I am not sure if this question has been asked before, but I recently had a friend, who is a NM, play a blitz match against stockfish (level 12 I believe) and commentate its playing style. What struck me was the fact that he kept mentioning that it was making “weird moves”; moves that a human player would likely never play. I asked him to explain further, and as an example he mentioned the fact about pushing pawns only one square instead of two when it could have clearly pushed two (maybe losing tempo on purpose?). I have played against it many times, and the fact that I have an ELO of about 1750, does not really help me in finding these “weird moves”. I would probably categorize them as “very strong moves” that have some deeper/intricate scope that I would probably not be able to grasp in a blitz match. My question is, how can higher rated players notice this non-human playing style? What makes a move weird? I am not sure if “weird move” is subjective to the player (here of course I am referring primarily to +2300 ELO) or if it’s something more on the objective side? Meaning that it is agreed amongst most that a certain move is weird.
-
9Perhaps you could find room in your question for a concrete example or two?– bofCommented Apr 27, 2020 at 23:10
-
3I would assume that when your friend says that the computer makes a strange move, I'm pretty sure that they mean that they would not be able to explain the engine's decision. This may be because the move chosen by the engine is not in line with the overarching plan the engine seemingly was gunning for earlier, or it may be some sort of "anti-positional" move.– ScoungedCommented Apr 27, 2020 at 23:55
2 Answers
Spotting "weird" computer moves is not difficult. Let me give you an example which you will agree is very clear. It is taken from the last round of the Boeblingen open in December 1999. The white player, Allwermann, was a 55 year-old German player who had been about 1900 strength (according to his German rating. FIDE ratings didn't go that low in 1999) for the previous 20 years. Going into the last round he was on 6.5/8 with the prospect of a first prize of DM 1,660 if he beat his grandmaster opponent. After 30 moves Allwermann was already winning and was to move in this position. See if you can guess his next move. Here are some candidates: Rd7, Rxb7, Rxf6, Qa7. Which one would you choose? Which one would be the last one you would choose?
[title "Allwermann,Clemens - Kalinitschev,Sergey (2505)"]
[fen "r6k/1p3Rpp/p2p1bq1/3N4/2P5/1P6/P1b2QPP/5RK1 w - - 0 1"]
1. Rxf6 (1. Qa7 Rg8 2. Qxb7 Be4) gxf6 2. Qxf6 Qxf6 3. Rxf6
In the Qa7 line Black's last move, Be4, threatens mate. The only move to stop it is Nf4.
Allwermann played the Qa7 line.
Fritz 5.32 evaluated Rd7 as +2.78 and Qa7 as +2.88.
This was written up in a Chessbase article in 2011. This is what the article said of this move:
If you switch on multiple-variation mode you will see that Fritz thinks 31.Qa7 is a tenth of a pawn better than the other alternatives. A computer program simply doesn't understand the difference between cast-iron moves that cannot fail, and a tight-rope walk on the edge of the precipice. I venture no human would undertake the latter course in the given tournament situation.
That sums up the difference between normal human moves and weird computer moves. At least it did up until AlphaZero came along.
Here is the full game. See how many obvious computer moves you can spot.
[title "Allwermann,Clemens - Kalinitschev,Sergey (2505)"]
[fen ""]
1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4 e5 5.Nb5 d6 6.c4 Be7 7.Be2 a6 8.N5c3 Nf6 9.O-O Be6 10.Be3 O-O 11.Na3 Nd7 12.Qd2 Nc5 13.Nc2 f5 14.exf5 Bxf5 15.Bf3 Kh8 16.Bd5 Qe8 17.Rad1 Qg6 18.Na3 e4 19.f3 exf3 20.Bxf3 Ne5 21.Nd5 Bh4 22.Nf4 Nxf3+ 23.Rxf3 Qe8 24.Nd5 Ne6 25.Rdf1 Qg6 26.b3 Rf7 27.Nc2 Ng5 28.Bxg5 Bxg5 29.Qf2 Bxc2 30.Rxf7 Bf6 31.Qa7 Rg8 32.Qxb7 Be4 33.Nf4 Qf5 34.Qd7 Qe5 35.Kh1 g5 36.Nh3 g4 37.Nf2 Bf5 38.Nxg4 Be4 39.R7xf6 Bxg2+ 40.Kxg2 Qe4+ 41.Kh3
In the final position after his GM opponent had resigned Allwermann couldn't resist telling his opponent that it was mate in 8. Of course he was right.
-
1Great answer. From the ChessBase story, I remember Anand saying “Fritzy!” about the Qa7 move and then laughing... Commented Apr 28, 2020 at 19:45
A classical example is the first game between Deep Blue and Garry Kasparov in 1996. This game is famous because it was the first time a machine had beaten the world champion in a classical time control game.
The 29th move by Deep Blue was a shocking and "weird" computer move. Deep Blue took in b7.
Why is it a weird move?
Deep Blue grabs a pawn far from the action. A human will consider this move as a loss of time while his king is under attack. The action is going to be in the kingside. The computer took in b7 because it had calculated deeply enough to get away with it!
What's the human move?
The "human move" is Qd5, centralizing the Queen, covering the kingside and threatening to take in f7 and b7.
[title "Deep Blue - Garry Kasparov, 1996 (m/1)"]
[fen "6rk/1p3p1p/2nN1q2/2Q5/3p1p2/PP5P/5PP1/2R3K1 w - - 0 1"]
1. Nxb7!! {the "computer move"} (1. Qd5! {is the "human move" covering the kingside and threatening to take in f7 and b7} 1... f3? (1... Rg6) {is more solid} 2.Nxf7 Kg7 3.Ng5+-) 1... Ne5 2.Qd5 f3 3.g3 (3...Qf4!? {with a double attack on g3 and c1} 4.Rc8!!+- {the key move behind this computer variation}) Nd3 4.Rc7 Re8 5.Nd6 Re1+ 6.Kh2 Nxf2 {Kasparov is close to mate but...} 7.Nxf7+ Kg7 8.Ng5+ Kh6 9.Rxh7+ {1–0}
