3 Champs Highlights

[Lyudmil Tsvetkov]

Oh, I see now, hope I will be playing a stronger Komodo soon

[Don]

I would agree with both solutions: the protocol one, and the engine-empowered one, as long as the issue is solved. It is indeed a pain for most normally advanced human players to follow such endings.

But I would not go as far as allowing an engine to offer and accept draws in rich positions with a lot of material, except when they are obvious draws, like for example totally locked. In distinction to humans, that need some rest from time to time, and therefore should be allowed to offer draws more or less at will, computers have enough resources of stamina to prolong the fight until it is very clear; and they should do that, as long as it is meaningful to continue playing.

Note that if one program offers a draw the other does not have to accept - so it would be up to the programs to decide.

Having said that, in computer vs computer tournaments it is often a rule to not offer or accept a draw OR EVEN resignation without the approval of the tournament director.

The origin of this rule happened many decades ago when there was a high profile computer chess tournament where 2 versions of the Spracklen Fidelity program were playing against each other. One version was an older version and the other version was the newest one they wanted to market. As it turned out I think the "wrong" program (from Fidelity's point of view) was winning - and so the operator resigned. There was no rule against it.

Even though this rule addressed that problem the bigger problem was that 2 Fidelity programs were in the same tournament - giving them twice the winning chances of any other program. And since they could control the outcome of Fidelity vs Fidelity it was a pretty strong bias.

We still have that issue in computer chess occasionally when we have tournaments with multiple family engines. Often the people running these tournament do not have the sophistication to even understand why that is bad.

If this were a tournament where programs played each other without knowledge of WHO they were playing and the operator were not allowed to make adjustments between matches I would see no problem with engine-managed draw offers. We could make a rule no offers before move 30 or something like that.

Concerning the timed potentially saved, a single fast game costs you nothing, but what about 30 000 fast games? In 10 000 of those at least 1/3 of the moves will be completely unnecessary.

[beram]

[d]r1b1r1k1/1pb2q2/2p1pn2/3p3p/PBPPp1p1/1Q2P1P1/P1RNNP1P/5RK1 b - - 0 21

This is a position from the 46th game Stockfish-Komodo.

Stockfish has just played 21.g3 and sees itself in advantage by some 0.74 centipawns, very decisive one, while Komodo thinks it is about equal.

Objectively, I think the position is already won for black, because of the following reasons:

- very weak backward (backward-fated) white pawn on f2, part of the king shelter; that makes the shelter very inflexible and black can gradually concentrate attacking resources
- 2 weak squares on the 3rd rank, part of the king shelter (f3 and h3), whereupon different enemy pieces, especially the knight, might intrude
- play is more or less closed, because of the big diagonally connected pawn chains, which gives black sufficient time to prepare a decisive strike

Black has some difficulties with developing its queen side, but I think those could be solved and are less weighty than the above-mentioned serious weaknesses of the white king shelter.

I tell you, I like this position, because it is a bit uncommon and also rich in resources that should be sought out. I could suggest a plan like, for example, Nh7-g5, Kh7, Rh8, Kg8, and only then h4, if necessary, prepare for e5 and bring the a rook in support of the attack, etc. In that case it is difficult to imagine how white could defend.

Komodo, however, proceeds to lose the already won game in just 2 moves by playing h5-h4-h3 instead, closing entirely the side where it is supposed to open files for attack.

[d]r1b1r1k1/1pb2q2/2p1pn2/3p4/PBPPp1p1/Q3P1Pp/P1RNNP1P/5RK1 w - - 0 23

This is the picture 2 moves later, where already white is lost. Black has a single semi-open file at its disposal (f), but both f2 and f3 are well guarded, white can also play Nf4, blocking this file to good avail, etc. It takes Komodo, however, some 60 moves to understand it is losing; and the same amount of moves it takes Stockfish to find a way to finally break through, but it does it in great fashion. When I think about that, it makes me happy that it is not only me that can lose games in 2 moves, but also Komodo can do that. If Komodo can do that, I have the right to do it too

[d]r1b1r1k1/1pb2q2/2p1pn2/3p4/PBPPpNp1/Q3P1Pp/P1RN1P1P/5RK1 b - - 0 23

I think black should have attempted to play e5 here, whatever the result, there is simply no other option.

[d]k1N5/1p4R1/rP2p3/P1PpB2q/PQ2p3/4Pp1p/5P1P/7K b - - 0 101

But I do not like this: a Ponomaryov-style Komodo playing until the very last glimmer of hope.

Finally, let's go back for a while to the main diagram:

[d]r1b1r1k1/1pb2q2/2p1pn2/3p3p/PBPPp1p1/1Q2P1P1/P1RNNP1P/5RK1 b - - 0 21

Komodo played very well until now, both g5 and g4 are great moves, but the point is: how many engines will be able to judge correctly that black has a considerable, probably winning advantage, and to find the correct winning way?

Any distinct engine output on this position very much appreciated.

Best, Lyudmil

Well first of all I think that black shouldn't have played h3?! and that 21 . ..Nh7 instead of h4 was better.
But I dont think black was already won here. Because for instance white can neutralise blacks attacking possibilities by exchanging black bishop after Qa3.
Houdini 3 would have played 21 ..Nh7 instead of h4 and also 23. ..e5 ! (at depth 23) instead of 23. ..Bb6 (also Nh5 would have been better chance

[beram]

What about this position from game 92 Houdini 3 - Komodo 51 dev
[d] 2r1rn2/p3pk2/3p1pp1/q2P3p/2BQ1P1P/1P4P1/P5RK/3R4 w - - 0 33

Houdini had the momentum in this game but lost in a few moves his +1 advantage
Stockfish knows in reasonable to find a much better move than 33 Rdg1
Namely Qb2 ! Houdini find it also but needs more time

[Lyudmil Tsvetkov]

Hi Bram.

Thanks for the input.

I agree with everything you say, except that black is not won. My basic idea was about the superiority of the black setup, as black could have played even better, not sacrificing a pawn on a4, something other instead of Re8, etc., while obtaining the same or very similar setup. But I am almost fully convinced that even in the game line black should be winning easily, although the exact move order might be a bit difficult to find: the multiple weaknesses of the white king shelter are simply tremendous.

I am not able to analyse deeper right now, might post something later; but the point is that white threatens nothing, while black threatens almost everything. Black should change threats constantly, but can proceed with:

- Qd7 to avoid exchange of dark-coloured bishops that would favour white or, even worse, penetration on e5/h4, etc. of the Bb4 or the white queen
- dc4 in some lines, paradoxically, if Nc3 is played
- h4-h3 and Nf3 in some lines, but with care
- the grand avenue way, not paying any attention to white threats, with Kg7, Rh8, h4, etc.

I do not know, I would quit playing chess if this is not won for black, objectively.

Would appreciate your deeper analysis on that.

[Lyudmil Tsvetkov]

What about this position from game 92 Houdini 3 - Komodo 51 dev
[d] 2r1rn2/p3pk2/3p1pp1/q2P3p/2BQ1P1P/1P4P1/P5RK/3R4 w - - 0 33

Houdini had the momentum in this game but lost in a few moves his +1 advantage
Stockfish knows in reasonable to find a much better move than 33 Rdg1
Namely Qb2 ! Houdini find it also but needs more time

Hi Bram, I am still not that advanced as game 92, but, for me, this position is already a slaughter.

What about g4, is not that winning instantly? (Again g4, obviously the main chess rule should be play g4 to win)

[Lyudmil Tsvetkov]

[d]5rk1/1Q6/5qp1/8/5pRp/5P1P/5PK1/8 b - - 0 77

This is a position from the 35th game, pitting Komodo with white against Houdini.

It is very obvious the position is a dead draw, but Houdini now plays 77...Rd8, followed by 78...g5, and already white is winning, as suddenly the black king remains without any pawns to shelter it, and white has to only transfer its rook via g1 to the queen side to weave a mating net. Of course, g5 is somewhat tempting, as the g5 pawn defends 2 friendly pawns, but its ominous downsides are evident. Black could have shuffled its king back and forth from g8 to h8 instead.

Obviously, a king always needs a cover of pawns, whenever this side does not enjoy big advantage, or does not forcefully storm and attack, which is not the case here. So, playing g6-g5 in the above position is really a fatal mistake, as the g6 pawn is still a good shelter, but not the g5 pawn. Somehow I do not believe Houdini would be able to make such a move in long time control, maybe some settings for the tournament are not optimal; but most probably the opponents have become too tough and that matters in a way.

I whole-heartedly hope that no other engines would repeat such a move; but, if there are such culprits under the engine heaven of Kaissa, I would be glad to look at some shared output.

Best, Lyudmil

[Lyudmil Tsvetkov]

[d]r3r1k1/2pq1ppp/pbppbn2/8/P2PP3/5N1P/1PQN1PP1/R1B1R1K1 b - - 0 14

This is a position from the 37th game, pitting Komodo with white against Stockfish.

Stockfish seems to have some advantage out of the opening, and now decides to sacrifice a bishop on h3.

2 moves later, the following position arises:

[d]r3r1k1/2p2ppp/pbpp1n2/8/P2PP3/R4N1q/1PQN1P2/2B1R1K1 b - - 0 16

Stockfish now plays 16...Qg4 and the white king slips away to the queen side, and after some still interesting battle, it is Stockfish that gets mated instead of mating its opponent.

Once I thought that computers are weak in only very specific areas, in which humans excel, having to do with positionality, closed pawn structures, etc. But now I know this is not so and only actually a myth: computers, just as humans, misplay every possible kind of positions, including open and very sharp ones, tactically rich in content, but humans in general have bigger difficulty to make sense of their mistakes in such fields.

I must admit that I am fully clueless in a large variety of sharp tactical positions, but still sometimes my assessment might prove a bit more correct, just occasionally. That is why, although almost certainly there is something wrong with my suggestion, I would dare present to your attention an alternative continuation of the black attack.

What about 16...Ng4 instead of Qg4?

Any engine output/substantiated and unsubstantiated comments very much appreciated, as this is a really intriguing position not only humans, but also computers might misplay. Usually engines are very sharp in comparison to humans, but sometimes just not a bit that sharp.

I did not analyse the position extensively, as that takes a lot of time and you need for many sharp lines a board with pieces to move, but I saw just 2 main lines:

- 16...Ng4 17. Qc6 Bd4 (already)
- 16...Ng4 17. Nf1 Re6 18. N3h2 Qh4 19. Rg3 Nh2 20. Nh2 Bd4,

both not very clear to me.

There are also some intermediate a5 pushes, also not very clear to me.

Again, maybe someone will know more, please help.

Best, Lyudmil

[Lyudmil Tsvetkov]

[d]8/2kb2pp/1n2p3/1p1pP3/pP1N2PP/P1K5/1NP5/8 b - - 0 44

This is a position from the 31st game, Komodo against Stockfish.

Stockfish now plays 44...g6, and after 45.g5 of Komodo all black pawns are fixed on squares the colour of the black bishop, which should be a bad sign. Also, in this way the f6 square becomes available for the white knight, which is even more dangerous.

I am still undecided if 44...h6 instead, guarding the f6 square and placing 2 pawns on squares opposite the colour of the own bishop, would save the game, but it was a reasonable option, based on a principled decision. It is difficult to believe that an engine like Stockfish, which most certainly has at least some understanding of the above-mentioned principle, would play g6 instead.

However, when I remember the number of games I have lost because I had neglected some obvious rules I had had knowledge of, I feel that engines are just like humans, and they would be able to make mistakes even in very simple endgame positions.

As I am undecided if this is the case, or the game has already been lost even with perfect play from black, I would be very happy to see some additional engine output/comments about this position.

Thanks in advance,
Lyudmil

[Lyudmil Tsvetkov]

The 16th game, pitting Stockfish with white against Komodo, was very interesting in terms of defying widely accepted principles. It was a very fine example of the advantage of the 2 knights, which comes about only sporadically, but is very real.

[d]3k4/1p3p1R/1r2b3/p2pPp2/P2P1Pp1/bP1N2P1/3K4/2N5 b - - 0 45

Position at move 45: Komodo thinks it is perfectly equal, while Stockfish sees itself correctly in advantage by almost a full pawn, but probably for the wrong reasons.

[d]8/5p2/p1k1b3/p2pPp2/P2P1Pp1/bP1NK1P1/N7/8 b - - 0 64

Position at move 64: Stockfish persists in its assessment, while Komodo already sees some edge for white, though not big.

I think this is an excellent example of the advantage of the 2 knights: the 2 knights will actually be stronger than 2 minors and 2 bishops, whenever there are a large number of fixed pawns, and here this is exactly the case, as out of 6 pairs of pawns 4 or 5 (depending on how you consider b) are fixed. I think for each pair of fixed pawns a knight should increase its nominal value by some 5%. 4 pairs of fixed pawns would increase thus the knights' value by some 20%, and if a knight is scored some 3 pawns, that would make some additional overvalue of some +60cps.

In this case, in most closed positions with a large number of fixed pawns knights are actually stronger than bishops.

It is also not clear, in terms of eval, if f7 and f5 are only double, or double horizontally isolated pawns, which would increase the penalty tremendously. Obviously, g4 makes the f5 pawn part of a group, and thus it is not horizontally isolated, but this is so just at first glance. Actually, I think f7 and f5 are double horizontally isolated pawns, because the most advanced of those pawns is fixed and the only friendly pawn that makes it part of a group is placed on a square in front of this pawn vertically, meaning horizontal isolation/lack of support/structural deficiency is felt very strongly. In any case, those double pawns would be due at least a large portion of the standard one for double horizontally isolated pawns.

[d]8/4bp2/p1k1b3/p2pPp2/P2P1Pp1/1PNNK1P1/8/8 b - - 0 65

The position at move 65: I think black should try now 65...f6, to activate its bishops even at the cost of a pawn (for example d5, by playing d5-d4), maybe this could still provide a draw, but maybe not.

[d]8/5p2/4b3/1k1pPp2/3P1Pp1/1K4P1/8/3N4 b - - 0 109

The position at move 109: black is hopeless; f7, f5, d5 are all major weaknesses.

I hope that after this game the 2 knights advantage will enjoy a bit more popularity among fellow forumers

Any comments, whistle-blowing, etc., very much appreciated.

Best, Lyudmil