Hairetikos annotates a bot-on-bot battle..."Bots are human too?"
ChessManiac.com unrated challenge, 23.06.2006
Here's an exciting game that I think you will enjoy. It was an unofficial game/1hour challenge from my fellow "bot" WONDER+BOT to me. The game shows how humans teamed with chess engines (which is known as Advanced or Freestyle Chess) can create beautiful and instructive games. I'll also dispel some common misconceptions about chess engine analysis.
1.d4 Nf6 2.c4 g6 3.Bg5 After the game, I found some analysis by GM Anand, who called this, "A strange and provocative move," which is exactly what I was intending.
3...Ne4 4.Bf4 I just decided to keep playing in a Trompowsky fashion.
4...Nc6 This looked odd to me. Isn't Black's c6 Knight vulnerable with his center pawns still at home vs my advanced pair? Indeed, GM Elvest suggests 4...e5 instead.
More provocative play. I'm trying to slap the misplaced Knight, but probably over-extending my center so early in the game.
5...e5!? Yikes! Up to now I was just playing without my database or engines. But what to do after this surprise? It's time to turn on the engines -- literally.
6.Be3!? A common weakness of computers is their weak positional play. Because it seems so apositional to block the e pawn here, this looks like such a so-called "computer move". But it was actually my own invention! This might only work if I can fianchetto on the kingside. And as the game proceeds, it will become clear that I can't even do that. Each of the alternatives have flaws: My engines recommended 6.dxc6 exf4 as the best. However, I'd rather not trade my Bishop for a Knight with no other compensation. 6.Bg3?? is embarrassed by ...Bb4+; computers never miss such shots. And 6.Bd2 and 6.Bc1 both seem to give Black the initiative with no compensation, as I would be falling behind in development.
6...Bb4+ 7.Nd2 Nd4 8.Ngf3
Complicated! And cramped. I can't blame my engines for this; I'm only in this awkward position because of my own adventurous opening play.
8...0-0 9.a3 9.Bxd4 would open the e-file for Black 9...Nxd2 10.Nxd2 exd4; 9.Rc1 is passive and possibly even more cramping; and 9.Nxe5 invites the attack 9...Qh4 10.Nd3 Nf5.
9...Nxd2 10.Bxd2 Bc5 11.Nxe5 Now this works, because ...Nf5 is not as strong with my dark-squared Bishop no longer on e3. Therefore, I over-ruled my engines' recommended 11.e3, which I thought was positionally weak because 1) it blocks the d2B, and 2) I plan to fianchetto kingside, when the e pawn advance would leave a hole on f3. With more time to analyze, I now realize that my thinking was flawed: the d2 Bishop now belongs on c3 and I should no longer have been fixated on a kingside fianchetto! It's always risky to over-ride the engines, especially in blitz. But this was an unofficial game and I decided to keep it interesting.
11...d6 12.Nd3 This also looks like a so-called "computer move", but it protects f2, threatens Nxc5, and is better than the only real alternative. Many people naively criticize computer "style", citing moves such as this one, without appreciating that the "standard" move is simply weaker! 12.Nf3 is answered well by ...Qf6 when the indirect pressure on f2 is difficult for White to defend: 13.Bc3 Nxf3+ 14.exf3 Qh4 15.g3 Re8+.
12...Re8 13.e3 OK, now I'm depressed. I played this very reluctantly because of the e8R on the open file. I was being stubborn, based on my previous judgment about e2-e3 at move 11. But I finally had to concede that this is the only way I can get developed, even though it will still be difficult. I thus have to abandon my kingside fianchetto plan. All this over-riding and second-guessing now has me more than 10 minutes behind WONDER+BOT.
Another criticism of computers is that they play inconsistently. I disagree with that term. It's true that a computer usually can't be counted on to follow a long-term plan or strategy, and in some positions they seem truly bewildered. But in other cases, apparently "inconsistent" moves can be a great strength compared to human stubbornness! In other words, computers have the advantage of being extremely objective on every move. They have no fixations, no pride, no ego.
It looks and feels like I'm losing here, but the engines are saying I'm OK or even a bit ahead. I suspect that they just can't see far enough in this fast game to identify a specific fatal threat for Black. Neither can I, of course, but a human can sometimes judge the general long-term potential of a position better than an engine. This is known as the "horizon problem" which means an engine's vision is limited by the number of moves it has time to analyze. My engines were probably seeing about 14 half-moves deep here. So if Black's killer thrusts don't come until move 21 or later, my engine might be misjudging the position.
14.h4! What!? One of my engines wants to attack Black's kingside! It's not desperation, so it must be part of a plan. But wait a minute, didn't I just say that computers don't plan or strategize? Well, a computer can appear to be following a strategy because it identifies tactical advantages deep in specific variations. The computer doesn't think, "because Black's kingside is weak, attack it"; rather, it has just determined that a favorable position is reached deep in a particular variation that begins 14.h4. Some like to highlight this different reasoning process between computers and humans, but I think the distinction is largely unimportant. Deep tactical analysis might virtually equate to strategic play. Whether a computer reaches the decision to play 14.h4 as a result of tactical analysis, or a human decides to play it after determining that Black's kingside is weak, both have arrived at the same strong move. The game continuation will show that Black's kingside is vulnerable to a Pawn attack, but we'll also see that there is another reason behind this move.
14...Bb6 This tells me that my position is not so bad after all. WONDER+BOT can't find a winning attack either (if there is one)!
15.Bc3 The Bishop has great potential here, because Black is so weak on the kingside dark squares. But can I really abandon the protection of e3? I'm trusting my engines on this one. I'm scared. My engines are not.
15...Be4 16.h5 Qg5 White didn't need to fear 16...Nf5 17.hxg6 fxg6 18.Nf4 Bxe3 19.Ne6.
17.Kd2!? This looks desperate. I was still thinking I'm busted.
Now another reason for 14.h4 is revealed. White's position looks crazy, but this is actually a very good move. It protects e3, and also helps prevent ...Bxg2, as we shall soon see. Furthermore, White's ideas of castling are no more, so why not activate a powerful piece this way?
18...Bxg2?! It turns out that this is unsound. Perhaps White's goofy position somehow confused WONDER+BOT's engine(s). Good alternatives are hard to find. 18...c6 may be best.
19.f4 And now we see another reason for 18.Rh3: to prevent 19...Qg3.
19...Bxf1 20.fxg5 Bxh3
Here I assessed the position and tried to consider plans again. Even though the e pawn is likely lost, I think White has achieved a small advantage based on material and the strong c3 Bishop. My engines seem to agree. But how to exploit that advantage?
The GMs teach us to envision a winning position and then play to make it happen. If I can get my Knight to f6, it should be powerful, especially in combination with the c3 Bishop. Plus my Queen and Rook are poised to slide to the kingside for an attack. Seems like a simple and convincing plan. So I will watch and possibly over-ride the engines' analysis from here, to see that the moves are consistent with my long-term strategy. That is, as long as tactical threats allow it. I often play this way: the computer deals with the tactics, but I can make it play even stronger by providing a long-term plan.
21.c5! I sure didn't like this at first! I assumed it was a mindless tactic to protect my dismal e pawn. But then I saw that it's not even effective in doing that. So I took a moment to analyze 21...Bxc5 to see why the computer thinks it doesn't work: 22.Nxc5 dxc5 23.Qh1! White can safely sac the e pawn! 23...Bg4 (23...Rxe3 24.Re1 Rae8 25.Rxe3 Rxe3 26.d6! cxd6 27.Qxb7 is winning) 24.hxg6 fxg6 25.Re1 is winning. My engines expect 21...dxc5, which blocks the b6B, and much more importantly, frees up e5 for the White Knight. That's not the path to f6 I was expecting, but Ne5-d7-f6 will get it done.
21...dxc5 22.Ne5 Bg2 Allowing the Knight yet another route to Nf6. Maybe 22...Ba5!? was better, but White is still clearly ahead: 23.Nd7 Bxc3+ 24.bxc3 Rxe3 25.Nf6+ Kh8 26.Qa4.
23.Ng4 Or 23.Nd7 Rxe3 24.Nf6+ Kf8, which transposes to the game line.
23...Kf8 24.Nf6! Sacrificing the cursed e pawn once and for all. Hopefully it's a worthwhile sac because White has now achieved the desired position discussed after move 20. It's notable that I did not have to over-ride my engines' analysis to accomplish this. It's as if we had the same vision. But of course we know that the computer just appeared to be following a strategy because it sees the greatest advantage in this position. Still, that's not too different from the GMs' technique I mentioned after move 20!
24...Rxe3 25.h6! Restricts the Black King and Knight, and threatens Nxh7+ followed by Nf6+ and then h6-h7.
25...Bf3 26.Qf1 26.Nxh7+! might be even stronger; it seems to make the point more directly.
26...Re2+ 27.Kc1 Hmmm... I missed that this was the only move, which derails White's plans to sweep the Rook over to the kingside. I remember thinking during the game that Black might have enough compensation for allowing me to play my Knight to f6 after all. 27.Kd3 fails to c4+ 28.Kxc4 Ne3+.
27...Re3 28.a4! Threatens a4-a5 to nab the Bishop, and also provides an outlet for the a1 Rook. One thing I've learned from playing with and against computers is to more fully appreciate multiple purposes when evaluating the best moves. Also note that computers are very creative in finding ways to activate pieces.
28...a6 29.Nxh7+ Kg8 Worse is 29...Ke7 where the Black King will not be able to help prevent the h pawn from queening. 30.Ra3 Bxd5 31.Nf6 Be6 32.a5 (or 32.h7 ) 32...Ba7 33.h7 threatening Ng8+ followed by h7-h8=Q.
30.Nf6+ Kh8 Black is forced to endure the x-ray attack of my c3 Bishop. This discovered-check alignment (any Knight move is a discovered check on the Black King) effectively allows my Knight to dance around the board with immunity. And note how difficult it is for Black to break out of the x-ray; his King and kingside pawns can't move at all! My idea to get the Knight to f6 was even more powerful than I envisioned. 30...Kf8 is no better: 31.h7 Kg7 because there is no way for Black to avoid the terrible x-ray of the c3 Bishop. Black could have resigned here.
31.d6 a5 It's hard to find anything good for Black here, in light of the x-ray. 31...cxd6 32.Nd7+ Kh7 33.Nxb6; or 31...Rf8? 32.Nd7+.
32.Qc4 Re6 33.Qf4 Be2 34.dxc7 Rc8 35.Ra3 Ba7. If 35...Bxc7 then 36.Nd5+.
36.Rb3 The a1 Rook gets involved after all!
36...Ba6 37.Qf3 c4 38.Rxb7 Be3+ 39.Kb1 Bd4 40.Bxd4 Nxd4 41.Qd5 Nc6
Here one of my engines announced forced mate.
42.Rb8! Nxb8 43.Qd8+ Re8 44.Nxe8 Nc6 45.Qf6+ Kh7 46.Qg7# 1-0