the best human players" already has the reply below (celeje 10/13/2018 01:27 and lajosarpad 10/13/2018 05:02), that they are better because they are crunching brute force through their calculations to high depth as they always do, while the humans are forced to calculate and avoid tactical blunders from move 1, which they don't cope well with. Again, we're back to lajosarpad's worry about chess engines.

Petrarlsen: "I had already noticed that you are not theoretically opposed to some changes being implemented in Chess960, but only if these changes could only occur in many years..."

No, I'm not consciously or subconsciously wanting everything to be delayed many years for no reason. I'll explain in a new comment.

Petrarlsen 10/14/2018 02:33

@ celeje:

- "I think the working hypothesis should be that all positions are fair in the sense of being draws with perfect play."

In my opinion, this isn't at all the problem. The real problem is to know if one player begins the game with much more chances to win it. Players are human; perfect play isn't possible - what is really the interest to know that, with perfect play, it would be a draw? For example, very frequently, in Carlsen's endings, with perfect play, the game would end with a draw, but Carlsen senses very well these positions in which he has a PRACTICAL advantage - I remember one of these positions against So, last year if I remember well; the annotating GM explained that, theoretically, this would normally have been a draw, but that, in practice, it was nearly impossible to defend for So.

So, if, in some Chess960's starting positions, one of the players has a really significant advantage, these positions are not fair, even if they are draws with perfect play - "we are not discussing God playing chess against himself" but human play; I don't see at all the interest to know that a position would theoretically be a draw with perfect play if it is much more difficult to play for one (human) player than for the other. Even between computers, this wouldn't be really interesting (computers are - for the moment, at least... - far from playing perfect chess, apart from the tablebases' positions), and, between humans, this is even worse...

- "The best evidence is from professional competitive tournament games by the best players."

How would it be better than a computer playing against himself, if we agree that computers play Chess960 at a better level that the best human players? All the more if, as lajosarpad suggested, you add "human insight" by having GMs participating to the evaluation.

My opinion is that the real explanation for which you consider that the only means to evaluate Chess960's starting positions is to have tournaments between human players is that, very probably unconsciously, you have some sort of an "internal dogma" according to which Chess960 musn't be modified, and all the 960 starting positions must be played - if we would consider that the best evaluation would be to have tournaments between human players, this would have two great benefits, in your persepective: 1) that a possible elimination of some starting positions would only occur in many years, so the problem would be put off until a much later date; 2) that in practice, Chess960 would be played as it is, criticized or not criticized, considered as imperfect or considered as perfect... it would be played, and this is in my opinion the only thing that really counts for you (even if, perhaps, you don't realize this fully yourself). I had already noticed that you are not theoretically opposed to some changes being implemented in Chess960, but only if these changes could only occur in many years... so that the possibility of these changes is in fact only theoretical, and that, in practice, Chess960 would continue to be played exactly as it is, in the slightest details...

celeje 10/14/2018 12:38

@genem:

genem: "Yes, the basic idea expressed in the quotation is the future of chess960.
Discard the 'Random' from Fischer Random Chess!"

genem, if you discard the "random", then the players will be forced to do full-time opening preparation taking up most of their waking hours to play chess960. How many superGMs and GMs do you think are willing to do that?
If they are not willing, they won't play. If they don't play, chess960 won't grow.

celeje 10/14/2018 12:33

@lajosarpad (@lajosarpad's most recent comment below):

That's not where I think our most important difference of opinion lies. I tried to explain before but was too brief. I think it's interesting to think about engine proficiency in chess960 vs traditional chess, but it's not required right now to "validate" chess960, unless we think engine confirmation of acceptable fairness for the 960 positions is required now. Because...

lajosarpad: "But I prefer playing chess over Poker. No offence to Poker players. "

... Even in the worst case, it's not like poker. The players play two games with the same position, colors reversed. That's what they currently do in all these matches. Your arguments for needing to know for sure how fair/unfair positions are are just an argument for needing to continue doing what they are already doing (playing pairs of games).

(I know you mentioned single casual friendly games. I'll get back to that in another comment. Petrarlsen is also not satisfied in a philosophical way by the play-two-games solution. I'll get to back to that in another comment too.)

I think the more important difference of opinion between you and me is in what you think "should be the null hypothesis", not for the chess engine question, but for the chess960 starting positions. I think the working hypothesis should be that all positions are fair in the sense of being draws with perfect play. I think that is the most likely outcome independently of what the engines would say. If I understand you correctly, you do not think this should be the chess960 "null hypothesis".

Humans+computers will never prove or disprove it even for traditional chess. All we can do is get the best evidence we can for this. The best evidence is from professional competitive tournament games by the best players. (You and I seemed to agree on this last sentence before, but maybe we need to rethink it... I'll put this in another comment.)

lajosarpad 10/13/2018 05:02

@celeje

I gave you an argument that the engines are most probably having legacy logic and I gave you the doubt of all the starting positions being thoroughly tested. I find it strange to doubt these highly probable issues, since this should be the null hypothesis, until we see that such engines appear and scientific papers are published. Did we hear of an engine which was planned, implemented and tested with the specific purpose of supporting chess960? I did not hear about any such engines, but if you did, then please, let us know which are these engines. Chessbase did not provide us even the UI for chess960, so I find it strange to believe (because that's what it is: a faith) that the engines are already satisfyingly covering all the nuances chess960 has and chess does not have in all the possible positions, nonetheless. As about tests, it is virtually impossible that all the 960 positions were thogoughly tested. If you have any factual knowledge that all these were covered and the methodology used, then I await your arguments, of course. Grandmasters did not play enough games of chess960 to even consider human insight into the results of the test. And what are the tests compared against? Of course if there is legacy code, affecting the correctness of evaluation, it is far from trivial, but once somebody finds it, the position will become unplayable at the top. Petrarlsen argued that somehow these possible findings should be managed. He did not limit this to balance, but added philosophical aspects as important elements of analysis. I quite agree with him about this.

"But why would anyone expect such a bias to be not just significant but biasing the eval. differently for different positions, so that the differences in the evaluations for the 960 positions are significantly suppressed? It may be. It may not be. "

We did not exclude that possibility and your conclusion is true. But I prefer playing chess over Poker. No offence to Poker players.

"I don't think we need to prove engines can evaluate all starting positions accurately and then use them to prove that all starting positions are not so different in advantage for one side, before we can seriously play the game. I think that's the difference in our opinions on this. "

The game can be seriously played before all these guarantees are provided. For instance, Poker is a very successful game and many players play it seriously, it is fun and has a lot of bluffing, which increases the drama.

In your comment you accurately predicted my argument. Indeed, humans do not have theoretical knowledge about most of the positions in chess960 (this is the precise cause the variant was suggested in the first place) and therefore they are more likely to play the opening inaccurately, or, if they are accurate, they will use up a lot of time reducing their chances against an engine.

@Petrarlsen

I agree with your points, but would add that human insight is needed as well, if the 960 positions would be divided between 60 very strong players, ideally GMs to analyze, then they would have 16 positions each. If each position would be rigourously analyzed by experts, then we would have human insight. If FIDE or a sponsor would pay for such a project, then at the end we would have a knowledge base. Currently, due to financial reasons, human insight is excluded from the evaluation of the 960 positions.

celeje 10/13/2018 01:27

@ Petrarlsen & lajosarpad:

Yes, I think it's definitely true that chess engines will gain in relative strength in chess960. That's because the humans have to calculate & get all the tactics right from the start, while the engines brute force calculate to high depth as they always do.

But I think lajosarpad would say that doesn't mean chess engines are stronger in chess960 than they are in traditional chess (or equal in both), but just that humans are much worse in chess960 than they are in traditional chess.

genem 10/13/2018 12:52

Quote from the article:
{
One proposed compromise solution is to have an "official Chess960 position" selected each year by, for instance, FIDE, that would be used in tournaments for the following calendar year.
}

Yes, the basic idea expressed in the quotation is the future of chess960.
Discard the 'Random' from Fischer Random Chess!

Petrarlsen 10/13/2018 11:31

@ lajosarpad and celeje:

In the previous article about Chess960 (https://en.chessbase.com/post/the-problem-with-chess960), in the section "Why does anyone need this new chess variant?", Frederic Friedel explains that the difference in strength between the best human players and the computer programs is even greater in Chess960 than in traditional chess.

I would think that this would mean that the present-day programs should be sufficient to evaluate the difference between White and Black for the Chess960's starting positions. All the more since it is possible to reevaluate regularly these positions, to follow the progress of the Chess960's programs...

Petrarlsen 10/13/2018 10:28

@ lajosarpad and celeje:

I will try to write some posts about castling tomorrow.

About the "Zero sum game" aspect:

Thinking of it after reading lajosarpad's posts, I think that this indeed something to take into account.

Initially, my reasoning was that, be it in traditional chess or with any Chess960 starting position, the situation is the same: the two players begin with the same starting position, and the difference between the two sides is only that one player plays first. So I considered that, in a way, the difference between the situations of the two players is the same: a one-move advantage for White (taking the "number of moves in advance" unit to measure the advantage of the White player, it is the same for every position: one unit, i.e. one move...).

But it is true also that, if, in practice, some positions would give a significant advantage to one player, it would be a problem. And I don't agree with celeje that a +0.6 advantage is a trifle; rounding off the evaluation to the nearest pawn unit, the difference between 0.2 and 0.6 is that 0.2 is nearer to 0 and 0.6 nearer to 1; in my opinion, above 0.5, it is too much: rounded off, it amounts to a 1 pawn deficit and this is a significant difference in my opinion.

When celeje writes: "It won't make sense for them (note: the SuperGMs) to complain (and I've never heard any of them complain) that it MIGHT be a starting position that is +0.6 instead of +0.2, when the moves of the two opponents swing it from +0.6 to -0.9 to +0.8 to -1.1 in the opening.", I don't agree at all. There is absolutely NO reason to think that some sort of an equilibrium around 0.0 will arise spontaneously. So what would have been a fluctuation between +1.0 and -1.0 with a 0.0 starting position would become a fluctuation between +1.2 and -0.8 for a starting position with a 0.2 difference between White and Black - more or less similar to the difference in traditional chess' starting position according to the programs (which I find acceptable - nearer to 0 pawn than to 1 pawn). But, with a 0.6 difference between White and Black, this fluctuation between +1.0 and -1.0 would become a fluctuation between +1.6 and -0.4 - and it is obvious that this would favor very much the White player: even at the worst moment, the White player would have only a 0.4 handicap (nothing much), while at the best moment, a +1.6 advantage is beginning to seriously "smell victory" for White; nothing final, but Black will have to seriously work to save a draw, and can probably more or less forget any dreams of winning the game (at least with classical time controls).

To play two games back to back with reverse colors is a solution as for the sports dimension of chess (the result will be more or less fair...), but not for the "work of art" dimension of a game of chess: if a player begins a game with a significant advantage, it can mar this game as a "work of art": if the White player wins, is it because he played superior chess? or essentially because of his initial advantage?

I think that, probably, the best solution would be to use an engine (or even, if possible, several engines) to evaluate the Chess960's starting positions (if or when a sufficiently efficient engine exists or will exist), probably with a mixt of 1) the engine evaluation for each starting position and 2) series of games played by the engine against itself for each starting position. (It seems to stand to reason that such an evaluation will necessarily be more reliable than simply to study the statistics from several tournaments - two players playing one against the other are never of exactly the same level, etc.). And, in my opinion, all the positions with a difference in evaluation of more than 0.5 between the players should be eliminated from rated games (...admitting that a Chess960 rating would be created...). And all the starting positions should be reevaluated regularly, to use the most precise evaluation possible.

celeje 10/13/2018 07:55

@lajosarpad:

lajosarpad: "The engines are not ready to get rid of assuming the starting position of chess if I am not mistaken and the different starting positions are not well tested."

I don't want to guess too much without hearing from an engine programmer. Maybe that wouldn't answer the question either. I don't think a traditional-position bias will be there in an obvious way, or we could look at the source code for it. It hypothetically could be there in a non-obvious way, just in the exact tuning of the weights of all the terms in the eval function, when the tuning is done by games testing, which will all be from the traditional starting position.

But why would anyone expect such a bias to be not just significant but biasing the eval. differently for different positions, so that the differences in the evaluations for the 960 positions are significantly suppressed? It may be. It may not be.

I don't think we need to prove engines can evaluate all starting positions accurately and then use them to prove that all starting positions are not so different in advantage for one side, before we can seriously play the game. I think that's the difference in our opinions on this.

lajosarpad 10/12/2018 04:37

@celeje

If the starting positions are known in advance, then they can prepare for them. If they are unkown, then they can prepare by analyzing the style of the other players and finding out what kinds of positions they like and what kinds of positions they dislike. Chess960 would change the way preparation goes greatly, but would not erase preparation at all.

The starting position of chess is well tested in practice. We do not know whether white wins due to the extra tempo, the game is drawn or black wins due to forcing white into zugzwang if the perfect moves are played. However, we cannot know for sure whether the other games are partially solvable. The engines are not ready to get rid of assuming the starting position of chess if I am not mistaken and the different starting positions are not well tested. I think tournaments having the same chess960 starting positions in all games would help us gathering practical knowledge about how humans fare with the given position and if a color is very much dominating, then investigation for the causes might be needed.

"I don't think that's the right view of how the game would spread. I think the game will spread if the top professionals play it, and their games are advertised, broadcasted, covered, etc. Then people at lower levels will see it and become interested in it.

Its popularity will spread downwards from the top, not upwards from the bottom. "

I partially disagree. The voices of strong amateurs will be heard if there are many, discouraging players to join the game. The solution is to avoid giving reasons for disappointment and here I agree with Petrarlsen, we should not make a religion of chess960, but analyze the positions with rigour. Non chess players will not really watch tournaments and if they do by chance, they will not really understand the game. They need strong amateurs to play with.

Let me give you a hint of legacy logic, by giving you an example:

If a player can still castle, then the possibility raises the evaluation of heuristics, but the actual value of castling is difficult to determine. In many cases the evaluation of the last level of depth depends on positional values and if at the depth of 39, castling was not done yet, then a general concept of castling usefulness has a part in the formula. If that's inaccurate, then the evaluation is inaccurate as well. And we should not forget the horizon effect. An evaluation of today might be changed when one studies the given starting position in depth.

"And we don't know what the White advantage is exactly for the standard position."

We do not even know whether there is an advantage and if there is, whether white has the advantage. Seemingly white should have the advantage if there is an advantage at the starting position, but if a sequence of perfect moves leads to a zugzwang, where white is lost, then black has the advantage.

"They didn't worry centuries ago that this might be e.g. +1.3 and refuse to play the game."

Centuries ago they did not have internet, chess engines, an army of seconds and a plethora of articles.

"We can only start finding out by having many games played by superGMs and GMs. "

Agreed.

celeje 10/12/2018 01:50

@lajosarpad:

lajosarpad: ""Meaningless difference"? How do you know it will not cause a problem? How do you know it will be +0.6, +0.2? All the positions should be evaluated carefully and if some are quite unbalanced, then they should be excluded from potential starting positions. ... As of the engine: chess engines were initially programmed assuming the starting position of classical chess. Are we sure this evaluation is no longer there in the programs? I think a completely rewritten chess960 engine would be needed to make sure no such legacy assumptions remain."

I meant that computer engine evaluations that show +0.6 vs +0.2 at depth 39 are "meaningless differences". I don't think standard chess engines are skewed so they cannot evaluate chess960 positions properly. There's nothing really opening-specific in their evaluation functions.. I cannot remember what the exact range was for the 960 positions. (It can be looked up.) If there was too much White advantage, my guess is it'd be something concrete, e.g. tactically forced, and so computers should find it at high depth. If they're only showing something like +0.6, I don't think it's there. Of course we can't be sure. I don't think we need to be sure. Again, it's standard so far in these matches to play the same position twice, once with White & once with Black.

And we don't know what the White advantage is exactly for the standard position. They didn't worry centuries ago that this might be e.g. +1.3 and refuse to play the game. We can only start finding out by having many games played by superGMs and GMs.

celeje 10/11/2018 04:36

@lajosarpad:

lajosarpad (@ Petrarlsen): "So, while the problem is solvable for tournaments, it is unsolvable for friendly games. And this is a big problem, because friendly games are the hotbed of new players. If the number of hobby players is reduced, then the game will slowly die."

lajosarpad: "... then a lot of us would stop playing and new players will not be introduced through us. If there are lesser and lesser strong hobby players, then the chess culture will be weakened. Newcomers will not think about the evaluation, but very good players will stop playing the game. And their exodus would be followed by less strong hobby players as well. "

I don't think that's the right view of how the game would spread. I think the game will spread if the top professionals play it, and their games are advertised, broadcasted, covered, etc. Then people at lower levels will see it and become interested in it.

Its popularity will spread downwards from the top, not upwards from the bottom.

& at the top, they will do what they think makes it fair. ATM, that means they play matches in pairs of games with reversed colors.

celeje 10/10/2018 09:08

@lajosarpad

lajosarpad: " "Another thing: If they just play one game, whatever the starting position, the White player will have a small advantage anyway, so if these two casual players are so worried, they'll have to play two games anyway. "

Please, show me the perfect chess game, where White and Black are always playing the strongest possible moves and where White beats Black. If you are not able to do that at this point, then your statement above is nothing more than an assumption. "

But that's another big reason why differences in White's opening advantage (White always has a small opening advantage) in different chess960 starting positions don't matter practically. Your point about non-perfect moves is much more true of chess960. SuperGMs will make moves that flip the whole eval. in the opening of chess960, probably at least 2 or 3 times in the opening. It won't make sense for them to complain (and I've never heard any of them complain) that it MIGHT be a starting position that is +0.6 instead of +0.2, when the moves of the two opponents swing it from +0.6 to -0.9 to +0.8 to -1.1 in the opening.

celeje 10/10/2018 01:07

@lajosarpad:

lajosarpad: "Celeje, when I was speaking about tournaments and matches I was of course referring to serious tournaments and matches, of classical time control"

Ah, okay, I understand you. I agree that classical time controls would be better. (But not "with the players preparing for many days beforehand". How would they prepare differently from now?)

That's one problem with having two games with same position, colors reversed. Obviously they should be on the same day, which means each game cannot be at long time controls, which could take too much time.

lajosarpad 10/10/2018 11:35

@celeje

Celeje, when I was speaking about tournaments and matches I was of course referring to serious tournaments and matches, of classical time control and with the players preparing for many days beforehand and which are publicized in a way that the whole world follows them. At this point the world audience does not see the games, does not understand the rules, so we are in prehistoric times, considering chess960. The large audience at this point is not able to follow the games live, the softwares are not ready, the news articles lack methodology. Suggestions are only possible if one actually sees the game and in order to make people watch the games, they should be highly publicized, I think we can agree on that. Nowadays, non-chess players in most cases do not even know chess960 exists. In many cases amateours, or weaker players do not know about it either. It is just not an important part of the chess life yet.

lajosarpad 10/10/2018 11:35

@celeje

I mean being personally present, or watching live games. If there are 30 games I am interested in, it will be difficult to remember where they have already castled and where not, especially if they are blitz games. This is difficult, even if you see the moves they already made. And I prefer not to have to watch the moves they have already played for almost all games. Somehow it should be evident whether they had castled.

Friendly games can be played between strong players as well. But regardless of the actual strength of the players, the loser might say that he or she did not lose because the other player played better moves, but because of the disadvantage at the start of the position. This would start never ending discussions. For instance, I frequently play friendly games and both me and my opponents play quite decently. If the game would stop to be fair, then a lot of us would stop playing and new players will not be introduced through us. If there are lesser and lesser strong hobby players, then the chess culture will be weakened. Newcomers will not think about the evaluation, but very good players will stop playing the game. And their exodus would be followed by less strong hobby players as well.

The person of the opponent is a matter of luck, but there is still an objective game where, with excellent performance even the best or even the computer can be beaten. I might play Carlsen, then Caruana and then Karjakin, who are all very very strong and if I get 0 points, then it is because I could not perform well. But if the objective evaluation of the position favors me or them greatly and a line is known to get the advantage to someone, which is certainly possible with the current approach of "let's believe everything is all right", then the game is no longer fair. So let's make a clear distinction between "luck" in terms of the opponent's potential performance against the individual, or "luck" in terms of getting an unbalanced starting position.

"Meaningless difference"? How do you know it will not cause a problem? How do you know it will be +0.6, +0.2? All the positions should be evaluated carefully and if some are quite unbalanced, then they should be excluded from potential starting positions. And the evaluation needs to include human grandmasters, who devote a significant amount of time for a given position, maybe a month. As of the engine: chess engines were initially programmed assuming the starting position of classical chess. Are we sure this evaluation is no longer there in the programs? I think a completely rewritten chess960 engine would be needed to make sure no such legacy assumptions remain.

"Another thing: If they just play one game, whatever the starting position, the White player will have a small advantage anyway, so if these two casual players are so worried, they'll have to play two games anyway. "

Please, show me the perfect chess game, where White and Black are always playing the strongest possible moves and where White beats Black. If you are not able to do that at this point, then your statement above is nothing more than an assumption.

celeje 10/10/2018 10:27

[continued]

At Mainz alone 15 matches & tournaments...

Chess960 Rapid Chess World Championship
2001 Peter Leko, 2003 Peter Svidler, 2004 Peter Svidler, 2005 Peter Svidler,
2006 Levon Aronian, 2007 Levon Aronian, 2009 Hikaru Nakamura

Open Chess960
2002 Peter Svidler, 2003 Levon Aronian, 2004 Zoltan Almasi, 2005 Levon Aronian, 2006 Etienne Bacrot, 2007 Victor Bologan, 2008 Hikaru Nakamura,
2009 Alexander Grischuk

celeje 10/10/2018 08:12

@lajosarpad:

lajosarpad: "Anyway, I would be VERY surprised if after the first 15 tournaments of chess960 we would not have a very very long list of sensible suggestions of improvement."

Without looking it up, I'm pretty sure there have been more than 15 tournaments+matches already. In past years, before the Carlsen-Nakamura unofficial championship match, there were matches with Svidler, Aronian, Leko, etc. and open tournaments.

I think they did/do/can show sensible suggestions for improvement of the PLAYING CONDITIONS (match or tournament rules), not the rules of Chess960 themselves.

celeje 10/10/2018 01:42

@ lajosarpad:

On fairness of opening advantage of different starting positions.

lajosarpad (@ Petrarlsen): "... if we play a friendly game and we randomize the position and you have a disadvantage, then we will not necessarily play another game which would neutralize your disadvantage. ... this is a big problem, because friendly games are the hotbed of new players. If the number of hobby players is reduced, then the game will slowly die."

Do you really think in a friendly game that means nothing players will worry about the opening advantage being e.g. +0.6 according to depth 39 calculations of some program rather than +0.2 for the standard opening position? I don't think they will think about that at all.

At a higher level, open tournaments, this is no problem at all for this reason:

in open tournaments, you're not playing the same opponents and definitely not the same opponents with the same color assignments. Who the leaders face is a huge amount of luck. That is FAR MORE "unfair" than the POSSIBLE but UNKNOWN slight theoretical difference in white advantage for a different starting position.

(I said this in the comments to the first chess960 article.)

Let me return to your casual friendly game example. If they are so worried by a meaningless difference of +0.6 - +0.2, can't they just agree at the start to play twice as fast and fit in two games with the same starting position and colors reversed?
Another thing: If they just play one game, whatever the starting position, the White player will have a small advantage anyway, so if these two casual players are so worried, they'll have to play two games anyway.

celeje 10/10/2018 12:45

@lajosarpad:

lajosarpad: "I have a problem with the possibility that I join a game as a spectator, where the king is between the two rooks in the first or eight's row, respectively. In that case I will not know whether the king already castled or not, for instance."

I guess you mean being a spectator in person at a tournament. To me, this is just a small practical problem. Online coverage of course will let you see the whole game score so far, so you know the full game state. The small problem is with tournament conditions, etc., not with the game itself. If players have to keep score, you could see, but only if they write clearly enough and large enough. As a spectator, another thing you'd definitely like to know is the starting position. But that is another feature of chess960, it looks like: once you get into the middlegame, you cannot tell what opening position it came from.

lajosarpad 10/9/2018 02:23

@celeje

while I agree with Petrarlsen that the starting position should be logical and philosophically defendable, but I think keeping chess as close to a zero sum game as possible is the most important factor. I understand your point that currently it seems that the positions are fair, but I think we have no reasons to be convinced at this moment. This seems to be a very interesting field of study, where, with very simple methodology important questions can be answered. If a university would start the research, I think in one or two years we will have a lot of important information. The philosophical aspect gives a feel of aesthetics, at least to me. To be fair, fairness is a particular branch of the question tree regarding philosophical motifs and I think this is the most important. While we can argue whether castling has a problem in chess960 and I certainly think it has a problem, the game is playable. If, in some positions castling is possible from the first move and it is better than other moves, then the games will be repetitive and not castling at the first move in that position would be a gamble. Also, if Petrarlsen is right with his opinion that in general the king will be weaker if it is closer to the centre, which - according to him - would defeat the purpose of castling (and I tend to agree, but am not convinced), then these problems will be discovered in tournaments and they could become game changers. This is absolutely logical, but if the game stays fair, it might survive even these changes. I have a problem with the possibility that I join a game as a spectator, where the king is between the two rooks in the first or eight's row, respectively. In that case I will not know whether the king already castled or not, for instance. It is true that the same problem exists in chess, but it is far less serious than in chess960. I think this would greatly diminish the entertainment quality. Also, I believe that in the case that a move is known that it will virtually never be played, then the move is most probably unnecessary. However, again, from all these problems I find the problem of fairness to be the most serious.

My estimations were given defining the time I would need to implement these, but I do not know how many issues Chessbase has to solve, what is their priority list, what is the development model they use, but anyway, the fact that since the last time this tournament was organized a year has passed and chess960 is still not supported even minimally makes me think that something might be wrong in their plans. But this is only my humble opinion. I would not be a regular visitor of this site if I did not consider it to be great, but, as everything, the site has its faults as well.

@Petrarlsen

Thank you, I am interested to see your arguments as well. I find them to be instructive even when I disagree.

celeje 10/9/2018 11:15

@lajosarpad (& Petrarlsen):

lajosarpad (@ Petrarlsen): "About handling the starting position as the only possible starting position when we validate it my point was that since there are more possible starting positions, the possibility of getting a more satisfactory position might be less problematic, just like in Poker, when you get bad cards, you are unlucky, but it is not unjust if the randomization was correct. However, randomity as a possible deciding factor would decrease the value of the game in my opinion. Inferior randomized positions could be remedied by playing a similarly inferior position, where the imbalance favors the player who was unfavored by the imbalance...."

If I understand what you write correctly, you are talking about something different from Petrarlsen. Petrarlsen is judging a starting position "validated" based on philosophical or conceptual ideas (mainly about the castling). You, I think, were thinking about fairness, i.e. the starting position not favoring one side too much. (The rest of your paragraph not quoted above discusses that further.)

I don't think possible different advantages (e.g. for White) for different starting positions is a problem. Someone showed supercomputer analysis of the 960 positions to high depth, and the differences in eval were small. For the Carlsen-Nakamura & St. Louis matches, they just played the same position with colors reversed, anyway. There are other ways of making things equal too, but even if you do nothing at all I don't think right now it would make a difference in a match.

celeje 10/9/2018 08:35

P.S. on Chessbase tech support comment below.:

Can chess960 now be played on playchess.com? If yes, how is it? How many people play there?

celeje 10/9/2018 02:37

@lajosarpad:

lajosarpad: "The fact that Chess960 games are not as supported is a technical question. In order to do so, they need to implement the rule checker (maybe 6 hours of development), they need to accomodate the UI (this might take 32 -40 hours of development). I think this could be done in a week or so. Of course, they need to adapt the database on the long run, which is a difficult task, but that should not prevent them from providing the games. ... So, I find your question to be perfectly valid, but the possible answer might be that there are more urgent issues, more pressing matters, at least that was the reason to postpone the implementation of some implementable features in many cases."

Thanks for your thoughts & time estimates on this. I feel Chessbase needs to be "encouraged" to support Chess960 technically. I'm not sure what the oldest versions of Fritz & Chessbase that had Chess960 capabilities were, but I think it's at least as far back as Fritz12 but not as far back as Fritz8. Of course, Fritz & Chessbase are not Javascript, etc., but you'd have hoped they would have had some webviewer code by now.
.
The file they link to in the article above is pgn. I'm not sure what cbv Chess960 possibilities exist now. I wish Chessbase would make this all clear.

celeje 10/9/2018 02:22

@ Petrarlsen:

Thanks for the message. I've been trying to space out comments so people have time to read & reply, but I'll be adding more.

Petrarlsen 10/9/2018 01:35

@ celeje and lajosarpad : I WILL answer!... but probably not until one day or two - there is already quite a heap of new comments, and I quite "suspect" that several others will appear rather soon, so it will take me some time to answer...

@ lajosarpad : Quite interesting that you gave once more your viewpoint - as you don't seem to fully agree with either celeje nor I, it certainly improves this debate!

lajosarpad 10/8/2018 12:04

@celeje

I agree with you: the king in the center being less safe is not necessarily untrue, but at this point it is an unproven assumption if it is applied to the 960 positions and even if we narrow it to the traditional starting position. It seems to be true, it is intuitive, but, if it were not true, this would not be the very first intuitive thought to be invalid. For instance, the flat Earth concept was very intuitive, but proven wrong eventually, even though it starts to get roots again, unfortunately.

lajosarpad 10/8/2018 12:03

@Petrarlsen

For instance, a very common tactic is to move a (possibly defended) knight on c2, f2, c7 or f7 where it can win a smothered rook. But if petrarlsening is possible, this tactic would not longer be such powerful by itself. If a check would be involved, that would of course change the tide. Also, in many cases the threat is more powerful than its execution. The fact that one can be petrarlsening any time makes it difficult for the opponent to build up an attack against his or her position. In many cases such attacks would backfire. I am not advocating petrarlsening to be introduced, I am just explaining that it could be an integral part of the game with the concept of military deception used as an explanation.

"My impression was that it was rather obvious that the King is less safe in the center that on the flank."

I know, but it was not proven yet. We have practical knowledge that in many cases "leaving the king in the centre" makes it unsafe, some of the finest attacking combinations is involving this motif. But if the players know from the start what castling means in the position, they can prepare it. Normal castling can be unsafe as well, see the many mates done against castled kings.

I understand your points about the King being safer at the corner in general and they are quite valid. I disagree though that the fact that players are frequently castling by hand would be an argument. The fact that players are doing it does not necessarily mean it is correct. In fact, the difference that the Rook will be on a different file after the castling in comparison to its file before the castling might greatly change the evaluation of the position, it could defend vulnerabilities, for instance. In the corner there are fewer possibilities to attack indeed, but we should not forget that there are fewer possibilities to run as well...

"A supplementary element : When you say : "I do not think that having a very difficult castling possibility would have any effect on a position where a good castle can be done at the first move. People will simply not care about the other castling possibility.", I think that this also constitutes an argument against positions for which castling is possible from move one[...]"

Indeed, one of my main intentions was to argue against castles which are possible from the starting positions. I think you had a great point when you brought up this problem.

@celeje

Thank you for quoting Jacob Woge's reasoning, it is interesting indeed. I also agree with you that the reasons Petrarlsen gave about the king being safer in the corner seem to mainly deal with our current knowledge of the starting position of classical chess.

The fact that Chess960 games are not as supported is a technical question. In order to do so, they need to implement the rule checker (maybe 6 hours of development), they need to accomodate the UI (this might take 32 -40 hours of development). I think this could be done in a week or so. Of course, they need to adapt the database on the long run, which is a difficult task, but that should not prevent them from providing the games. Ten years ago with a friend we implemented a playable and replayable chess game (without engine) in 3 days. And ten years ago I was very inexperienced in professional programming. So, I find your question to be perfectly valid, but the possible answer might be that there are more urgent issues, more pressing matters, at least that was the reason to postpone the implementation of some implementable features in many cases.

lajosarpad 10/8/2018 11:12

@Petrarlsen

About handling the starting position as the only possible starting position when we validate it my point was that since there are more possible starting positions, the possibility of getting a more satisfactory position might be less problematic, just like in Poker, when you get bad cards, you are unlucky, but it is not unjust if the randomization was correct. However, randomity as a possible deciding factor would decrease the value of the game in my opinion. Inferior randomized positions could be remedied by playing a similarly inferior position, where the imbalance favors the player who was unfavored by the imbalance. For instance, we play position X and I have a disadvantage. By itself this is problematic, but if we then play position Y where you have the disadvantage and the disadvantage is similar, then the imbalances are neutralized. So, the problem is solvable in tournaments, provided that imbalances are addressed by the system, but if we play a friendly game and we randomize the position and you have a disadvantage, then we will not necessarily play another game which would neutralize your disadvantage. So, while the problem is solvable for tournaments, it is unsolvable for friendly games. And this is a big problem, because friendly games are the hotbed of new players. If the number of hobby players is reduced, then the game will slowly die. Your proposal would handle the problem of hobby games as well, so I tend to think your point is probably valid.

I understand your point about the concept behind castling and you are right when you say that the move can be explained, but I still think that if I was starting to play chess where castling is undefined and play it for 20 years, then the concepts explained for castling would seem to be very strange. As strange as petrarlsening seems to be, if we think about it analytically, I do find some concept which could be used as an explanation. It is a move which allows the knight to enter the castle. Now, if we assume that petrarlsening would be possible only in the case the involved rook and knight were not moved yet, we assume further that only neighbouring rooks and knights can be petrarlsening and finally, we assume that castling would not exist, then petrarlsening makes sense. Think about the game (in this petrarlsening variant of chess) where 1. b4 e5 moves were played. Now, petrarlsening becomes a considerable option. 2. P-P-P, defending the b4 pawn, preparing c4 and d3. Another possibility is to play a gambit, like 2. g3 Bxb4 3. P-P-P, preparing an attack, involving Bg2 and c4 as well. In contrast with castling, which is a mainly defensive move in most cases, petrarlsening would be an aggressive, attacking move. Also, take a look at the knight in the corner: after the b4, c4 and d3 construction was made, it can be developed behind the pawn skeleton, by moving it to b3 or c2, controlling d4. In my opinion petrarlsen would weaken the position if castling exists, because it makes castling impossible, but if castling doesn't exist, then petrarlsening becomes an interesting option. I think some experimental games where castling does not exist, but petrarlsening is possible would give us practical knowledge about the value of petrarlsening. Even though it would not be a final answer, it would at least give us a feeling. Our minds are too clouded with castling, pawn barriers in my opinion to be able to realistically estimate the value of petrarlsening. The strategic change which would come with petrarlsening would be immense.

celeje 10/8/2018 03:53

@ Patrarlsen:

Correction --- two comments below,
"I'm not saying your "K moving closer to the center" is definitely untrue..."
was meant to be
"I'm not saying your "K moving closer to the center is less safe" is definitely untrue..."

celeje 10/8/2018 01:55

@ Petrarlsen (2):

On to your argument #2.

Petrarlsen: "Yasser Seirawan: "Castling allows a player to move his King out of the center (the main theater of action in the opening) to the flank, where the King can be protected by pawns." "

Again, Yasser's talking about the traditional starting position when he says the center is "the main theater of action".

But the other thing is related to our previous discussion about folktale history vs real history. Especially for non-advanced teaching books, authors don't think too carefully before they write general statements like that. This is just the book-writing business. When they write those types of books, they are not thinking of them as the pinnacle of their chess contribution to mankind. They write them like a reporter writes a newspaper column. They oversimplify all the time. Even if they're written more carefully, general principles are just a crutch, anyway.

I mean that in general about authors about chess books. I'm not saying Yasser is terrible compared with other authors of the same level of chess book. Even Nimzo's book, which is claimed to be so important, blah, blah, blah, gets attacked by some superGMs. The attackers say that Nimzo actually wasn't motivated to write some immortal classic. They say he needed money and got this book deal, so he wrote some sophisticated-sounding stuff and got the money he needed.

celeje 10/8/2018 01:42

@ Petrarlsen:

It seems your arguments are influenced by our beliefs about the traditional starting position. I can't remember how extreme the flank experiments were in Carlsen-Nakamura. Of course some P move may be necessary just to move the pieces out of the way to allow castling too. But we don't know that some extreme flank opening isn't very good in certain positions. As I said in a previous comment, I think that's good, not bad. I want to know how the other 959 positions change ideas about the opening. That can only seriously be discovered by concrete play and analysis.

The fianchettoed B is not the only thing with a P move in front of the K. It's very common for a3 or h3 to be played by White (a6 or h6 for Black). No one panics and says the K. was fatally weakened by these moves.

I'm not saying your "K moving closer to the center" is definitely untrue. I'm just saying we don't definitely know and finding out is a reason FOR playing, NOT a reason AGAINST playing.

Petrarlsen 10/8/2018 12:18

@ celeje:

In fact, I had already taken into account your last post's counterarguments when I wrote my own last post.

- For 1a, I said: "(...) some of the central (in the broader sense of the word - c to f) pawns (...)". Yes, it is obviously true that, be it in traditional chess or in Chess960, it is possible to play without putting the d and e pawns in the centre (as you said yourself, it is the most well-known theme from the hypermodern school...). But I think that it is only in a complete minority of hypotheses that it is possible to play the opening while keeping in their original positions at least three contiguous pawns between the four c to f pawns. (I write : "at least three contiguous pawns between the four c to f pawns" because, obviously, to keep only the d and e pawns on their original squares wouldn't constitute a sufficient protection for the King).

As for: "And P moves (one square at least) to allow development can be as safe or safer. The obvious example is K castled behind fianchettoed B.", I think that this is true only in some hypotheses: for example, in your own example about the King castled behind a fianchettoed Bishop, the King is quite safe because of the presence of the Bishop, but, if a player was to put his g-pawn on g3 without putting at any moment his Bishop on g2, in many cases, he would have very serious problems with the safety of his King. So I don't think at all that to move pawns before the King is of so little importance as that.

- "The counterargument to your argument 1b) is that if the K is in the center it can run away to either flank. If it's in the corner, it better be safe there, because it's not so easy for it to escape. So I don't find argument 1b) on its own a decisive argument."

I must said I am not convinced! The King is a very slow piece (none is slower), and, without castling, an "Houdini act" would certainly not be the King's prime speciality; in my opinion, in most cases, it would be much safer to have the King in a corner than to keep it in the middle of the board hoping to be able to escape to one side or another. To give a slightly fanciful, but nonetheless illustrative - in my opinion... - example, it is as if a medieval knight, alone in a valley and knowing that an ennemy army will attack very shortly, would stay in the middle of the valley, this while he could have taken refuge in any of two castles on both sides of this valley, each of these castles being 5 km away from him. Yes, when he will face the ennemy army, he will have the choice of going into the direction of either castle, but he will be killed or taken prisoner before he will ever have a chance to reach one of these castles! And, in my opinion, in a large majority of cases, at the end of the opening or the beginning of the middlegame in chess, the King is more or less in the situation of this knight!

celeje 10/7/2018 03:00

@ Macauley Peterson, @ Chessbase:

"... Carlsen-Nakamura ... Kasparov-Topalov ... "

Speaking of which, it'd be better if Chessbase can make these games playable on their webpages, like this webpage. I don't know why we can't play over the St. Louis games above, just as we can for regular chess games.

"Downloading a PGN of the whole event to replay in ChessBase 14 or Fritz 16", suggested in the article, is better than nothing, but not as convenient as playing it over in the article. If it can be done in Chessbase or Fritz, what technical problems are there putting it in the webpage?

celeje 10/6/2018 08:11

@ Petrarlsen:

I kind of gave the historical (according to Jacob woge) view, which is also Jacob woge's view. It feels like that may be a bit short-term, though. By short-term, I mean obviously what pieces are on the same file can change.
So I'm not sure long-term how much K being opposed to Q or R initially matters.
(I wasn't trying to address your four arguments to lajosarpad there.)

As for your numbered arguments to lajosarpad (why you think "the King is less safe in the center"), some of them seem to be based on our beliefs about the traditional starting position. What of our beliefs about opening theory are specific just to the traditional starting position is one of the things that is theoretically very interesting about the other 959 positions.

e.g. 1a) This may be true, or it may be only for certain of the 960 positions. Already in the Carlsen-Nakamura match, both tried flank openings. It shows they are open to the idea that flank openings may be best for certain starting positions. Kasparov-Topalov & the other St. Louis games maybe also had some flank openings. I'd have to look at them to see.

Even for the traditional starting position, historical attitudes changed from believing the centre had to be occupied by pawns to hypermodern opening ideas. And P moves (one square at least) to allow development can be as safe or safer. The obvious example is K castled behind fianchettoed B.

The counterargument to your argument 1b) is that if the K is in the center it can run away to either flank. If it's in the corner, it better be safe there, because it's not so easy for it to escape. So I don't find argument 1b) on its own a decisive argument.

Petrarlsen 10/6/2018 05:48

@ celeje:

As for what you developed in your last post, my arguments n° 1 ("As a preamble, my reasoning: If the King is in the center: (...)") and n° 3 (" In quite a number of opening's lines featuring queenside castling (...)") still hold completely. (About argument n° 3, it must be noted that there isn't any Queen or Rook either on the b-file or the c-file, in the traditional chess' starting position, so this cannot have any link whatsoever with a displacement of the King from the c8 square to the b8 square.)

As for argument n° 4, it doesn't seem to be much linked to your own arguments: 1) When castling "by hand", in general, the King goes from a square which doesn't face the Queen or a Rook to a square which doesn't face either the Queen or a Rook. 2) The players must spend quite a number of move to obtain this, so it seems to be quite useful to take the King from the center, and to bring it to the flank, otherwise, they wouldn't do it.

What would be quite interesting would be to have the opinion of several GMs on this subject (but not only GMs that are diehard proponents of Chess960, as my very strong impression is that many Chess960's proponents, GMs or not GMs take as some sort of a dogma that Chess960 is perfect as it is, and it could quite well warp their reasoning for all the themes linked to Chess960).

celeje 10/6/2018 02:29

@ Petrarlsen:

Petrarlsen (to lajosarpad): "My impression was that it was rather obvious that the King is less safe in the center that on the flank."

Jacob woge already had doubts about that in previous discussion. He claimed that historically it was about what file you're on. He claimed when the Q got increased power, suddenly the K was in danger on the same file as the Q. This was in the starting position with K&Q opposed, which I later commented seems to be called the "crosswise" configuration.

In a later post, Jacob woge replies (to you I think):
"With the pieces shuffled, those squares are not necessarily safe. You may well castle into the file of the enemy Queen, or Rook."
So clearly Jacob woge agrees with the belief that safe squares are about what enemy pieces are aligned on the same file.

Of course, the apparent existence of the crosswise configuration is significant itself. It suggests the starting position we are used to was not considered everywhere as how it must be. It also suggests that a perfectly symmetrical start position was not essential to them.

Petrarlsen 10/6/2018 02:21

@ celeje :

Thanks !

I will also continue to follow the comments on this page.

News

ChessBase '26 - Mega Package

Expand your Chess Horizon The perfect equipment for 2026 with the latest ChessBase program '26, Mega Database, ChessBase Magazine and Premium-Account!

€349.90

Reversed Sicilian Power - Win with 1.c4 e5!

In this course, IM Nico Zwirs presents the Reversed Sicilian as a powerful and practical weapon for White

€39.90

ChessBase Magazine Subscription (Digital)

The magazine for professional chess

€19.90

Master Class Tactics - Train your combination skills! Vol.4

In almost every chess game there comes a moment when you just can’t go on without tactics. You must strike to not giving away the advantage you have worked for the whole game.

€39.90

ChessBase Magazine Extra 229

Opening videos: Daniel King presents new ideas against Caro-Kann with 3.Nc3 dxe4 4.Nxe4 Nf6 5.Nxf6+. ‘Mikhalchishin's Miniatures’: Najdorf, Petroff and Scotch. ‘Move by Move’ with Robert Ris. ‘Lucky bag’ with 37 analyses by Ganguly, Illingworth et al.

€14.90

SHOP

SHOP

Champions Showdown: Improving the format

ONLINE SHOP

ChessBase Magazine 224

Chess960 is here to stay

Sinquefield is sold

Room for improvement

Picking the position

Kasparov's final thoughts on the Champions Showdown

Links

Discuss

ChessBase '26 - Mega Package

Reversed Sicilian Power - Win with 1.c4 e5!

ChessBase Magazine Subscription (Digital)

Master Class Tactics - Train your combination skills! Vol.4

ChessBase Magazine Extra 229

Understanding Middlegame Strategy Vol 13: Reversed Colour Systems – Grünfeld & Dutch

A powerful 1.e4 Repertoire

ChessBase Magazine 229

Pop-up for detailed settings