How can anyone claim this is optimized by putting the least mobile minor piece on the outskirts of the board and creating positions where Bishops can't fianchetto?
ChessHistorian 21:05, 3 August 2007 (UTC)[reply]
- I apologize for accidentally deleting your question and for the delay this caused you in receiving an answer.
- 1. The ability for bishops to fianchetto is a trivial, game-specific tactic to Chess.
- 2. The move(s) required to bring knights into more readily-usable positions than the corner squares is just a reality of the game.
- Every CRC position has built-in trade-offs of some sort. It improves the defensive posture of both players with regard to knights in measure as it degrades the offensive posture of both players with regard to knights. If you like to use knights, this is bad. If you dislike being harrassed by knights, this is good.
- 3. This CRC position is optimized for its defensive stability or strength- not for its offensive playability, tactics or rapid development. The former criterion is more important than the latter criterion. It is one of the two best CRC positions available. See the select CRC analysis tool for details. You need to read this work thoroughly if you intend to continue this discussion with me.
- Note- Personally, I prefer the CRC position Embassy Chess a little for a couple of subtle reasons you may agree with. Do you have anything bad to say about that game (on its page)?
- --InfoCheck
The ability to fianchetto is a strategy, not a tactic. A tactic refers to an exact line of play that terminates in a result, such as Nxd5 cxd5 Bxd5 Bxd5 Rxd5 winning a pawn for white. A strategy is like a summary of an experience, like place your rooks in an open file.
Fianchettoing is important, and not just for the reason of parking bishops on a long diagonal. Typically you would have pawns at f2, g3, and h2 with a Bishop on g2. This is a very sturdy formation that is tough to crack. It also allows castling with ease, again adding stability to a game designed to help you destroy your enemy! Fianchettoing is a sword with many blades!
Embassy Chess looks OK, maybe too lopsided on the right side of the board. Also, it makes castling to the short side more like a queenside castle rather than a kingside castle. So, to me, it looks like the colors are reversed. You castle "kingside" to the left in regular chess when you have the black pieces. Castling to the right looks funny in Embassy.
ChessHistorian 01:01, 6 August 2007 (UTC)[reply]
I downloaded the calc.pdf file and examined all 58 pages. There seems to be a great deal of superflous information in it, as well as some underlying assumptions that are just plain incorrect. There is a discussion about the exchange vales for a Queen vs. Rook + Bishop in regular chess. The advice is to keep the Queen, but the reasons listed were not correct. While a Queen is composed of a Rook and a Bishop, the author of this paper overlooks a very important mathematical detail.
A Rook and Bishop occupy two squares on the chessboard. A Queen has the same power and requires only one square.
The power density of a Queen is much greater than the same power diluted through the occupation of two squares.
ChessHistorian 01:30, 6 August 2007 (UTC)[reply]
- I agree that the "power density" theory is the most probable cause of the observable, measurable "supreme pieces enhancement". This is an interesting point that the author should consider explaining if he wants to take the reader into that much abstract detail.
- A converse point for consideration ...
- Any piece that captures by replacement is powerless to capture on the space it occupies (obviously). It is absolutely vulnerable there.
- Two moves are required for the opponent to capture both one rook and one bishop. Only one move is required for the opponent to capture one queen.
- --InfoCheck
I have forwarded the calc.pdf file to a few people, I only heard back from one person so far. They said whoever came up with this must not be a tournament chess player (and I am being nice in paraphrasing it that way.) So, does the author of the calc.pdf file have a rating? Has this paper ever been reviewed by those who publish papers? Or, was it just created by the author and then never sent anywhere for a peer review? It seems to me that it is just a bunch of theoretically nonsense that "sounds good" to a chess novice, but it has no scientfic method. Furthermore, has a program been designed where these values were used? If so, how does it compare to other programs?
ChessHistorian 15:06, 6 August 2007 (UTC)[reply]
- It was revised and expanded into its present form mainly thru one year's intensive playtesting in cooperation with Reinhard Scharnagl using SMIRF and Gregory Strong using ChessV. It is as accurate for FRC & CRC pieces as RS's model or any other multi-variant model published on the internet and more accurate than GS's and Ed Trice's models (which are very similar to one another). It is still a work in progress.
- The author is a very good (but not great) player only at a few chess variants- not Chess. --InfoCheck
I am the publisher of Gothic Chess Review for the Gothic Chess Federation. My name is John Cartaphilus. You can see all of my Gothic Chess games here: http://www.gothicchesslive.com/one-players-games.php?id=174 and Reinhard can tell you I never lost a game to SMIRF. My last game with it was October 2006 and it was just crushed: http://www.gothicchesslive.com/javascript/game.php?gameid=2017
I had a USCF rating of 2076 after 260 tournament games played over a 6 year period. Just so you know, when it comes to chess and variants, I know what I am talking about.
ChessHistorian is a reporter for a newspaper. He also has a Master's Degree in Mathematics. He can tell you, unless you submit your research to a University to have it reviewed, it is just your unverified opinion. Unless you get it published by a reputable source, it's importance exists only in your own mind. You say your data is "more accurate" than Ed Trice's data, which was published by the International Computer Games Association review board after people of renowned such as David Levy and Jonathan Schaeffer deemed it worthy of print. I just don't understand that at all. Anybody can say something of theirs is "better" or "more accurate", but until it is published, it's just your own unverified opinion.
Bear in mind, the values that are in this published paper is not the Gothic Vortex evaluation function. Trust me, that program is extremely strong, and it does not just use those values. I have had about 100 conversation with Trice about what makes that thing tick. It changes the values of its pieces based on the pawn population. For example, did you ever try to mate with Archbishop vs. King? It's pretty hard if you start with the pieces placed in each corner of the board. Yet, look how easy it is to mate with Rook vs. King with the same arrangement. Does that mean a Rook is stronger than an Archbishop? No way! It means it can mate easier in the endgame. Vortex uses the "ease of mate" parameter to change the piece values. In a game full of pawns, the Archbishop is stronger than a Queen. Just look how many times it delivers fast checkmates when opponents screw up. As pawns come off the board, the Archbishop gets weaker, the Queen, and Rook get stronger. Chancellors get stronger, then weaker at some point. Why? They work best with some pawns still impeding the path of other pieces. On an open board, the Queen outperforms the Chancellor. On a board half filled with pawns, the Chancellor outperforms the Queen.
None of this is in that paper. It was not the purpose of the paper to discuss Gothic Vortex.
Just so you know, I have never lost to SMIRF, and I have never beat Gothic Vortex. The guy who designed SMIRF is not a chess player. The guy who designed Gothic Vortex is the last human player to have defeated Computer World Champions in both the games of chess (Deep Thought, 1989) and Checkers (Chinook, 1996). He knows how to handle programs and he imparts his skill to his own creations. There are times I am not sure how to describe who is winning in a Gothic Chess game I am annotating, and I read off the moves to Ed. Without setting up a board, he'll say things like "The bishop pair can tame the opposing Archbishop if the g-pawn can be exchanged, so white needs to secure it at all costs, even if it means playing Nj3 so that the pawn can be pushed to h3 to support g4..." and he continues what he was doing. I'd have to stare at the position, play a few moves into the game, return to the position in question, try some other variations, and 15 minutes later I would be able to summarize it like he did.
Yet here you are taking a swipe at him with your "mine is more accurate" remark.
You need to back up claims like that. And you can't use "because I said so" or "because SMIRF agrees" as statements. You need to debate Ed Trice is an open forum for all to see if you think you have something worthwhile. And Ed is someone who corrected, in his spare time, one of the world's leading researchers in artificial intelligence (the guy who solved checkers.) And I'm not just saying nice things about him just because he is my boss.
GothicEnthusiast 04:26, 7 August 2007 (UTC)[reply]
- It would be a complete waste of my time to try to filter thru this voluminous mess of partially-incoherent bragging (the truthfulness of which I have no means to verify) and unfair, unprovoked insults, most of it off-the-topic of the material values of pieces. Instead, I am going to explain some basic facts, establish some ground rules and then, from the maturity of your reaction, assess whether or not a productive discussion can possibly resume.
- Material values are the demarcation point for the universal calculation of piece values. The author is not a programmer who has a powerful chess variant program at his command to customize and test with any time he likes. So, positional values, the other vital half of relative piece values, are to some extent, the exclusive luxury of the developers of SMIRF, ChessV and Gothic Vortex. So, the model by Derek Nalls is compared to published material values for SMIRF (until recently), unpublished material values for ChessV communicated privately and published material values for Gothic Chess (used within Gothic Vortex). Since this comparison is of a restrictive, incomplete nature, there is no need for anyone associated with any of these three excellent programs to get upset. In fact, the author respects their accomplishments greatly.
- The varying degrees of skill with which the developers of these three programs have handled positional values as well ultimately determines how sound their overall, relative piece values are which are foundational to playing strength. Note that it is even possible for a programmer to attain average playing strength either by using a strong model of material values with a weak model of positional values or using a weak model of material values with a strong model of positional values. Therefore, some complex mysteries are likely to remain even after years of work.
- For material values exclusively of FRC & CRC pieces upon the 8x8 and 10x8 boards, respectively, the approximately equal quality of the model of Derek Nalls to Reinhard Scharnagl has been established thru intensive playtesting using SMIRF. [In fact, I have a reason to suspect the DN model is better although this has not yet been proven. See the link further down page for details.]
- From what I have read and been told about the function of Gothic Vortex, I suspect that it has extremely-accurate positional values, partially due to its heavy reliance upon special, game-specific rules statistically likely to be advantageous in playing Gothic Chess that it possesses EXCLUSIVELY. Note that this leaves the accuracy of its material values unknown. I suspect that the accuracy of its material values is good (a/k/a- average) but not so good that it cannot be improved upon- much less, perfect. Yes, I can explain this assessment.
- The Chess Variant Pages | Comments | Shatranj Values
- The main concepts I want you to note are the meaning of the terms "direct inversion" and "indirect inversion". In my assessment, the material values for CRC pieces upon the 10x8 board used in Gothic Chess experience neither of these two types of inaccuracies. Congratulations! However, there is another type of inaccuracy that was not relevant to this post and not mentioned: "compression of scale" or "expansion of scale".
- Compression of scale may be defined as a situation where, although no direct or indirect inversions of material values occur, the scale of material values of pieces is distorted by compression from their optimum, flat state.
- Expansion of scale may be defined as a situation where, although no direct or indirect inversions of material values occur, the scale of material values of pieces is distorted by expansion from their optimum, flat state.
- Correctly calculating the material value of the pawn is critically important since it establishes the baseline for the entire set of piece values. Correctly calculating the material value of the pawn (at opening setup) is also complicated and tricky due to its capture-only moves, moves without capture and promotional potential.
- If the material value of the pawn is calculated too high, then the material values of all other pieces will be too low. This is compression of scale.
- If the material value of the pawn is calculated too low, then the material values of all other pieces will be too high. This is expansion of scale.
- It is compression of scale that you need to pay attention to since the material value of the pawn amongst CRC pieces upon the 10x8 board is too high via the Ed Trice model.
- Gothic Chess | Piece Values
- CRC | Piece Values
- The first clue that the true scale is compressed comes via the comparison of the material values of the pawn (least valuable piece) and the queen (most valuable piece). The maximum range of values on the 10x8 board should be greater than upon the 8x8 board due to the increased mobility of the queen. Instead, it is slightly less than the standard material value for the queen in Chess on the 8x8 board of appr. 9.00- pawn 1.00 & queen 8.73 on the 10x8 board (ET model)! By comparison, RS model- pawn 1.000 & queen 9.6005 & DN model- pawn 1.00 & queen 10.16.
- The second clue that the true scale is compressed comes via the comparison of the material values of the pawn and the knight- two limited-range pieces that do not require any special adjustment to compare directly under the DN or RS models. Note that the material value of the pawn should be very nearly the same on the 10x8 board as the 8x8 board due to both having 8 ranks. Note that the material value of the knight should be mildly greater upon the 10x8 board than the 8x8 board due to it having more spaces and a higher percentage of spaces available from which it can make its maximum of 8 choices of moves. Instead, it is significantly less than the standard material value for the knight in Chess on the 8x8 board of appr. 3.00- pawn 1.00 & knight 2.50 on the 10x8 board (ET model)! By comparison, RS model- pawn 1.000 & knight 3.0556 & DN model- pawn 1.00 & knight 3.08. It is impossible to correctly calculate a material value this low for the knight upon the 10x8 board.
- Am I supposed to be in awe of someone who would proudly, irresponsibly publish such junky, out-of-range material piece values for CRC in a journal?
- It is probable that Gothic Vortex could still maintain superior playing strength overall against SMIRF and ChessV with extremely-accurate positional values and average-accurate material values.
- The playing strength of Gothic Vortex would improve if its material values were revised (as well as any-all interrelated positional values) in accordance with this advice. Only open and appreciative minds are capable of learning, though. I expect that you "super-genii" will ignore this advice, instead.
- The author has not even began to explain to you how totally inappropriate the "safe check" method is for calculating material piece values except sometimes during the endgame phase. You will not dig what he has to say (related by me) IF this discussion goes any further (which is doubtful).
- --InfoCheck
Just for your own information, EVERY chess programmer I have ever interviewed (Hyatt, Kittinger, Spracklen, Levy, Kaufman) says that pieces get weaker on a larger board. Are you telling me that on a 100x100 board that a Queen with so much "mobility" is stronger than a Queen on an 8x8 board? There's just no way! The mobility is a linear function of order of magnitude 1, yet the dilution of the piece strength grows as a function of board length x board width, which is like squaring the weakness if the length = width. Just look at the percentage of squares a queen can reach on a 100x100 board. It has 99 + 99 + 99 + 99 in the center of the board, which is 396/10,000 = 3.96% of the total square count. On an 8x8 chessboard, the queen can reach 28/64 = 43.75% of the squares.
How can reaching fewer than 4% of the squares make a piece stronger than when it can reach 43.75% of the squares?
The rest of your arguments don't make sense. Gothic Vortex won like 89 out of 96 games against programs like SMIRF, ChessV, Zillions, and others, yet you say your UNTESTED theories could make it even better. Can you explain why Gothic Vortex destroys all of these programs if its evaluation of the pieces is so far off base? And, by the way, Knights are weaker on a rectangular board than a square one. Vortex sacrifices its Knight for only 2 pawns in most of the games where it destroys its human prey. And look at this game
http://www.gothicchesslive.com/javascript/game.php?gameid=2028
where SMIRF was checkmated in a mere 32 moves. Can you explain why Vortex wins so much with such bad evaluations??
ChessHistorian 04:59, 12 August 2007 (UTC)[reply]
- Since you are sincerely trying to make a rational point, I will reply once more and constructively.
- The pawn has appr. the same material value on the 8H x 10W board as the 8H x 8W board since it is a limited-range piece that can hit only the SAME NUMBER of spaces as an absolute number on a single move upon the larger board. Moreover, the height (number of ranks) of the board (which is the same for both) is most determinant of its material value.
- The queen has more material value on the 8H x 10W board than the 8H x 8W board because it is an unlimited-range piece that can hit MORE spaces as an absolute number on a single move upon the larger board.
- In terms of the ratios of the material values of the pawn and the queen, this yields an even larger relative value of the queen to the pawn (with a fixed value at 1.0) on the larger board. [Note- The complete, adjusted formula is more complicated.]
- When you speak of the percentage of spaces the queen can hit on different sized boards, you are using relative measures- probably in accordance with the "safe check" method which is only useful sometimes during the tertiarily-important endgame [Piece values are meaningless whenever checkmate becomes irrefutable in the endgame regardless of the material sacrifices it requires.] and produces very faulty values for the primarily-important opening game and secondarily-important mid-game. By sharp contrast, the correct formula involves absolute measures as inputs.
- By the way, the percentage of spaces the queen can hit decreases inversely with the distance from its origin. For example, 100% at 1 space, 50% at 2 spaces, 33-1/3% at 3 spaces, etc.
- Finally, the author has mentioned many times, many places that his material piece values model has been tested intensively using two high-quality chess variant programs- SMIRF & ChessV. Please pay attention?
- --InfoCheck
- As it stands, it is much too easy for you to disregard my counter-reply as purely argumentative. So, I decided to take the next step from the speculative to the experimental realm. This is possible using the Zillions Of Games program. Of course, I realize it does not produce accurate piece values upon any board with any piece set. However, with ingenuity, this tool can be useful in a fundamental yet important way. After all, the ZOG program was logically designed by two chess engine experts- Mallett & Lefler. It can be reliably used to confirm basic correlations for pieces values across boards of greatly varying sizes in games that are related to Chess.
- The ratio of material values for the queen and pawn can be demonstrated to change in the way you predict OR I predict using widely disparate boards. So, I ZOG implemented 4 simplistic games that have only queens and pawns on 08H X 04W, 08H X 08W, 08H X 12W, 08H X 16W boards: Material Piece Values Tests. Push the "download now" button.
- Although these boring games can be played, there is no point in doing so. They were created expressly to have their piece values read by right-clicking on the pieces and then, selecting "properties". In this manner, you and anyone else who owns the ZOG 2.0 program can verify the truthfulness of the piece values given in my report. [Thanks to Derek Nalls for hosting this upon his web site]. Note that queen:pawn ratio gets larger as the board gets larger.
- There is nothing important left to argue about. Go back to the drawing board. You can hit paydirt in improving the playing strength of Gothic Vortex IF you revise your material piece values correctly.
- --InfoCheck
This is Derek Nalls. If you need, you may confirm the truth in this claim by sending a confirmation E-mail to me via "nethomeless@mail.com" (which is the same E-mail address currently published upon my web site [1]). I promise to reply.
I only wish to associate my legal name with what I consider my best game invention "Hex Chess SS".
All other games that I consider of lesser quality are formally invented by "OmegaMan", instead, even though I make no attempt to keep it secret that I am responsible for the gameworks (11- to date) of "OmegaMan". Perhaps it is eccentric on my part but please respect my wishes.
Thank you!
--Derek Nalls —The preceding comment was added by 24.117.209.55 (talk • contribs) 8 August 2007.