Champions Showdown: Improving the format

by Macauley Peterson
9/18/2018 – Reflecting on the Champions Showdown in St. Louis, Macauley Peterson pulls a few highlight videos from the commentary webcast and breaks down what appeals to fans of the Chess960 variant, for a look at how the next event like this from the Saint Louis Chess Club could be improved. Chess960 may have its problems gaining traction, but as Peter Svidler notes, there's plenty of space on the chess schedule to try new ideas. | ChessBase via Saint Louis Chess Club YouTube

Endgames of the World Champions from Fischer to Carlsen Endgames of the World Champions from Fischer to Carlsen

Let endgame expert Dr Karsten Müller show and explain the finesses of the world champions. Although they had different styles each and every one of them played the endgame exceptionally well, so take the opportunity to enjoy and learn from some of the best endgames in the history of chess.

More...

Chess960 is here to stay

Personally, I'm a big fan of Chess960 — ever since first playing it online in the late 90s — and especially after experiencing it over the board at the Mainz Chess Classic Chess960 FiNet Open in 2008. What makes Chess960 so much fun? This question was put to the players at the recently completed Champions Showdown in St. Louis.

Clearly, the main draw is the "absolute freshness" (to quote Peter Svidler) of being able to throw out opening theory. Svidler describes the appeal as "dogfights from move one" and notes that you can expect to find yourself in a "Martian landscape" from time to time. Even Garry Kasparov has been won over:

"People enjoy the best players in the world being so creative from move one...It's still the same — the same number of squares, the same number of pieces — just reshuffling the pieces on the first and last rank, so that you become an inventor again." 

Players discuss the draw of Chess960 (a.k.a. Fischer Random Chess)

Not so fast, argue the naysayers (my own colleagues and readers alike)! The unending quest for perfection in the opening is part of the scientific and artistic merit of (classical) chess. Without it, many players — especially beginners and amateurs — will be lost. The balanced nature of the traditional starting position ("position 518" in Chess960 parlance) is part and parcel of the aesthetic harmony to be found. Disconnecting from centuries of history is throwing the baby out with the bathwater. (And probably a dozen other cliches we needn't mention.)

Veselin Topalov echoes the concern for non-expert players in the video above:

"Professional chess players will adapt to Fischer Random but for the normal chess fan, who only plays on the weekends, it will be a big problem."

Chess960 positions used

All five Chess960 positions used throughout the competition | Graphic: Saint Louis Chess Club

Our February article, "the problem with Chess960", delves into these issues and sparked a huge debate which has put it atop the list of most commented articles ever on ChessBase (albeit mainly thanks to a few super-eager readers).

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. This would allow players and fans alike to get accustomed to rudimentary opening theory to a far greater extent than the 30-minutes to one hour of lead time given in St. Louis.

Of course, this could well undermine one of the other chief motivations behind the format, that Nakamura points out in the video: The increase in the number of decisive games. The average number of draws across the five 20-game matches was just 8.4 or 42% (58 decisive games out of 100), while in classical chess the draw rate is historically around 50% (see also: "Has the number of draws in chess increased?").

Having more decisive games surely appeals to some readers of our earlier Champions Showdown post (although perhaps more so to those predisposed to exaggeration):

Abraxas79 9/13/2018 07:42
Chess960 has to be the future. Way more exciting to watch than classical chess where 90% of the top games now end in draws. 

Of course, it's also partially a matter of taste:

yesenadam 9/14/2018 09:06
I don't get the "Draws=Bad" thing, at all. As if that's all that matters. You could just decrease the time allowed until the draw ratio is down to your preferred amount, but that would be ridiculous. A decisive game decided by a blunder, error, flagfall etc isn't much fun either. What is bad are boring games where there's no fight. Some players almost never play boring games, some players nearly always do.

Sinquefield is sold

The Saint Louis Chess Club founder and patron, Rex Sinquefield, seems to have embraced the format, although as an amateur player he does find it extra challenging, as this exchange with Maurice Ashley highlights:

Sinquefield: As a club player, it's much more difficult than watching regular chess because you're immediately into all tactics. You're solving tactical problems from the first or second move. And there's no repeat formations — every one is de novo. In regular classical chess you can sit back and say OK I know this opening, I know the strategy, I know what's going to happen for the next 15 or 20 moves. Here you don't know anything. The fireworks start on move one.

Ashley: Seems like the amateurs like having the crutch, having the opening theory that they can lean on, saying at least I know the French, or the Caro-Kann — that gives me some measure of comfort. With this, there are no names for the openings that are going to come out of it.

Sinquefield: Yes, fiddle-dee-diddly-dee — no two openings alike. That's true, they might like that crutch, but after a while, they're going to see how exciting this is. I think it's just wild.

Rex Sinquefield and Maurice Ashley

Rex Sinquefield, being debriefed by Maurice Ashley after the Champions Showdown | Saint Louis Chess Club webcast

Room for improvement

There were two big problems with this year's event from a spectator and webcast producer's perspective. One is just the unfortunate reality of having five rapid and blitz games running in parallel. The show necessarily focused on one game each round, for the most part, which meant fans missed quite a lot of the live action.

The flip side, of course, is you have more great players participating in total and you can always go back and review the games independently, for instance by downloading a PGN of the whole event to replay in ChessBase 14 or Fritz 16.

The other problem is that all matches were decided before the final rounds, and most were not even close heading into Day 4. During the last day's webcast, there was a discussion about how to maintain excitement in the face of blowouts that can occur with a match format:

"It's not fun for the players, obviously, who are getting killed, for the fans [or] for the commentators", said Ashley, who suggested alternatives such as mini-matches played between different players each day, or team Scheveningen style matches (e.g. "USA vs the World"). Knockout matches were favoured by Jennifer Shahade:

"We actually don't have a lot of prestigious knockouts, only the World Cup. And KO actually is that beautiful combination of a tournament and a match — it's exactly the solution to [blowouts] — that you have short matches, so it's almost impossible to get totally blown out because it's so short that if you get totally blown out the match is over".

But Ashley worried about players from abroad (or, for instance, Kasparov) being eliminated too early, noting that having a "losers" bracket is undesirable. In a team event (like the 2011 Kings vs Queens Chess960 experiment) even a lopsided individual game has no negative impact on the dynamics of subsequent games in the event.

Commentators on tweaking the format | Saint Louis Chess Club

I think there's a much simpler option: Just to have the matches end when decided, allowing commentary to focus on matches with more sporting drama. None of the players was mathematically eliminated until the last day, but some of the blitz games in matches that were not close seemed a bit perfunctory.

Picking the position

Jennifer Shahade and Roulette chessIn the video above, there's a brief back-and-forth on how the starting position is — or should be — selected. Jennifer mentions that for Kings vs Queens they actually used a "roulette chess" wheel at one point (she co-created one in 2009, pictured), which I remember well (as the producer of the webcast in those days). It was a fun, if gimmicky, solution. 

Other suggestions were to choose two positions and have players vote on which one to play, or increase the lead time for preparation by revealing the starting position 24 hours in advance. This idea would have the benefit of allowing fans to play the position and conduct crowd-sourced opening research in advance of the professionals' games, while being less extreme than the year-long position approach, which honestly strikes me as a bit antithetical to the Chess960 concept.

But there's also an opportunity to tap into the scholastic mission of the club by having the starting position chosen by school kids, perhaps in the various home countries of the players. This way the position for the next day could become known at approximately the same time each day (say 9:00 AM in the USA or 15:00 in Europe), and video recordings made on-site could document the event for use in the webcast, adding a little more global flavour to the mix.

Kasparov's final thoughts on the Champions Showdown

After getting over the shock of having the wrong initial position at the start of Day 4, Kasparov finished the blitz session of his match with Topalov on a high note by scoring back-to-back wins, which tightened the final score to 10½:9½. Of course, everyone wants to know, will we see Kasparov back at the board again in the future? When asked, he demurred:

"It's not the greatest moment to make any promises, but look it's fun, so again, I'm not here to win I'm here just to have fun...I'm quite happy that we did something I believe historic because it's the beginning of a new era. Innovations and exploration come from St. Louis." 

Kasparov suggests that the URS™ rating (Universal Rating System) should incorporate professional Chess960 games. Will that help drive adoption of the variant outside of these rare exhibition events? Hard to say, but I'd bet the odds are greater than 1/960.

"Classical chess is position number 518 in Chess960"


Links




Macauley is Editor in Chief of ChessBase News in Hamburg, Germany, and producer of The Full English Breakfast chess podcast. He was an Associate Producer of the 2016 feature documentary, Magnus.
Discussion and Feedback Join the public discussion or submit your feedback to the editors


Discuss

Rules for reader comments

 
 

Not registered yet? Register

Petrarlsen Petrarlsen 12/10/2018 03:14
For now, as I participated on the debates on the World Championship format on several ChessBase pages, I hadn't enough time to post on this page; I will return in one or two days (if the debate on the World Championship format stops quickly, or else, after the end of this other discussion).
celeje celeje 11/26/2018 07:45
Very long discussion by Grischuk, Svidler & Giri.
Grischuk strongly favors Chess960.
Says opponents only throw insults & he's never heard an argument against Chess960.
Petrarlsen Petrarlsen 11/19/2018 11:14
@ celeje and lajosarpad: Sorry; obviously, I will not find time to come back before the end of the World Championship (...and perhaps 2 or 3 days after it, because there are other things that will be late for me just after the World Championship!...).
celeje celeje 11/15/2018 01:36
@ Petrarlsen:

No problem. I'll keep returning.
Petrarlsen Petrarlsen 11/10/2018 07:27
@ lajosarpad and celeje:

As I try to follow the World Championship as closely as possible, I couldn't find time for the moment to comment on this page; I will nonetheless return to it in a few days.
lajosarpad lajosarpad 11/10/2018 06:58
@celeje

I have explained what should we do with the evaluation function and how could we use it to achieve improvements. You are right that the exact function was not agreed upon up to this point, but I'm not sure what is your argument behind this. You are acting as if you had an intention not to improve anything and it requires quite an amount of self-discipline on my part to not focus on your approach. But to react to your argument(?) we do not have 10 million dollars, but we are perfectly capable of imagining what would we do with that amount of money if we had it. Similarly, we do not have a function at this point which would evaluate the value of starting positions or rules, but we can of course imagine how could we use it. It is enough to know that it would depend on the rules and starting positions and it would yield a numeric value as output. If we had such a function, then it could be used to compare the values of different inputs. This way answers could be found.

"Yes, let's see, but that also depends on your interpretation of what he's saying and his interpretation of what you're saying. "

If there was a misunderstanding, then Petrarlsen can clarify it. I consider king safety to be the general aim of castling by definition. If we ask grandmasters why are they castling in general, they will answer to achieve king safety. So, if chess is the particular and chess960 is the generalization, then castling should have similar general result in all the starting positions. If chess960 did not manage to properly generalize chess in all the aspects, then the process of generalization is not 100% successful, so there is room for improvement.

I did not call moves which will never be played redundant. Redundancy is quite different from low quality and I would never mix these up.

"I think that if a move were so bad generally that it's never played that is a far bigger problem than if it's fine but has no fable, so these are very different arguments to me. It's just that I don't think it's likely that any move is so bad generally that everyone avoids it."

In the example of castling logical coherence might go hand in hand with the general quality of the move. If castling does not improve king safety in some positions, then the move will be much less attractive as a result, unless another reason is found to consider the move to be good. I do not recall anyone besides you using the word "fable". I consider the logical explanation quite different from fables and not requiring any animals, therefore I consider your whole "fable" argument to be a red herring and as such, I will ignore it in my future comments.

"No, that's not what I mean. Again, that is mixing up "coaching heuristics" (that may or may not be valid in general though not always) with "explanation of game rules". The two are totally different."

I disagree the mixing of these concepts is present in my argument. I consider the aims of moves explainable by the game phase when they are moved. Since I do not see any value in the quote above, I will not allocate further time discussing it.
celeje celeje 11/10/2018 05:43
@ lajosarpad: (2)

celeje: "First, how do I explain any move (in any game) or its purpose to a novice..."

lajosarpad: "I expected this question. The explanation is that the purpose of moves in the opening is to develop the pieces..."


No, that's not what I mean. Again, that is mixing up "coaching heuristics" (that may or may not be valid in general though not always) with "explanation of game rules". The two are totally different.

As I wrote:

"I certainly don't try to give fables. I don't do that even for children and definitely not for adults. That was not how I learnt the rules of traditional chess as a child. Do children really want the fables? I don't think they do. I didn't.

If I get someone now to teach me poker or some poker variant, I do not want them to give some fable "explaining" why a royal flush is so good, or why X hand beats Y hand. I just want them to tell me the rules, even if the fable has historical basis. If the fable has no historical basis, it's even worse."

I did NOT write that it was impossible for me to be told fables as a child learning the game. I did not write that it is impossible for me to be told fables by someone teaching some type of Poker. I said I did not get fables, and I did not want fables, and I'm not sure any other child wanted fables.
I do not want Poker fables. I just want the Poker rules.


If a fable has historical basis, meaning that was really historically in our ancestors' minds when the game was developed, then that's interesting, because history is interesting. It's interesting as "history", but still not relevant as "logic". When it's not even real history, then it's just an interesting idea and no more. The King's Leap was nothing to do with King safety historically.
celeje celeje 11/9/2018 02:31
@lajosarpad:

lajosarpad: "My approach was not assuming we would randomly choose the improvements with Monte Carlo"

I just meant that we are NOT doing any numerical optimization problem by ANY method. It sounded like you were referring to getting stuck in a local minimum in such a problem. I don't think that's anything like what we're doing.


celeje: "Yours appears to be a re-interpretation of his argument."

lajosarpad: "I would like to see Petrarlsen's reaction to this. "

Yes, let's see, but that also depends on your interpretation of what he's saying and his interpretation of what you're saying.

What I meant about your view is it sounds a bit different from his (in your recent comments to me). He thinks that without the explaining "fable", the move is illogical. Maybe you think that too, but you added that if the move is generally bad then no one will play it and it will be redundant. That's a related but different point. Then it's like the fable is just a hint that the move might be generally useful and not redundant, and it's not the fable on its own that's the main thing. (I think that if a move were so bad generally that it's never played that is a far bigger problem than if it's fine but has no fable, so these are very different arguments to me. It's just that I don't think it's likely that any move is so bad generally that everyone avoids it.)
lajosarpad lajosarpad 11/9/2018 12:37
@celeje

My approach was not assuming we would randomly choose the improvements with Monte Carlo, so I find it strange that you reacted to something I did not say. I was speaking about gradient approach to find local optimums and additional studies to find better places in the problem space.

"We are not agreeing on the function to optimize."

We are discussing the components of the function. It is natural at this stage to not agree. Also, we do not need a function to optimize per se. We need a function to evaluate the rules. Optimization is a natural result of the comparison of evaluations via the gradient approach on rules close to each-other and further analysis when the proposed deviation is greater than a tolerance level.

"Yours appears to be a re-interpretation of his argument."

I would like to see Petrarlsen's reaction to this.

"First, how do I explain any move (in any game) or its purpose to a novice..."

I expected this question. The explanation is that the purpose of moves in the opening is to develop the pieces, that is to put them into squares which improves the position. This, in general applies to pawn moves, piece moves and castling in the case of chess. In the middlegame the purpose of the moves is to outmanouvre the opponent. And in the endgame the purpose of the moves is to improve the position via technical means. Most of the moves are of general purpose depending on the stage of the game, while special moves, like castling, en passant require further explanation, which is possible in chess, while in the case of chess960 it is difficult to give the further explanation for castling. If there is nothing to achieve, players will simply not play the move. Castling as a move which in most cases improves king safety, by bringing the king to a place where it is better defended is not a fable. Opening lines confirm this. A Royal Flush in Poker is good, because most possible card combinations in one's hand are worse. Castling is good, because it improves the position by improving king safety in chess. The two cases are quite different.

I agree with you that chess960 is a generalization of chess. But I'm not sure it is the best possible generalization.
celeje celeje 11/9/2018 06:06
@ lajosarpad: (2)

lajosarpad: "Okay, let me put this differently: if you show the rule of castling to a chess960 novice, how do you explain the move? What is its purpose?"

First, how do I explain any move (in any game) or its purpose to a novice... I just explain the rules and that's it. I explain what the goal of the game is, i.e. how to win, then the legal things the players can do (e.g. moves) and that's it.

Again, there's a HUGE difference between explaining the rules of the game and trying to "coach" someone how to play well.

I certainly don't try to give fables. I don't do that even for children and definitely not for adults. That was not how I learnt the rules of traditional chess as a child. Do children really want the fables? I don't think they do. I didn't.

If I get someone now to teach me poker or some poker variant, I do not want them to give some fable "explaining" why a royal flush is so good, or why X hand beats Y hand. I just want them to tell me the rules, even if the fable has historical basis. If the fable has no historical basis, it's even worse.

As for chess960, it is a generalization of traditional chess. This is another point of disagreement. Petrarlsen advocates it being totally divorced from traditional chess. For me it's very important that it is a generalization, which e.g. is a huge negative of b/g-file castling.
I therefore would teach chess960 rules to players of traditional chess. If the "novice" knows no chess at all, I teach him traditional chess first and then chess960, possibly immediately after with no break.
celeje celeje 11/9/2018 05:48
@ lajosarpad:

celeje: "We have the bigger problem here that we clearly are not agreeing that the change is even a local improvement."

lajosarpad: "In optimization theory local optimums might be worse than the status quo. However, with a gradient approach a better optimum can be reached than the status quo, unless it is already a local or global optimum. Local optimum is not synonymous with improvement. "

This is not relevant because we are not randomly (e.g. Monte Carlo) trying to sample the whole space of possible rule sets. If we were trying to do some sort of optimization that sampled over the whole space, we'd first need a well-defined function to optimize. We are not agreeing on the function to optimize. Even you and Petrarlsen appear not to be arguing the same thing re. his "logical coherence" argument. Yours appears to be a re-interpretation of his argument.
lajosarpad lajosarpad 11/8/2018 03:09
@celeje

"We have the bigger problem here that we clearly are not agreeing that the change is even a local improvement."

In optimization theory local optimums might be worse than the status quo. However, with a gradient approach a better optimum can be reached than the status quo, unless it is already a local or global optimum. Local optimum is not synonymous with improvement.

"I strongly disagree with this statement. This is part of Petrarlsen's "logical coherence" argument. But the final rules are just abstract rules. Such an "idea behind a rule" AT BEST is a serious consideration if it has historical basis. (Not that we have to let history dictate everything.) We see no such historical belief linking castling to K safety. We had this discussion with jacob woge in the comments to the first article. Murray seems the best English source. Jacob woge had a German-language source. "

Okay, let me put this differently: if you show the rule of castling to a chess960 novice, how do you explain the move? What is its purpose?

"And even IF we decide now that castling should be about K safety (and there's no reason we have to), there's no reason EVERY aspect of it should be safer. Overall, with the R on the "correct" side, the K may be safer. If it's not safer in EVERY aspect, I don't think the rule is "incoherent", because even normal castling in traditional chess is not "coherent" in that sense in general. It only could seem so if you consider only the K's final square, which is NOT the only aspect of safety. "

In chess we can say that the aim of castling in general is to improve king safety. There are exceptions though, but from the castling moves the aim in more than 90% of the cases was to improve king safety.

"You want a fairy tale or fable that can explain it."

This is quite rude, "analyzing" intentions. It is better to calm down before one writes a comment.

@Petrarlsen

I have meant only king and rook moves necessary in order to do the castling. If there is another piece blocking the way or the king would go into check, then castle would not yet be possible, just like in chess. The idea is to remedy the problem of people avoiding to castle if it takes additional king moves. It was meant to be a possible addition to your idea. Also, your idea has another problem to solve. What if in order to do a castling one needs to move with the king first, but the person moves with the king, then makes some other 10 moves and then castles? These are the kinds of problems my additional idea aims to remedy.
celeje celeje 11/8/2018 11:44
@ Petrarlsen: (2)

celeje: "We know "why there is a move called castling in traditional chess", and it has NOTHING to do with King safety."

Petrarlsen: "In modern-day chess, castling has EVERYTHING to do with the King's safety. Ask to any GM."


No, the EXISTENCE of a move rule, not whether it is a good move to make.
You fail to distinguish between the two. Rules exist and then it is for the players to decide when, how, why etc. to act within those rules.

The EXISTENCE of the traditional castling rule is not because of K. safety.


And modern-day GMs do not think in fables or fairy tales anyway. (Not that it matters, because, as I said, deciding what legal moves are good is completely different from deciding what moves should be legal.) GMs calculate. Even the ones who don't like things tactical are just calculating. Otherwise, the time control would not matter. You don't sit there for 5 minutes or 10 minutes or 30 minutes for a move trying to remember the "correct fairy tale" to apply to the position on the board. You sit there for 5 minutes or 10 minutes or 30 minutes CALCULATING.

We discussed this before. The fairy tales are just for the quickly written teaching books they churn out.
celeje celeje 11/8/2018 07:01
@ Petrarlsen:

celeje: "So how do we convince them to do that?"

Petrarlsen: "As I was speaking of computers, obviously, there isn't any need to convince a computer... "

Given that we're not in an AI world, we need to convince human owners of clusters of computers and organizers of computer chess tournaments...
If you and I and lajosarpad had unlimited computer resources, we could just run whatever tests we want. But we don't. I don't think we even have large resources.


Petrarlsen: "Obviously, a 2400+ IM (for example) isn't a low-level player, but the difference with a 2800+ GM is still absolutely enormous, and we couldn't ever be sure that an IM's idea wouldn't be refuted in 5 seconds by a 2800+ GM, so I wouldn't see how this would be useful for us. "

Just the standard elo/ tournament performance calculations, etc. How do we learn about traditional chess from all these imperfect IMs? The same way. The elo calculations take that into account. They are not perfect, but that's what we have and it works well. And things are easier still when we have same-level players against each other. IM vs IM, etc.

We don't ignore their contributions to opening theory in traditional chess, or their tournament results, just because they are "only 2500 elo IMs".
Petrarlsen Petrarlsen 11/8/2018 06:54
@ celeje:

- "By "low level", I mean like IMs and 2500 elo GMs. (Maybe 2200, 2300 elo players too?) That's not really low."

Obviously, a 2400+ IM (for example) isn't a low-level player, but the difference with a 2800+ GM is still absolutely enormous, and we couldn't ever be sure that an IM's idea wouldn't be refuted in 5 seconds by a 2800+ GM, so I wouldn't see how this would be useful for us.

- "So how do we convince them to do that?"

As I was speaking of computers, obviously, there isn't any need to convince a computer...

- "You did not calculate anything in that comment. I mean, how many human-human + computer-computer games will get the error bars calculated by the standard elo calculators small enough."

I calculated about high-level human games, and this gave more or less one starting position being played for one single round of a tournament every 20 years - so there wouldn't be any usable data before MANY decades.

For computers, there isn't really any need to calculate anything; you take some computers, you make them work 24/24, 7/7 for some weeks or months, and the result is there. But, one more time, I wasn't speaking of computer games, but of human games, as I always said that to use computers would be useful.

- "We know "why there is a move called castling in traditional chess", and it has NOTHING to do with King safety."

In modern-day chess, castling has EVERYTHING to do with the King's safety. Ask to any GM; you will see what they will answer (...apart, probably, from GMs who are diehard Chess960's proponents, as they would quite well know that this answer would be devastating for this game...).

- "You appear to want to ignore mathematical abstraction, which after the dust has settled is all the rules are." "You refuse to acknowledge mathematical abstraction."

In my opinion, to invoke "mathematical abstraction" to explain a rule is just simply completely absurd. A rule is necessarily the concrete translation of a given idea. I can find you easily all the ideas which explain all the rules, in traditional chess, and this is just the prerequisite for a rule to have its place in a game such as chess.

But, in a way, the Chess960's castling correspond to some sort of an idea, this idea being to keep a rule outwardly resembling to the traditional chess' castling, this while keeping the same arrival squares that the traditional chess' castling. And as the Chess960's starting positions are all quite different one from the others, this gives (unsurprisingly) completely strange and absurd results; as I said before, the Chess960's castling is just a bad copy-and-paste of the traditional chess' castling; nothing more... Some outward characteristics have been kept, but not the original ideas; this is more or less as if, emptying an egg of its content and replacing it by water, mud, sand, or whatever, you would say that it still is an egg... no, it wouldn't be anymore an egg; just an eggshell with something else inside; nothing more...
celeje celeje 11/8/2018 06:35
@ Petrarlsen: (2)

Petrarlsen: " If you don't know why there is a move called "castling" in Chess960, then this move hasn't any justification for existing and it must be suppressed or transformed so as to become a logical part of the whole. "

We know "why there is a move called castling in traditional chess", and it has NOTHING to do with King safety. We have that rule in traditional chess because there was a rule called the King's Leap, and because people wanted to speed up the game. There's no reason we have to follow history, but you talked about "why there is a move called castling", and that is "why there is a move called castling in TRADITIONAL chess", to the best of our knowledge, from Murray, etc.


And even IF we ignore all of that, it's not true that there is no reason "why there is a move called castling" in chess960. The more accurate statement is you are unhappy with the reason "why there is a move called castling" in chess960. You want a fairy tale or fable that can explain it. But that is not the right way to think about the rules. You appear to want to ignore mathematical abstraction, which after the dust has settled is all the rules are. No one playing traditional chess now would think about the explaining fairy tale behind a rule (even if it's a true fairy tale) when playing it. You appear very reluctant to acknowledge mathematical abstraction. And that is after you ignore history (which we are free to do, but we're also free not to do).

I think the fairy tale aspect is interesting to discuss, but it cannot stand on its own in some sort of vacuum. We cannot ignore everything else when that everything else clashes with the fairy tale.
Petrarlsen Petrarlsen 11/8/2018 06:19
@ lajosarpad:

About your proposal regarding Chess480M, there is still something I am not sure of: when you say, for example: "Also, we might consider adding that the whole process could be a single move", do you mean that, with your proposal, for example, castling would be possible from move one? That, with one move, the player could at the same time move the King, the Rook, and all the pieces that would block them? This is what I have understood, but I am not at all sure to have understood it correctly...
celeje celeje 11/8/2018 06:17
@ Petrarlsen:

Petrarlsen: " I wouldn't see the interest to include low-level human games"

By "low level", I mean like IMs and 2500 elo GMs. (Maybe 2200, 2300 elo players too?) That's not really low.

Petrarlsen: " there isn't any need to wait for anything else for such an evaluation. "

So how do we convince them to do that? People will be most persuaded to study the openings if it's played widely. They don't do that out of theoretical interest. They do that because it has sporting relevance.

I'm all for doing that AS WELL AS playing, but not INSTEAD OF playing. But as I just said above, I don't think we'll convince them to do that INSTEAD OF playing, anyway.


Petrarlsen: "It is obvious with some simple calculations (cf. my previous posts) that to have enough high-level games would take centuries"

Not with computer-computer games. There are MANY computer chess tournaments being run around the clock as we speak. They are all traditional chess, because high-level human games are almost all traditional chess. If high-level human chess960 becomes common, computer chess tournaments around the clock will also commonly be chess960.

You did not calculate anything in that comment. I mean, how many human-human + computer-computer games will get the error bars calculated by the standard elo calculators small enough.
Petrarlsen Petrarlsen 11/8/2018 06:00
@ celeje:

- "You just keep repeating that it would take centuries to find out, without any justification at all."

It is obvious with some simple calculations (cf. my previous posts) that to have enough high-level games would take centuries. I wouldn't see the interest to include low-level human games; no-one could know if the players didn't miss simple things; this couldn't be useful to really know more deeply the Chess960's starting positions. As for computer games, I remind you that lajosarpad and I precisely consider that computers + a team of GMs would be the best solution to evaluate Chess960's starting positions, and there isn't any need to wait for anything else for such an evaluation.

- "(...) for you wanting to hack off limbs from the body like Frankenstein."

I quite like this quote; it is so extreme and so completely without any link to reality that one cannot lack to be quite impressed by it!!

Taking a concrete example: if a tournament organizer, unbeknownst to the public, would suppress some starting positions, when organizing a Chess960 tournament, following you, this would be a complete catastrophe; some sort of a crime against Chess960. But, in fact, if this wasn't revealed to the public, you would applaud the courageous organizer who would promote your beloved Chess960, not knowing he is in fact the vilest of criminals! And, even if this tournament would go on for years and years, in view of the enormous number of starting positions in Chess960, you could NEVER have any idea that this tournament organizer would suppress some starting positions! So you have invented a new category: the catastrophe without any possibly perceivable negative consequence! To the best of my belief, if, for example, you hack some limbs of a human being, there ARE some perceivable consequences!!!

- "lajosarpad: "The main aim of castling in chess is to improve king safety." I strongly disagree with this statement. This is part of Petrarlsen's "logical coherence" argument. But the final rules are just abstract rules."

If there isn't any idea justifying a rule, it must be suppressed; it is just obvious, just simple common sense. No need to debate about this. If you don't know why there is a move called "castling" in Chess960, then this move hasn't any justification for existing and it must be suppressed or transformed so as to become a logical part of the whole.

...The fact is that, for a Chess960's proponent, the only way to defend Chess960 is to completely refuse any idea of logic or coherence, about the Chess960's castling, as there isn't any in it, and they know this quite well... this is certainly why logic and conceptual coherence are so despised by Chess960's proponents: their game couldn't survive one single minute in its present form if they would accept such ideas...
celeje celeje 11/7/2018 12:27
@ lajosarpad: (ctd.)

And even IF we decide now that castling should be about K safety (and there's no reason we have to), there's no reason EVERY aspect of it should be safer. Overall, with the R on the "correct" side, the K may be safer. If it's not safer in EVERY aspect, I don't think the rule is "incoherent", because even normal castling in traditional chess is not "coherent" in that sense in general. It only could seem so if you consider only the K's final square, which is NOT the only aspect of safety.
celeje celeje 11/7/2018 12:18
@lajosarpad:

(BTW, our last two comments were almost at the same time & I did not see yours before I wrote mine.).

lajosarpad: "The main aim of castling in chess is to improve king safety."

I strongly disagree with this statement. This is part of Petrarlsen's "logical coherence" argument. But the final rules are just abstract rules. Such an "idea behind a rule" AT BEST is a serious consideration if it has historical basis. (Not that we have to let history dictate everything.) We see no such historical belief linking castling to K safety. We had this discussion with jacob woge in the comments to the first article. Murray seems the best English source. Jacob woge had a German-language source.

Castling had all these links to the King's leap. There were different versions. One K's leap rule allowed the fantastic move 1.Kd3!!!
This is not about K safety.

Without that historical link, an "idea behind a rule" is just like a fable or fairytale. It's interesting to consider but not everything. The final product is more abstract than that. It's mathematics, a mathematical problem to study.
celeje celeje 11/7/2018 12:02
@ lajosarpad: (2)

lajosarpad: " There is indeed a danger when a change is issued that we will enter into a local optimum instead of a global optimum..."

We have the bigger problem here that we clearly are not agreeing that the change is even a local improvement.
lajosarpad lajosarpad 11/7/2018 11:59
@celeje

The main aim of castling in chess is to improve king safety. If while castling king safety is not improved, then castling is logically incoherent with the castling of chess. Since the aim of castling is to improve king safety, if king safety is not improved by castling, then we do not see the point why castling should be made. I also advice not to blindly believe the engines. Their source-code was written by programmers and I would be VERY surprised if chess960 engines of today would not have some legacy logic from their predecessors, i.e. the chess engines. Since castling is justifiable in chess in many cases, it would be perfectly normal to anticipate that this is taken into account by the engine. But what if the engine takes this into account in positions where castling is not as justifiable? Again, I am not saying that Petrarlsen is right or wrong. But if he is right, then we have a serious problem in logical coherence, which will be observable by players avoiding to castle if they realize such a defect. So, a good explanation is needed of what one can achieve by castling in chess960.
celeje celeje 11/6/2018 04:17
@ lajosarpad:

lajosarpad: "Petrarlsen's problem, as far as I know is not that castling towards the center is not pleasing to the eye of a human being, but rather his opinion that in some positions castling does not seem to be a viable move. If your king ends up in a worse position, going towards the center while castling and your position gets worse, then players will not aim to castle at all in the given position, which is a defect."

That has not been my understanding of his argument. It's not that he worries about "pleasing to the eye". It's that he believes in his concept of "logical coherence", i.e. that the concept of the move is not explainable and therefore "logically incoherent". I believe that is not a correct way to think about it, for many reasons. (I need to explain these.)


As for the "players will not play it" (because they think it's a bad move), I didn't get the impression Petrarlsen was worried about that. I don't think players will feel that way about "K towards the center" castling, but this is quick-and-dirty testable by us. Just pick some of the positions where this happens and play a few computer-computer games and see whether the computers refuse to castle. I bet they will castle in most games. If I had an engine (old Fritz 12 would do) installed on the machine I'm typing this comment from, I'd do this right now, but I don't.
lajosarpad lajosarpad 11/6/2018 10:55
@celeje

Aesthetics is an important matter and in a previous post I have given the example of chess problems where aesthetics are objectively defined according to a standard. However, the argument about castling is not about aesthetics. Petrarlsen's problem, as far as I know is not that castling towards the center is not pleasing to the eye of a human being, but rather his opinion that in some positions castling does not seem to be a viable move. If your king ends up in a worse position, going towards the center while castling and your position gets worse, then players will not aim to castle at all in the given position, which is a defect.

As about the Nirvana fallacy: I would not like to enter into the discussion about whether the "experiment" will take years, decades or centuries. This is a backstreet. My argument is that if some rules/starting positions seem to have serious defects, then changes done to cope with the problem are not premature and this is true regardless of the time we waste waiting for the "experiment" to yield results. We do not need to wait years/decades/centuries to implement an improvement, just to confirm what is already very probable. We can implement changes now and if later we find out that an introduced change was unnecessary or bad, then we can change the rules again. The problem is an optimum problem, where the topology's geography is determined by several factors, such as evaluation's distance from zero sum, conceptual coherence, aesthetics. There is indeed a danger when a change is issued that we will enter into a local optimum instead of a global optimum (which, by the way is still an improvement, at least probably), but the ever increasing research material ensures the possibility to eventually find out where we are wrong. Not going anywhere and preserving the status quo is effectively staying in a place for a long while, waiting for some answers which might never come, which is much less probable to be a global optimum, or even an optimum at all than applying some kind of gradient approach.
celeje celeje 11/6/2018 02:56
@ Petrarlsen:

Petrarlsen: "And as I said in this same post, no significant data will be obtained before centuries for Chess960 with what you call the current "experiment" about Chess960."

As I said in a previous post: We haven't done an estimate of how long it would take practically. Maybe we should.

It's not easy to estimate how long it would take. It depends what you're happy to include. I assume we'd include everything including lower-level human games and computer-computer games. The point is: if chess960 really takes off at the highest level, its popularity will spread to all other levels, including computer-computer. Once you include all of that, the number of games is exponentially larger. e.g. how many traditional chess computer-computer games are being played now? A huge number.

You just keep repeating that it would take centuries to find out, without any justification at all. And then you just keep repeating the words "Nirvana Fallacy", presumably based on this made-up "centuries". It's simply not true.


But anyway, as I said previously, that does not seem to me the motivation for you wanting to hack off limbs from the body like Frankenstein. Your motivation seems to be castling aesthetics, not "fairness of White advantage". (Even the fairness of White advantage could be thought of as aesthetic, because they are playing games in pairs with colors reversed. But it's far more concrete than whether someone likes a castling rule.)
Petrarlsen Petrarlsen 11/5/2018 09:53
@ lajosarpad:

I didn't forget your proposal about Chess480M; I will answer it in one or two days!...
Petrarlsen Petrarlsen 11/5/2018 09:53
@ celeje (2/2):

- "This is quite related to one of my replies to lajosarpad. Please look again at celeje 11/3/2018 12:22"

I suppose you mean this passage: "The proposals for arbitrarily amputating limbs (...) are motivated by preconceived ideas about what you previously called "aesthetics". I think some of those aesthetic ideas are misguided. I'll write about that separately."

And this must be taken with this other passage from your "celeje 11/4/2018 04:20" post: "It matters what sort of comparison we're talking about. There are multiple questions. For "aesthetics" of the rule on how pieces can move (e.g. castling aesthetics), there will not be agreement before or after the experiments. For other questions (e.g. someone claims position 567 is "too unfair" for Black), the experiments are helpful and necessary."

I don't see how this answers to what I said ("if I can choose between these plans without any experimenting, then it means that, in general, it is POSSIBLE to make choices without experimenting in practice. And thus, it is also possible to try to compare different Chess960's starting positions. Which doesn't necessarily mean that this comparison will give such a result that some positions will be considered as inferior and eliminated, but to try to compare these starting positions IS possible.").

I think that what I wrote in this last passage still holds completely: in general, it is POSSIBLE to compare different Chess960's starting positions. Whether this will permit to attain some conclusions about whether some Chess960's starting positions must be discarded or not is another matter. And to wait, one more time, "until cows have wings" is anyway not a solution for improving Chess960...

- One more point: when you say that an experiment is under way about Chess960, I don't agree, in particular because, to the best of my belief, none of these games are played with classical time controls, so they cannot be very useful to know really well the Chess960's starting positions; to use the tandem computers + collective reflexion by a team of GMs would be much more useful about this than your so-called "experiment".

- I don't agree that the criticisms about the Chess960's castling represent essentially a question of "aesthetics". In my opinion, it is much rather a question of conceptual coherence. If you design a plane, and there is a problem of conceptual coherence in your design, quite probably, the plane will finish its course on the ground, and not the ground of an airport... For something like the Chess960's castling, there will be no explosion, no fire, no sudden deaths... but this doesn't prevent conceptual coherence to be an important element for assessing it. And, quite simply, I don't see ANY conceptual coherence in the present Chess960's castling, so, for me, this constitute a big problem indeed.
Petrarlsen Petrarlsen 11/5/2018 09:47
@ celeje (1/2):

- "Look again at the post celeje 11/1/2018 12:57"

In this post, you wrote in particular: "There's a lot in between knowing nothing and just speculating with no evidence and complete mathematical proof. Just because we will not have complete mathematical proof doesn't mean we have to go to the opposite end of the scale."

As I said in one of my last posts, approximately, each position will, at the best, be played for one round every 20 years in a super-tournament with classical time controls. With this sort of "experimenting", you will not have anything usable before centuries. And to wait for centuries for improve Chess960 is simply just absurd...

- "I think on that topic the Nirvana Fallacy is held by you and lajosarpad (...)"

A good starting point for this discussion would be that you explain us why our arguments demonstrating that your reasoning is flawed by the Nirvana Fallacy are wrong... And as I said in this same post, no significant data will be obtained before centuries for Chess960 with what you call the current "experiment" about Chess960, so when you say: "there's a lot of room in between stopping with statistically insignificant results now and waiting forever", your are wrong: even if we would stop the experiment in 50 years, the results would still be completely "statistically insignificant"... so, yes indeed, we will necessarily "wait forever" with your approach...
lajosarpad lajosarpad 11/5/2018 10:00
Aesthetics is something subjective indeed. However, an agreement on this can be reached, see the quasi standard for aesthetic values in chess problems.

You are right that my thought experiment was about something completely objective, the aim was to show that in the imaginary case I presented it is easy to reach a consensus for discarding the given starting position. The real case is much more subtle, but we may reach a consensus, either to change some rules or discard some positions, or, to keep things as they are.
celeje celeje 11/4/2018 01:43
@ lajosarpad:

I agree uniqueness on its own does not guarantee a position should be kept. But "aesthetic" arguments on their own are not strong evidence it should be discarded. It is telling that your thought experiment was not about aesthetics, because aesthetics is something that people won't all agree on. Your thought experiment was about something objective. And you had to make the example extreme (mate in 1) to make it completely objective, because otherwise we still have to decide how much is "too much" White advantage.
lajosarpad lajosarpad 11/4/2018 11:21
@celeje

The thought experiment shows that the uniqueness of each position by itself is not a final, decisive argument to keep them all. It is an argument, but not a final one and indeed we can imagine some cases in which a given starting position is discarded.
celeje celeje 11/4/2018 04:20
@ Petrarlsen: (2)

Petrarlsen: "if I can choose between these plans without any experimenting, then it means that, in general, it is POSSIBLE to make choices without experimenting in practice. And thus, it is also possible to try to compare different Chess960's starting positions. Which doesn't necessarily mean that this comparison will give such a result that some positions will be considered as inferior and eliminated, but to try to compare these starting positions IS possible. "

This is quite related to one of my replies to lajosarpad. Please look again at
celeje 11/3/2018 12:22
@ lajosarpad: ...

It matters what sort of comparison we're talking about. There are multiple questions. For "aesthetics" of the rule on how pieces can move (e.g. castling aesthetics), there will not be agreement before or after the experiments. For other questions (e.g. someone claims position 567 is "too unfair" for Black), the experiments are helpful and necessary.
celeje celeje 11/3/2018 04:23
@ Petrarlsen:

Petrarlsen: " Some time ago, you said that you don't consider the Chess960's starting positions as being randomly generated."

Hang on. What do you mean there by "randomly generated"?? The one that is chosen for a specific game of course is randomly selected.


Petrarlsen: "And, furthermore, if we would wait to have sufficient reliable data about all the Chess960's starting positions with the games of one super-tournament's round every 20 years, to obtain a real result, we would have to wait until cows have wings! (...and perhaps even wait until they will be able to cross the Atlantic Ocean by air...) (And, by the way, Welcome Onboard, Nirvana Fallacy!!!...)"

We haven't done an estimate of how long it would take practically. Maybe we should. It's just not true. I already responded below to @lajosarpad about some of that. Look again at the post
celeje 11/1/2018 12:57
@ lajosarpad: ...

I think on that topic the Nirvana Fallacy is held by you and lajosarpad (for the reasons in that post --- there's a lot of room in between stopping with statistically insignificant results now and waiting forever, and not waiting forever does not mean we need to stop now)!
Petrarlsen Petrarlsen 11/3/2018 03:59
@ celeje: Some time ago, you said that you don't consider the Chess960's starting positions as being randomly generated (...I didn't agree, but this was nonetheless what you said...), but that there were the result of a choice. And you didn't at all consider this to be something negative.

If it was possible to make this given choice, it is also possible to reconsider this choice, and to make a different choice, thus discarding some positions (...and it would also be possible, theoretically, to envisage including some positions which were initially discarded...).
Petrarlsen Petrarlsen 11/3/2018 03:47
@ celeje:

- About my comparison about the writing of a law book, I don't see at all why it wouldn't be a good comparison: if I write a law book, and find two or three different general plans, for the book (in law, the plans are always considered as being very important), and if I can choose between these plans without any experimenting, then it means that, in general, it is POSSIBLE to make choices without experimenting in practice. And thus, it is also possible to try to compare different Chess960's starting positions. Which doesn't necessarily mean that this comparison will give such a result that some positions will be considered as inferior and eliminated, but to try to compare these starting positions IS possible.

- You always insist to affirm that it would be some sort of Mega-Catastrophe if some starting positions were taken out of Chess960, but all this is in fact completely absurd.

To take a concrete example: If Chess960 was to be very successful, and if there was, for example, 5 classical time controls super-tournaments per year, with 10 rounds per tournament (...it would more probably be 9 or 11 rounds, but I choose 10 to make this simpler...), with one starting position for each round, this would mean (approximately) that there would be 50 starting positions per year in super-tournaments, and that each starting position wouldn't be used in average more than for one isolated round of one super-tournament every 20 years. So, if a given position was discarded from Chess960, the result would only be, more or less, that in one isolated round of one super-tournament every 20 years, instead of a given starting position, another starting position would be used (and there wouldn't even be any special reason to think that the discarded starting position would be any better than the position effectively used).

In my opinion, to consider this as a serious problem is really, one more time, just completely absurd...

What can be a problem is NOT to discard some positions: this hasn't any real negative consequence if there still remains a sufficient number of starting positions. The sole REAL possible problem can be if some inferior positions are kept (because this can mar the games that are played with these positions). So, if there is a significant doubt about some positions, the right solution is to DISCARD them and not to KEEP them.

And, furthermore, if we would wait to have sufficient reliable data about all the Chess960's starting positions with the games of one super-tournament's round every 20 years, to obtain a real result, we would have to wait until cows have wings! (...and perhaps even wait until they will be able to cross the Atlantic Ocean by air...) (And, by the way, Welcome Onboard, Nirvana Fallacy!!!...)
Petrarlsen Petrarlsen 11/3/2018 03:47
@ lajosararpad and celeje:

"I think Petrarlsen thinks a possible rule change would improve the rule of castling because the logic behind the current rule cannot be known and it seems to be counter-intuitive." (lajosarpad)

I quite agree... Or else "I am waiting for someone to explain this logic" (convincingly, if possible!!)...
Petrarlsen Petrarlsen 11/3/2018 03:46
@ lajosarpad: I will reflect more on your proposal about Chess480M, and I will answer you later about it...
celeje celeje 11/3/2018 12:22
@ lajosarpad:

lajosarpad: "To reflect on your third answer to Petrarlsen: Let's have a thought experiment of a starting position where White has a mate in 1. If there were no other starting positions, then this would be a lot different and unique in comparison to the others. Yet, the position would be instantly discarded."

No, this is mixing up two different concerns we have been discussing. Your thought experiment is about fairness of the opening advantage (which in practice is not a problem even if it does exist, which we don't know, because they play games in pairs).

The proposals for arbitrarily amputating limbs are not motivated by the fairness question. They are motivated by preconceived ideas about what you previously called "aesthetics". I think some of those aesthetic ideas are misguided. I'll write about that separately.
lajosarpad lajosarpad 11/3/2018 11:37
@celeje

Grandmasters were not stating concerns indeed. It is possible that they do not have concerns, or they have, but they did not express it, or they have, they expressed it but we are not aware of it. In any case, our lack of knowledge of any explicit concern is much weaker than the knowledge of explicit unkoncern, but even if Carlsen, for instance would say he is not concerned, this would not mean much, unless he would add that he thought a lot of it. And even in the case when he thought a lot of it, if he did not use a powerful methodology I would not be sure his explicit lack of concern would be strong enough to shift the null hypothesis. However, we do not know about a concerning statement pro or contra at this moment. I think the null hypothesis should be that chess960 as it is is playable, but since there are quite plausible suggestions for improvement, I think time allocated into thinking and debating them would be a time well spent.

"If there are lots of games played with enough variety in them, the horizon affect should be less of a problem, because the computers are actually playing moves, so the horizon keeps on advancing."

Let's say they see 20 plies ahead. If there is a move they discard in a given position which would be better much later, but seems to be less than satisfactory using the 20 plies, then the move will be discarded. And if something else will be played, then the discarded (better) move will not be analysed as the game goes on. The horizon effect is lowered only in the case when a discarded idea is still possible after the move is being made. If the opportunity is lost, then the move will not be found by the computer at all. So, the most serious cases of the horizon effect will not be solved by the engines continuing to analyse the game as it advances, nor by an increased number of analysis. Humans have a shorter horizon and due to this a lot of opening variations were discarded for decades, until computers with their better horizons were able to discard the human evaluation.

To reflect on your third answer to Petrarlsen: Let's have a thought experiment of a starting position where White has a mate in 1. If there were no other starting positions, then this would be a lot different and unique in comparison to the others. Yet, the position would be instantly discarded. Now, we know that there are no starting positions with the exact defect specified in the thought experiment, but this does not exclude the possibility of other serious defects. A position might be very unique and different from the others and having a very serious defect at the same time. My point is that the uniqueness of a position is not a decisive argument to keep it.