The Chess Variant Pages




[ Help | Earliest Comments | Latest Comments ]
[ List All Subjects of Discussion | Create New Subject of Discussion ]
[ List Latest Comments Only For Pages | Games | Rated Pages | Rated Games | Subjects of Discussion ]

Comments/Ratings for a Single Item

Later Reverse Order EarlierEarliest
This item is a game information page
It belongs to categories: Two dimensional, Orthodox chess set but with different winning conditions
It was last modified on: 2020-12-22
 Author: Hans L. Bodlaender. Inventor: Bruce  Zimov. Knightmate. Win by mating the knight. (8x8, Cells: 64) [All Comments] [Add Comment or Rating]
H. G. Muller wrote on 2020-12-22 UTC

I suppose I should drop the requirement that the piece that castles must be royal. As at the moment the Applet generates the code the Diagram in it still thinks the royal piece is the King. I don't think there is any downside to that: if one specifies a piece can castle, that piece will obviously need partners, and the Diagram should figure out how wide the board is at that point. Even if the piece is non-royal.


Greg Strong wrote on 2020-12-22 UTC

Thanks! I'll read through the tutorial again.

I was able to make a Knightmate preset, although I did have one other issue. I set the king's move to K and the knight's move to NisO2 but castling still did not work. The issue was that the set partners line was not generated. So there may be a but where the partners variable isn't set if the castling piece isn't the king.


H. G. Muller wrote on 2020-12-22 UTC

No, the Applet always considers the King royal, and keeps the order as it is in the selection table (so that Pawns will lead the list). This is just because I tried to keep the interface as simple as possible, and an overwhelming majority of all variants would need it this way.

When making a Diagram it is easy enough to edit the generated HTML definition for swapping the order of the piece lines, or altering the royal parameter value. In the generated GAME code such post-editing is also the easiest solution to the rare cases where you would want it differently. I described this in the Game-code-generation tutorial ('Multiple royals' section). You just have to put your own assignments to the (array) variables wroyal and broyal, to overrule the default setting the include file gives them:

set wroyal (N);          // royal pieces
set broyal (n);

 


Greg Strong wrote on 2020-12-22 UTC

Thanks! I'll get this on the page.

Can it make a GC preset for Knightmate? I don't see an option to set the royal parameter. If I place knights and move them around to capture the kings do the knights then become the last in the in the list?


H. G. Muller wrote on 2020-12-22 UTC
files=8 ranks=8 promoZone=1 promoChoice=QRKB graphicsDir=/graphics.dir/alfaeriePNG/ squareSize=50 graphicsType=png symmetry=none pawn:P:ifmnDfmWfceF:pawn:a2,b2,c2,d2,e2,f2,g2,h2,,a7,b7,c7,d7,e7,f7,g7,h7 bishop:B:B:bishop:c1,f1,,c8,f8 rook:R:R:rook:a1,h1,,a8,h8 queen:Q:Q:queen:d1,,d8 king:K:FW:king:b1,g1,,b8,g8 knight:N:isO2N:knight:e1,,e8

The problem was that you defined the Knight first. The order of pieces is important in two ways:

  • if maxPromote=N (default 1) the first N pieces will be promotable
  • if you don't explicitly assign royalty through a royal=N parameter, the piece defined last is the royal

I guess something went wrong in the attempt to promote the Knight when it left the 1st rank. This in combination with the fact that you defined invalid promoChoice (namely lower case, while the defined piece IDs were all upper case).


Greg Strong wrote on 2020-12-21 UTC

I'm trying to set this up with the interactive diagram but it's not allowing the kings to move at all.  The Knight is listed first, so it should be the extinction piece...

Here's my code:

files=8
ranks=8
promoZone=1
promoChoice=qrkbnp
graphicsDir=/graphics.dir/alfaeriePNG/
squareSize=50
graphicsType=png
symmetry=none
knight:N:isO2N:knight:e1,,e8
pawn:P:ifmnDfmWfceF:pawn:a2,b2,c2,d2,e2,f2,g2,h2,,a7,b7,c7,d7,e7,f7,g7,h7
bishop:B:B:bishop:c1,f1,,c8,f8
rook:R:R:rook:a1,h1,,a8,h8
queen:Q:Q:queen:d1,,d8
king:K:FW:king:b1,g1,,b8,g8

 


H. G. Muller wrote on 2015-12-28 UTC
I still consider that a dubious claim. From what I have seen from engines playing it, Shogi is all about attacking the King with drops. To alter the King move has a much larger impact in making it unlike Shogi than making a small minority of the pieces you could drop to catch that King vaguely resemble a Gold General makes it more Shogi-like. <p> E.g. dropping a Gold two squares in front of a King in Shogi often leads to an unavoidable mate (Hishi). Dropping a Commoner in front of a Royal Knight does nothing of the sort. It is not even a mate threat, despite you having other Commoners in hand.

Fergus Duniho wrote on 2015-12-28 UTC
I said "more like Shogi," not "much more like Shogi."

H. G. Muller wrote on 2015-12-28 UTC
OTOH, when the royal piece then jumps away like a Knight it looks nothing like Shogi. So this is an extremely dubious claim. <p> Crazyhouse drop rules will never really look like Shogi. Crazyhouse revolves all around dropping Pawns on 7th rank and promoting them. In Shogi Pawn drops are usually not possible because of the 1-Pawn-per-file restriction. It is that rule that really causes the main difference in character between Crazyhouse and Shogi games.

Fergus Duniho wrote on 2015-12-28 UTC
Richard Kochanski has suggested the game Knightmate Drop Chess, which he describes as "KnightMate Chess with Crazyhouse Drop rules." The advantage to such a game would be that dropping Kings instead of Knights would make the game feel more like Shogi.

Georg Spengler wrote on 2015-01-28 UTC
Years ago I had exactly the same idea, but when I tested it, I didn t like it at all. I guess, to design a game featuring a Royal knight, it needs more than just switching the roles of knight and king.

H. G. Muller wrote on 2010-08-16 UTC
> Anyway, your rating evaluation builds on matches between computers, 
> which won't work because Zillions can never accumulate rating by 
> beating weaker opponents as there aren't any.

This is not true. Even against stronger opponents you will win a game now and then. (e.g. against a 280 Elo stronger opponent you should score 16%). 

It is more reliable to also test against weaker opponents, of course. But one can create artificially weaker opponents by giving time odds. When I have only few engines that play a variant I make a tourney where each engine participates in several versions, one unhandicapped, the others with factors 3, 10, 30 and 100 time odds. (For the weakest engines I don't have to go that far.) Then you play a tourney and caluclate the performance ratings (not incrementally, but in one fitting procedure, e.g. with EloStat or BayesElo). The unhandicapped strongest and most handicapped engines might have (slightly) distorted ratings because of the one-sided testing, but you simply take a set that usesd a Time Control somewhere in the middle, that all have opponenets on either side.

But to get back to the main point: you do seem to admit that the current ChessV is stronger than Zillions. But I know for a fact that Fairy-Max is (somewhat) stronger than ChessV, and that Joker derivatives (and SMIRF, in Capablanca) are again some 400 Elo points above that. Unlike Zillions, all the other engines support a universal protocol and can be automatically played against each other, so that I have hundreds of games between them. So if you assign Zillions 2300, I am really curious what ratings you would assign to Fairy-Max, SMIRF and Joker. Note that in norml Chess, Joker is some 700 Elo behind the top engines, like Rybka and its clones. It seems you quickly would climb to unrealistically high values.

George Duke wrote on 2010-08-16 UTC
Commoner is just middle age Courier Chess Man 700 years old. Any piece may be made royal for a change, not just Knights. If Bruce Zimov invented this in 1972 before anyone else, even if new at that time it is not a very original Mutator. Knightmate is about on the level of Betza's Avalanche Chess or Schmittberger's Extinction Chess both of the same general 1970s time period. All three are interesting enough to try once or twice though not very creative. CVPage has newer CVs making Queen royal within certain restrictions that take more creativity to dampen her power. In Fergus Duniho's Caissa Britannia, Queen is royal but cannot cross check [done earlier the same year 2003 by another designer...the other prior art will be inserted]. In Charles Gilman's Magna Carta, for example, one side has no King, and an allowable win condition would checkmate one of that side's Carrera compounds including loss of either one of the two as a mate; and Gilman has other CVs with types besides King royal. Peter Aronson makes Falcon royal in Horus, as does Joe Joyce in Falcon King Chess. Battle Chieftain gets its royal piece from among the Rooks. Jeremy Good makes King's Pawn royal in Royal Pawn Chess. Aronson also has royal Amazon in CV of that name. Granted, all the CVs mentioned from Caissa Brittania on follow Knightmate chronologically. Yet either Knightmate does not start a Cluster (the way for instance 1962 Ultima does), but rather Knightmate continues the existing Cluster of alternate win conditions, under way in likes of Losing Chess, Odds chesses, and Annihilation.

M Winther wrote on 2010-08-16 UTC
No, I am speaking of ICC *standard* rating (typically 30 min each), not blitz rating, or bullet rating. Nobody has 3600 there. Standard ratings on ICC don't seem inflated. I'm not certain blitz ratings are either. 

I played a match between ChessV and Zillions in Capablanca Chess long ago and it ended 2½-2½. But I think ChessV is better today.

Anyway, your rating evaluation builds on matches between computers, which won't work because Zillions can never accumulate rating by beating weaker opponents as there aren't any. So its rating can only go in one direction. Should it play against humans its rating would be decidedly higher than you  estimate.
/Mats

H. G. Muller wrote on 2010-08-16 UTC
Well, ICC ratings are highly inflated. Even micro-Max was at 2160 in no time, and I think top players there are around 3600. And believe me, they are not Kasparov. So subtracting 1000 seems prudent... In blitz (5 sec/move is blitz!) computers have a big advantage over Humans, as Humans tend to overlook quite some low-depth tactics at those speeds. (This can be as blunt as simply hanging a piece.) Computers do not have that weakness, and mercilessly exploit it. Tord Romstad once tried to make his program (Glaurung) to simulate a weak Human player. To that end he completely crippled it, by randomly deleting a large fraction of the moves, hanging pieces here and there, to the point where it was scoring 0% against other weak programs. It was still above 2000 on ICC in no time. People did simply fail to collect the hanging pieces, or waited with doing it until they were facing a mate-in-one threat elsewhere on the board... It is true that there are some complications, but Joker (rated slightly below 2400 on the CCRL list) switches from Qxd4 to O-O after 2.18 sec (and that is on a 1.3Ghz machine, not 3GHz!). But at a much later stage the piece can still be saved at the expense of a Pawn by playing g4 (Nxg4 Qc3!) The absolute value of ratings is arbitrary, and thus hardly worth a debate. But the heart of the matter is that in terms of playing strength, according to the best information I have Zillions < ChessV < Fairy-Max << Joker << Crafty << Rybka I happen to roughly attach the labels 1600, 1800, 2000, 2400, 2800, 3200 to this, to conform to the CCRL scale. If you think Zillions belongs at another position in this ranking, I would be interested to see some evidense of it.

Anonymous wrote on 2010-08-16 UTC
Well, ICC ratings are highly inflated. Even micro-Max was at 2160 in no time, and I think top players there are around 3600. And believe me, they are not Kasparov. So subtracting 1000 seems prudent... In blitz (5 sec/move is blitz!) computers have a big advantage over Humans, as Humans tend to overlook quite some low-depth tactics at those speeds. (This can be as blunt as simply hanging a piece.) Computers do not have that weakness, and mercilessly exploit it. Tord Romstad once tried to make his program (Glaurung) to simulate a weak Human player. To that end he completely crippled it, by randomly deleting a large fraction of the moves, hanging pieces here and there, to the point where it was scoring 0% against other weak programs. It was still above 2000 on ICC in no time. People did simply fail to collect the hanging pieces, or waited with doing it until they were facing a mate-in-one threat elsewhere on the board... It is true that there are some complications, but Joker (rated slightly below 2400 on the CCRL list) switches from Qxd4 to O-O after 2.18 sec (and that is on a 1.3Ghz machine, not 3GHz!). But at a much later stage the piece can still be saved at the expense of a Pawn by playing g4 (Nxg4 Qc3!) The absolute value of ratings is arbitrary, and thus hardly worth a debate. But the heart of the matter is that in terms of playing strength, according to the best information I have Zillions < ChessV < Fairy-Max << Joker << Crafty << Rybka I happen to roughly attach the labels 1600, 1800, 2000, 2400, 2800, 3200 to this, to conform to the CCRL scale. If you think Zillions belongs at another position in this ranking, I would be interested to see some evidense of it.

M Winther wrote on 2010-08-16 UTC
You should know that there is no point in playing a match between programs when the strength differential is huge. In this case the game wasn't over when white lost piece. The game is probably equal, while black's king is very exposed. Zillions won anyway. That piece win is not easy to see because it involves e6-e5 + h7-h6 + g6-g5. It is difficult to see with 5 sec/move. The Saitek computer plays very fine chess. It is worth 2080. I most often draw against it in rapid chess and my rating at the ICC is 2175, with peak at 2195. When I played table chess I peaked at 2179. As it is a challenge for me to beat it I think the USCF evaluation is correct. I am surprised that you underestimate Zillions so. It is ludicrous to say that it's worth less than 1500. It is vastly stronger. 
/Mats

H. G. Muller wrote on 2010-08-16 UTC
That is incredibly stupid play by white, letting its Bishop be trapped like that. It seems n error not even a 1600-rated player would make. (Well, perhaps at blitz...) I severely doubt the rating quoted for this machine.

Fact is that ratings are reltive measures, so that a rating doesn't mean a whole lot if you don't know the zero point of the scale it is measured on. This Saitek is not on any computer rating scale I know, so I can't really judge the importance of winning a game from it. I would be more impressed if you showed me a game where Zillions would beat JokerKM at Knightmate. Or SMIRF at Capablanca.

(It won't happen in a Zillion years! :-))) Not even with 100-fold time odds. On your rating scale JokerKM must be rated 5000!)

Perhaps a better idea: have a variant-spanning match between Fairy-Max (or its dedicated derivatives) and Zillions, playing a game (or perhaps two, with black and white) of Chess, Xiangqi, Shatranj, Makruk, Capablanca, Knightmate, Superchess, Great Shatranj, Cylinder and Berolina?

Of course all this is side tracking a bit from the original point, the value of Commoner vs. Knight. For it is not even claimed that Zillions thins Commoners are better than Knights, in opening or end-game...

M Winther wrote on 2010-08-16 UTC
I played a standard chess game Saitek Travel Champion versus Zillions on a 3 GHz computer, 5 sec/move. The Saitek Travel Champion is an excellent tabletop rated 2080 by USCF. I agree fully with this evaluation. Zillions won. Zillions is worth at least 2080 on a 3 GHz computer. In chess variants it is perhaps as much as 2300. Here is the game, replayable in my Accessory Chess:

Zillions Save Game File Version 0.02 CH
RulesFile=AccessoryChess.zrf
VariantName=Accessory Chess - Swedish Cannon
1. White Indicator j8 = no-extra-piece
1. Black Indicator j8 = no-extra-piece
2. Pawn d2 - d4 White H Z4
2. Knight g8 - f6
3. Pawn c2 - c4
3. Pawn d7 - d6 Black H Z5
4. Bishop c1 - f4
4. Pawn g7 - g6 Black H Z6
5. Pawn g2 - g3
5. Knight b8 - d7
6. Knight g1 - f3
6. Pawn c7 - c5 Black H Z7
7. Bishop f1 - g2
7. Pawn c5 x d4
8. aQueen d1 x d4 = Queen
8. Pawn e7 - e5
9. Queen d4 - e3
9. Pawn h7 - h6
10. Bishop g2 - h3
10. Pawn g6 - g5
11. Bishop h3 x d7
11. Bishop c8 x d7
12. Knight f3 x g5
12. Knight f6 - g4
13. Queen e3 - d2
13. Pawn e5 x f4
14. Knight g5 - e4
14. Pawn f4 x g3
15. Pawn h2 x g3
15. Bishop d7 - c6
16. Knight b1 - c3
16. Knight g4 - e5
17. Knight c3 - d5
17. Pawn f7 - f5
18. Knight e4 - f6
18. King e8 - f7 @ f7 0 0
19. Queen d2 - f4
19. Bishop f8 - e7
20. Knight f6 - e4
20. King f7 - g6 @ g6 0 0
21. Knight e4 - c3
21. Bishop e7 - g5
22. Queen f4 - d4
22. Knight e5 - g4
23. Pawn e2 - e4
23. _Rook h8 - e8 = Rook
24. King e1 - g1 _Rook h1 - f1 White H M1 = Rook on f1 @ g1 0 0
24. Pawn f5 x e4
25. Knight c3 x e4
25. Knight g4 - f6
26. Knight e4 x g5
26. Pawn h6 x g5
27. _Rook a1 - e1 = Rook
27. Knight f6 x d5
28. Pawn c4 x d5
28. Bishop c6 - b5
29. Rook e1 x e8
29. aQueen d8 x e8 = Queen
30. Rook f1 - d1
30. King g6 - h7 @ h7 0 0
31. Pawn f2 - f4
31. Pawn g5 - g4
32. Rook d1 - c1
32. _Rook a8 - c8 = Rook
33. Rook c1 x c8
33. Queen e8 x c8
34. Queen d4 - e4
34. King h7 - h8 @ h8 0 0
35. Queen e4 - d4
35. King h8 - g8 @ g8 0 0
36. Pawn a2 - a3
36. Queen c8 - c1
37. King g1 - h2 @ h2 0 0
37. Queen c1 - e1
38. Queen d4 x a7
38. Queen e1 - d2
39. King h2 - h1 @ h1 0 0
39. Queen d2 x d5
40. King h1 - h2 @ h2 0 0
40. Queen d5 - d2
41. King h2 - g1 @ g1 0 0
41. Queen d2 - e1
42. King g1 - h2 @ h2 0 0
42. Queen e1 - e2
43. King h2 - g1 @ g1 0 0
43. Queen e2 - f1
44. King g1 - h2 @ h2 0 0
44. Queen f1 - h3
45. King h2 - g1 @ g1 0 0
45. Queen h3 x g3
46. King g1 - h1 @ h1 0 0
46. Bishop b5 - c6

H. G. Muller wrote on 2010-08-15 UTC
The board size and shape, and context of other pieces and rules surely have an effect on piece values. Large board make sliders more valuable w.r.t. to short-range leapers. (E.g. in FIDE B=N, while in Capablanca B-N = 1/2 P.) Small boards make long-rnge leapers virtually worthless (e.g. a Camel on 8x8). Wide boards (cylinder boards) make Bishops more valuable compared to Rooks.

The type of Pawns could have a large effect on piece values, as the dominant way in which Chess games are decided is in end-games where pieces help to promote, or try to stop promoting Pawns. I heve never attempted to measure this, but it would be very interesting to see how piece values change in Berolina Chess. To a lesser extent King type could affect piece value, as it certainly affects mating potential. In Knightmate Rooks have no mating potential, and this could very well suppress their value somewhet compared to the Bishop.

This is why I always report board size and context when I am quoting piece values.

To Mats: 

I am not denying the usefulness of Zillions at all. I am denying the uselfulness of the opinion of someone who loses to Zillions. If someone has a brilliant new strategy, which makes a Commoner worth more than a Knight, and using that strategy against Zillions he is clobbered by it, I would not see any merit in that strategy, and disbelieve the conclusion that it makes the piece as valuable as claimed.

I have some doubt if Zillions would rank above 1500 Elo in FIDE. In my estimate, it would hardly rank 1500 on a computer rating list. But computer lists might very well be somewhat expanded, as computers are on average much more similar than Humans, and it is well known that the more similar two programs are, the more extreme their mutual results are for the same perfrmance difference against a more varied set of opponents. So low computer ratings could very well be under-estimates compared to the Human rating scale, while the ratings of the top-engines (like 3300 for Rybka 4) would be over-estimates.

Zillions is incomparibly weak compared to Rybka in FIDE, but who isn't? I don't think you have to be as strong as Rybka to draw useful conclusions about FIDE piece values. Being a mere GM would already do. But a GM would crush Zillions at Chess, I am quite sure of that. 1900-rated club payers already have the upper hand against Fairy-Max at long time controls (although not at blitz, which is always more dfficult for Humans), and Fairy-Max is quite a bit stronger than Zillions in any variant it plays. (I hope you agree with that...) If Betza lost to Zillions with his Commoners, either he was very far from GM level in this variant, or the Commoners must have been so inferior to Zillion's Knights that even GM-level play could not avert the loss. In both cases the conclusion that the Commoners are better is suspect.

I do agree that Commoners get relatively better in the end-game, and I am sure that one can design posiions where a Commoner beats a Knight (or a Bishop, or a Rook), with the proper Pawn constellation. But on the averaege, I did not see any advantage of Commoners ove Knights, even in the late end-game. And I doubt if Zillions will (although this would not really be conclusive if a stronger engine does not).

M Winther wrote on 2010-08-15 UTC
When discussing the strength of Zillions one cannot compare with Rybka. One must compare with average human strength. In fact, many tabletop computers being sold today have around 1500 Elo. This seems to be good enough for many an amateur. Zillions on a 3 Ghz computer is far above 1500 Elo. It is a remarkably good software for chess variants. Nothing compares. Its versatility makes it immensely useful.
/Mats

Joe Joyce wrote on 2010-08-15 UTC
Ha, then I sure fooled a lot of people. Always said I couldn't play the game. ;-)

Regardless of my skills, HG, do you contend that the value of a specific chess piece is in some sense invariant? Well, invariant across the range of reasonable chess games, anyhow, on effectively standard 2D boards. Clearly, a non-standard board can drastically change piece values. Consider a rectangular lattice, a diagonal lattice, a Byzantine board, and a 3D board. The exact shape of the edges also makes a difference, most notably in mating. The file count can be even or odd, making a difference in types or [initial] placement of pieces, and the values of bound pieces are affected by rank and file counts. But that's not what we're looking at. Flat, 2D, essentially rectilinear boards [without too many holes in them] are the playing surface in 90+% of games, so we'll stick to that. And ignore the 'parity' of the board, always assuming the board is perfectly fair. It's a legitimate simplifying assumption.

Having disposed of terrain, we look at environment. There are 2 ways to look at environment. The first is the most common. To illustrate, the rook is of zero practical value in the beginning of the FIDE game, and it grows in power/usability as the game goes on. Betza makes a similar comment about the commoner. It seems to me both statements are obviously true. One could even say they are trivially true. In the past, HG, you have successfully argued this observation is essentially trivial, the rook is 5 in value and the knight is 3, even though there are tactical situations where a knight is worth more than a rook. Does that hold true for all environments, though? To be continued later.

H. G. Muller wrote on 2010-08-15 UTC
Your point is not necessarily contradicting mine. I did not make the claim that any Chess player could easily beat Zillions in any variant. Exactly for the reason that you mention, namely people being unfamiliar with unorthodox pieces, and handling them very poorly, Zillions, with at least some vague understanding, can easily have the upper hand.

But that doesn't imply play is anywhere near good. It is more a matter of One Eye being king in the land of the blind.

On an absolute scale, Zillions is quite a weak player. This is convincingly demonstrated in variants for which dedicated engines exist. Depending on how much effort went in the construction of these engines, they are then 400-1000 Elo points stronger than Zillions. This would not be possible if Zillions was anywhere near perfect play. Other general variant engines, like ChessV or Fairy-Max, both beat Zillions, where Fairy-Max is in general some 100-150 Elo stronger than ChessV. But a quite basic dedicated engine is usually some 400 Elo above the minimalist almost knowledgeless Fairy-Max. Have you ever tried playing Zilions against JokerKM (or Fairy-Max) in Knightmate?

M Winther wrote on 2010-08-15 UTC
Zillions is fairly good. It beats the average chess player. I challenge anyone to try any of my implementations which have been properly programmed to advance pawns in the opening and to castle early. It is a ludicrous claim to say that any chessplayer can easily beat Zillions in any chess variant. Facts are that the amateur player won't even have  a proper understanding of all the strange pieces, while Zillions will handle them with ease. 
http://hem.passagen.se/melki9/chessvar.htm (right now the server seems to be down, however)
/Mats

H. G. Muller wrote on 2010-08-14 UTC
The whole story by Betza suggests to me that his was not sufficiently thought through. If you suddenly discover a way of handling the pieces that makes them much stronger, it is clear that you must have been quite far from optimal handling indeed, or it would not be possible at all. And it is in fact very unlikely that you would bridge 90% of what separated you from optimal handling in one sudden flash of insight. Even bridging 50% in one step requires a stretch of the imagination. So the new method of handling is likely to be quite far from optimum as well. And who is to say that an equally dramatic change in the style of defending against these pieces would not push their value back to well below that of the Knight?

Losing against Zillions is a very bad sign. Zillions is one of the weakest engines that exist, in any Chess variant. Its importance is that in many variants it is the only engine that plays them, and thus the strongest, by default rather than performance. But if you manage to lose to Zillions, you cannot claim to have much insight in the variant. That you do well agains humans does not mean a whole lot either. Humans play by pattern recognition, and they have to learn the patterns. Which is not possible for unknown variants, where there is nothing to learn from. So they are likely to just blunder along.

Computer programs are not hndicapped by this. They are equally stupid with respect to orthodox and un-orthodox pieces alike. But they can compensate for this by deep search through millions of positions. Therefore I would believe the conclusions of a computer program much sooner than that of a human without real experience in the particuar game. Especially of a program that is some 1000 Elo stronger than Zillions. If you are that much stronger, you must do something right...

Unless someone can show he masters the variant so well, that he can easily beat the engine, of course. Then his words would carry some clout.

25 comments displayed

Later Reverse Order EarlierEarliest

Permalink to the exact comments currently displayed.