5

Here is a short game played online at chess.com.

[FEN ""]

1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. c3 d6 5. d4 Bb6 6. dxe5

Black resigned for some unknown reason; maybe a mouse slip or he/she needs to leave. When the game was analysed at chess.com, the game accuracy of white is 99.7%. In my humble opinion, every White move is accurate (unless one considers 3. Bb5 to be more accurate than 3. Bc4). But why is the game accuracy for white is not 100%? What is the 0.3% that White l missed?

Rewan Demontay
  • 16,942
  • 4
  • 65
  • 109
Zuriel
  • 8,749
  • 6
  • 39
  • 93
  • 1
    Can you show which entity did the analysis that got 99.7% accuracy? (a link to the analysis would be best) – Allure Aug 03 '20 at 07:06
  • 6
    After reading only the title, I thought this is a Breaking Bad question. – Shadov Aug 03 '20 at 11:12
  • 1
    @Shadov what would be the breaking bad connection? – pulsar512b Aug 03 '20 at 15:39
  • 2
    @pulsar512b: The main character's name is [Walter White](https://en.wikipedia.org/wiki/Walter_White_(Breaking_Bad)). IDK what Shadov had in mind for accuracy; possibly mixing chemicals (something the character was excellent at). – Peter Cordes Aug 03 '20 at 15:49
  • 1
    @pulsar512b As above. About accuracy - it's a big thing in the show that Walter White's meth purity is, if I remember correctly, 99.1%. That is the main reason he becomes so successful - it's a lot better than the regular street-meth. – Shadov Aug 03 '20 at 18:32
  • @allure, here it is: https://www.chess.com/analysis/game/live/5238304345 – Zuriel Aug 04 '20 at 04:32
  • 99.7% accuracy means the exact same thing as 100% accuracy – David Aug 04 '20 at 06:09

3 Answers3

19

Probably, as you pointed out, Bb5 instead of Bc4.

The Ruy Lopez is considered very very marginally better/more critical than the Italian by humans, so it'd make sense that the engine used here'd agree.

Certainly, however, the Italian is fine, and the difference is very marginal.

pulsar512b
  • 1,848
  • 4
  • 12
  • 7
    Running the moves on chess.com/analysis we see that all white’s moves match the engine’s top move except 3. Bc4. It seems reasonable to assume that the engine used to give an accuracy score is the same as the one they use for their analysis board. – DongKy Aug 02 '20 at 02:49
  • 1
    Thank you for the answer! Talking about be very marginal accuracies, after 1. e4, is e5 very marginally better than c5 or c6 or e6? – Zuriel Aug 02 '20 at 03:01
  • 3
    e5 is (probably) very marginally better than e6 and c6, but it's not so clear for c5. – pulsar512b Aug 02 '20 at 04:18
  • 1
    Stockfish actually has a thing for the French. – David Miedema Aug 03 '20 at 22:46
  • Apparently, the best move for Black against 1.e4 is 1...e6. –  Aug 12 '20 at 04:03
  • And Stockfish declares that the Caro-Kann is trash. –  Aug 12 '20 at 04:03
  • I think probably just SF being weird. It loves the French for some ungodly reason. – pulsar512b Aug 12 '20 at 16:17
  • French is pretty good, and Caro-Kann gives White a small dynamic edge. –  Aug 19 '20 at 22:37
  • If you just put `1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. c3 d6 5. d4 Bb6` into the analysis, it gives White 100%. So the speculation that it's downgrading the Italian doesn't seem to hold water. – D M May 24 '21 at 09:13
6

Well, we can trust the engine in some cases when they analyse, but in this case, we are looking at book moves. Since the Italian Game is perfectly reasonable, don't worry about the 0.3% off.

1

Assuming the engine performing the analysis is Stockfish (sounds reasonable, since the engine is presumably running on CPU) then it's probable the move it considers subpar is 4.c3. Left to its own devices Stockfish will occasionally play the Giuoco Piano, but it will usually play the Giuoco Pianissimo - i.e. 4.d3 instead of 4.c3.

Indeed, the cloud analysis by Stockfish 11 at lichess.org/analysis gives, at depth 45, +0.3 for 4.d3 compared to +0.2 for 4.c3. You can also check out the SF-Lc0 bookless bonus TCEC ran in 2019. There were ten Italian games, all played by Stockfish, and of these games 9/10 continued with 4.d3.

Allure
  • 25,298
  • 1
  • 65
  • 143
  • But `1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. c3 d6 5. d4 Bb6` gives a score of 100%. So it doesn't seem like 4.c3 is the culprit. – D M May 24 '21 at 09:15