hgm wrote:Oops, my mistake! Seems I played 2. Qba2 in stead of 2. Qda2, when entering the moves of your proposed plan. So the Pawn was not defended. The game QueeNy-QueeNy was cheating anyway, as it turned out that pondering was on. So the Knights side was not having nearly as much disadvantage of the time odds as was intended.
I now replayed the game with ponder off, 40 min for the Queens, 4 min for the Knights. Because I let QueeNy play from the very beginning there, it did actually play Qda2, and not the faulty move I entered. This makes life indeed a bit harder, as 5 Knights vs 2 Queens is in fact rather close to equality. Black will have to play accurately, and at 1 sec per move becomes prone to tactical mistakes.
Very funny, indeed.
Queeny started its first game with a 33 move win over Stockfish, then it won in 45 moves with 1.b5 against Stockfish and in 54 moves against itself, and finally, after the right sequence of moves, a whopping 75 moves are neccessary to Queeny to win the position against itself. By that rate, I am afraid to suggest another improved line, as this time Queeny might need much more than a 100 moves to finish the game.
Could you repeat the same experiment (of course, if you have time and feel like doing so) but with Stockfish playing white this time under the same conditions? I bet this time Queeny will not win.
Btw., 5.Qaa2, although looking quite natural, might not be the only meaningful move. Another possibility is 5.Qb8 for example.
I tested one of the stockfish developement versions from 29.9.2013 with a deep search and it suggests 1.Qda2 and not 1.b5 but it is interesting that the score goes down after many iterations that it was stable(the score initially dropped).
In many iterations(depths 30-42) the main line was
1.Qda2 Nb-d5 2.Qexd5 with slighly more than 5 pawns for white.
Only at depth 43 it fails low with the line 1.Qda2 Nb5 and I do not plan to wait for it to solve the fail low).
Here is the analysis from arena
FEN: 2n2nk1/n1p1nnp1/1n3p1n/8/1P1PQ3/1Q3P2/3Q4/6K1 w - - 0 1
Lyudmil Tsvetkov wrote:Could you repeat the same experiment (of course, if you have time and feel like doing so) but with Stockfish playing white this time under the same conditions? I bet this time Queeny will not win.
Btw., 5.Qaa2, although looking quite natural, might not be the only meaningful move. Another possibility is 5.Qb8 for example.
I can do that, (actually the computer is already working), but before we lose ourself in too much detail, let's consider what you are trying to prove here:
You claim that the way to win this is 'sacrifice' Q for 2N, and that engines don't win, (and then actually lose) is because they are biased against doing such trades. Which is to say that the trade is necessary and the only way to avoid a loss.
That in itself is already an admission that two Knights are worth more than one Queen, in this position.
Whether after the trade the position is really won or lost doesn't even matter, and can always be tuned by adding or deleting a Pawn here and there. But as we know that with more typical material signatures a Queen is worth at least 3 Knights, it still leaves for you to explain why the 6th and 7th Knight in this position are worth 50% more than usual, or the 3rd Queen 33% less than usual. The elephantiasis effect would explain that.
Analysing this position ad nauseam hardly serves a purpose, because, after all, it is just one position. To know whether 7 Knights in general beat 3 Queens one would have to analyze many positions. I claim that you picked a position here that is extremely unfavorable for the Knights: You put Knights at the edges, while all Queens are in high-mobility positions, the white Pawns are much more advanced than the black Pawns, white has a candidate passer... (And despite that, the Knights so far always won, and your only 'triumph' seems to be that they needed 100 moves, against a tactically superior opponent.)
And worst of all, you picked a position that is not tactically quiet, but where the Queens can force a Q for 2N trade from the start. So it isn't really 7 Knights vs 3 Queens at all, but 5 Knights vs 2 Queens, which the theory predicts to be less of an advantage. The possibility for this trade is highly uncharacteristic: it is very easy to set up positions where each Knight is protected at least twice, and each Pawn at least once. Those are the minimum requirements for a safe position, in the case that a Q vs 2N trade would be winning (which still seems very doubtful, thouugh).
Lyudmil Tsvetkov wrote:Could you repeat the same experiment (of course, if you have time and feel like doing so) but with Stockfish playing white this time under the same conditions? I bet this time Queeny will not win.
Btw., 5.Qaa2, although looking quite natural, might not be the only meaningful move. Another possibility is 5.Qb8 for example.
I can do that, (actually the computer is already working), but before we lose ourself in too much detail, let's consider what you are trying to prove here:
You claim that the way to win this is 'sacrifice' Q for 2N, and that engines don't win, (and then actually lose) is because they are biased against doing such trades. Which is to say that the trade is necessary and the only way to avoid a loss.
That in itself is already an admission that two Knights are worth more than one Queen, in this position.
Whether after the trade the position is really won or lost doesn't even matter, and can always be tuned by adding or deleting a Pawn here and there. But as we know that with more typical material signatures a Queen is worth at least 3 Knights, it still leaves for you to explain why the 6th and 7th Knight in this position are worth 50% more than usual, or the 3rd Queen 33% less than usual. The elephantiasis effect would explain that.
Analysing this position ad nauseam hardly serves a purpose, because, after all, it is just one position. To know whether 7 Knights in general beat 3 Queens one would have to analyze many positions. I claim that you picked a position here that is extremely unfavorable for the Knights: You put Knights at the edges, while all Queens are in high-mobility positions, the white Pawns are much more advanced than the black Pawns, white has a candidate passer... (And despite that, the Knights so far always won, and your only 'triumph' seems to be that they needed 100 moves, against a tactically superior opponent.)
And worst of all, you picked a position that is not tactically quiet, but where the Queens can force a Q for 2N trade from the start. So it isn't really 7 Knights vs 3 Queens at all, but 5 Knights vs 2 Queens, which the theory predicts to be less of an advantage. The possibility for this trade is highly uncharacteristic: it is very easy to set up positions where each Knight is protected at least twice, and each Pawn at least once. Those are the minimum requirements for a safe position, in the case that a Q vs 2N trade would be winning (which still seems very doubtful, thouugh).
I think you got it totally wrong (hope the mods will tolerate this, but one should argue in some way).
This is the most favourable position for the knights one could think of with an imbalance of 3 queens vs 7 knights. What favours black:
- white has 3 isolated pawns, black just one
- neither d4, nor b4 are candidates
- white has a single pawn of the king shelter, f3, while black has 2 such pawns
- black has 3 knights additionally defending the king - f8,f7,h6 - and this is one of the main reasons that the black king enjoys a relative peace with 3 opponent queens
- all black knights are defended by other knights, and Nf8 by the king; I think this is what actually makes the position very peculiar and makes somewhat stronger the knights - extremely good mutual defence, which is difficult to break, etc.
It is true that white queens are more appropriately placed than the black knights, but that is the only such feature.
Your point about elephantiasis was that a big number of pieces of lower strength are relatively stronger than a low number of pieces with bigger power, under material equality, because those pieces would limit the mobility of the bigger power pieces. I would argue that this is not so, even with fairy-tale circumstances, like this one. For me, a bigger number of pieces of lower power would do well only if they defend each other sufficiently, like in the above case. In many situations they would not do so, and then it is not clear if the rule is valid.
One obvious thing is that queens gain value with each piece and pawn, regardless if enemy or own, that disappears off the board, as they are quickest in movement. In the above situation, there are so many pieces still left, that queens have a bit difficult time to justify their real values, and I do not know if this is already an endgame. Knights, on the other hand, as the slowest in movement of all pieces, gain value with many pieces and pawns on the board, and the above situation favours them.
However, I think the crux of the matter is that black has excellent piece defence, when you have 7 or 10 knights defending other knights, and when you decide to assign 1/10 the value of the knight bonus for defence, you get an additional full 3 pawns material. That is something. If you and Mr. Schnarnagl would argue that most probably in a significant percentage of cases an extremely large, but unrealistic number of pieces of lower power would enjoy mutual defence simply because they are covering a large space and their trajectories would necessarily intersect, I would agree with that. Otherwise, with a realistic number of pieces of lower power similar arrangements would never be valid: the 3 queens would be stronger than 2 queens, 2 knights and a bishop.
And another diagram to claim my point:
[d]6k1/ppp2pp1/1nnnnn1p/8/8/7P/PPP2PP1/3QQ1K1 w - - 0 1
This is the most natural position of an imbalance of 2 queens vs 5 knights:
- both kings have good equal king shelter
- all pawns on initial squares, no special features for either of the sides
- all black knights are defended and better placed than the white queens
And yet, I would bet that even Queeny, not to mention Stockfish, would win this 99% of the games, regardless of the time control. Would you still think that black is better here?
Actually, what does queeny think of that?
Lyudmil Tsvetkov wrote:And yet, I would bet that even Queeny, not to mention Stockfish, would win this 99% of the games, regardless of the time control. Would you still think that black is better here?
Actually, what does queeny think of that?
Win this? What is 'this'? Win with the Queens, against whom? Just tell me which engine do you want to win with what side, against whom, with how much time odds, to conceed that you lost the bet.
Note that we are down now to arguing about whether 5 Knights beat 2 Queens, while my original statement was that 7 Knights would beat 3 Queens easily, and I didn't claim anything about 5 vs 2. No matter who you think is favored in the position you originally gave, fact is that there was initial tactics there that could make white force a trade to 5 vs 2. Which we seem to agree is favorable for white. So this was NOT a 7 vs 3 position anymore than KQKR with a Rook skewering K+Q is a KQKR position. Characteristics are not an advantage just because you say they are an advantage, but I don't even want to go into that, because it is a completely moot point. All that matters is that a trade could be forced, and that I can easily construct hundreds of positions where it can't.
That, plus the fact that 5 vs 2 still seems to be a sure win for the Knights even against perfect play. As the following game demonstrates again.
Results for QueeNy (4 min/40) vs Stockfish (40 min/40). Stockfish lasted only 48 moves:
Lyudmil Tsvetkov wrote:And yet, I would bet that even Queeny, not to mention Stockfish, would win this 99% of the games, regardless of the time control. Would you still think that black is better here?
Actually, what does queeny think of that?
Win this? What is 'this'? Win with the Queens, against whom? Just tell me which engine do you want to win with what side, against whom, with how much time odds, to conceed that you lost the bet.
Note that we are down now to arguing about whether 5 Knights beat 2 Queens, while my original statement was that 7 Knights would beat 3 Queens easily, and I didn't claim anything about 5 vs 2. No matter who you think is favored in the position you originally gave, fact is that there was initial tactics there that could make white force a trade to 5 vs 2. Which we seem to agree is favorable for white. So this was NOT a 7 vs 3 position anymore than KQKR with a Rook skewering K+Q is a KQKR position. Characteristics are not an advantage just because you say they are an advantage, but I don't even want to go into that, because it is a completely moot point. All that matters is that a trade could be forced, and that I can easily construct hundreds of positions where it can't.
That, plus the fact that 5 vs 2 still seems to be a sure win for the Knights even against perfect play. As the following game demonstrates again.
Results for QueeNy (4 min/40) vs Stockfish (40 min/40). Stockfish lasted only 48 moves:
[d]6k1/ppp2pp1/1nnnnn1p/8/8/7P/PPP2PP1/3QQ1K1 w - - 0 1
Play 10 games with the above position, featuring the imbalance of 2 queens vs 5 knights in a neutral environment, and report the result.
I would be happy with the following conditions: 1 to 5 minute games (1 minute per each side for the entire game would be fine), 10 games of Queeny with black vs Stockfish + 9 other engines of your choice that are not weaker than Queeny. If Queeny gets more than 2 points, I would concede that elephantiasis might have some meaning (unless you do not falsify the games ).
10 games would take some 20 minutes in all, quite doable.
I would consider only practically relevant rules: and 7 knights for one of the sides is absolutely unrealistic. Even if the rule might have some validity, which I am not convinced of, it is not applicable in real-life chess.
Regarding the initial game - that Queeny did not tire of winning game after game against Stockfish...I told you that the position favours black in terms of immaterial factors, but still, how would you explain the fact that Queeny wins vs Stockfish much quicker than against itself (45 vs 75 moves)?
Again, I do not have the time to analyse extensively, but my 5th move would be either Qb8, or Qa2.
Well, it would take a lot longer, as I would have to install all the engines. And as you already start to cast doubt on whether the results I report would be falsified, wouldn't it be more logical if you played those 20 games with engines of your choice? You would only have to download QueeNy.
There is little point in wasting my time if you don't trust me.
[d]8/8/5n1K/5n2/2p3q1/1p1k4/8/8 w - - 0 59 Fruit's demise
So there's your two points...
Lyudmil Tsvetkov wrote:Regarding the initial game - that Queeny did not tire of winning game after game against Stockfish...I told you that the position favours black in terms of immaterial factors, but still, how would you explain the fact that Queeny wins vs Stockfish much quicker than against itself (45 vs 75 moves)?
Of course I could claim that this is because QueeNy, with its superior evaluation, plays this better. But in fact number of moves means very little. More significant is when the game becomes clearly lost, but even that is dependent on a lot of random factors.
Lyudmil Tsvetkov wrote:Regarding the initial game - that Queeny did not tire of winning game after game against Stockfish...I told you that the position favours black in terms of immaterial factors, but still, how would you explain the fact that Queeny wins vs Stockfish much quicker than against itself (45 vs 75 moves)?
Of course I could claim that this is because QueeNy, with its superior evaluation, plays this better.
But for these types of positions that makes sense, right? How does Stockfish do against QueeNy when playing black?
[d]8/8/5n1K/5n2/2p3q1/1p1k4/8/8 w - - 0 59 Fruit's demise
So there's your two points...
Lyudmil Tsvetkov wrote:Regarding the initial game - that Queeny did not tire of winning game after game against Stockfish...I told you that the position favours black in terms of immaterial factors, but still, how would you explain the fact that Queeny wins vs Stockfish much quicker than against itself (45 vs 75 moves)?
Of course I could claim that this is because QueeNy, with its superior evaluation, plays this better. But in fact number of moves means very little. More significant is when the game becomes clearly lost, but even that is dependent on a lot of random factors.
[d]6k1/ppp2pp1/1nnnnn1p/8/8/7P/PPP2PP1/3QQ1K1 w - - 0 1
No one can convince me that this position is not won for white: it simply is, but maybe I should handle the white side myself.
2 queens worse than 5 knights: it absolutely makes no sense, it is almost the same as 2 queens against 4 normal minor pieces and a little bit more, which is obvious is a big advantage to the queen side.
But maybe the answer is in the strength of Queeny: is it really stronger than Stockfish and much stronger than Fruit? This might explain things.
As I do not run engine-engine games myself nowadays, and can not check your results (but I do not doubt you are earnest), probably the only way to make things straight would be to win a couple of games in a row against Queeny.
Seems I played 2. Qba2 in stead of 2. Qda2, when entering the moves of your proposed plan. So the Pawn was not defended. The game QueeNy-QueeNy was cheating anyway, as it turned out that pondering was on. So the Knights side was not having nearly as much disadvantage of the time odds as was intended.
(hope the mods will tolerate this, but one should argue in some way).
). 