Saturday, March 10, 2018

Jerome Gambit: The Science of the Draw?! (Part 2)


[continued from previous post]




Part 1, the last post, ended with my computer assistant, Stockfish 9, in a rage at foolish human behavior...

In the above position I captured Black's Knight on d5. The silicon choice for White was, instead, 17.Qh3+!?, preferring to capture the Bishop after 17...Kb8 with 18.Rxf7. This doesn't gain a pawn immediately, as Black has 18...Nb4 19.Nc3 Nc2 20.Rb1 Qd4+ 21.Kh1 Qxc4. 



Analysis diagram





Stockfish 9 now sees White as almost 4 pawns better - it took me a while to see that it does not think that Black's Knight on c2 will escape. Yet the computer rates lines without the mischief-making 18...Nb4 followed by 19...Nc2 as even worse.

By the way, in the analysis diagram the Knight at e7 is not hanging, due to the threat of checkmate at f1. White's best is 22.e6!? - but then I am back trying to figure out why the first player is "clearly" winning.

Oh, well, back to the game. From the first diagram on this page:

perrypawnpusher - Krisstianes_017

"Italian Battleground", Chess.com, 2018

17.cxd5 Bxd5 18.Nc3 Bc6 



White's development is better, but he has that nagging pawn on e5 that looks more weak than strong. Stockfish 9 sees an advantage for the first player after 19.Qh3+!? Kb8 20.Be3 b6 21.e6!? but, during the game, I didn't see it - and began to look for a draw. Clearly a failure of nerve!

Consider that: the "psychology" of the Jerome Gambit had been turned on its head. Again, props to Krisstianes_017.

19.Qg5 Ng6 

My opponent is no fool. His King will be safer after an exchange of Queens, even if his Rook is still trapped at a8.

20.Qxd8+

This is silly, but according to my plan.

20...Rxd8 21.Bg5 Rf8 22.Rad1 Rxf1+ 23.Kxf1 b6 24.Rd8+ Kb7 25.Rxa8 Kxa8

By dint of brute force (and ignorance) I have turned a won position into a (probably) drawn one. My next goal was to get rid of the pesky Knights.

26.e6 Kb7 27.Ne2 h6 28.Bd2 Bd5 29.Nf4 Bc4+ 30.Kf2 Nxf4 31.Bxf4 Bxe6 

My opponent has reached a pawn-up endgame, and was pleased. I could tell, because in a few moves I offered a draw, and he declined.

32.a3 c5 33.h4 Kc6 34.Be5 g6 35.Ke3 Kd5 36.Bg7 h5 37.Kd3 Bf5+ 38.Kc3 b5 

Black can advance his pawn majority, but White's King and Bishop (which can operate on the intersecting a1-h8 and b8-h2 diagonals) have the blockades in place.

39.g3 a5 40.Bf8 b4+ 41.axb4 axb4+ 


Or 41...cxb4+ 42.Kb3 Kc6 43.Ka2 Be6+ 44.b3 and the game will still be drawn.

42.Kd2 Kd4 43.Be7 b3 Drawn (my opponent offered)


2 comments:

  1. an ordinary chessplayerMarch 11, 2018 at 12:02 PM

    Stockfish 9 has a contempt factor of 0.2 pawns, pointed out by brabo on chesspub forum.

    ReplyDelete
  2. Readers,

    With almost a dozen years' worth of blog posts - nearly 2,550 - covering an oft-refuted chess opening, I was surprised to find the word "contempt" showed up only 5 times.

    Computer chess programs evaluate a position and assign a number to it, usually in pawn equivalents, giving a "+" or "-" to an advantage to White or Black.

    Similar to human players, programs would "prefer" not to draw with a much weaker opponent, but would be fine doing so with a much stronger opponent. Thus, programmers add a "contempt factor" (a numerical value) to shift a position's numerical assessment, taking into account the opponent's weakness/strength. (i.e. I have an even position against this guy, but I beat him all the time, no way I want a draw, so I will see myself as better.)

    As an ordinary chessplayer points out, there are some interesting discussions on the internet about contempt factors, with some posters suggesting that for analysis purposes the engine's factor should be set to 0, with others suggesting to leave the number alone (and still others suggesting toggling one way or the other, depending which side is being analyzed; although apparently this brings up concerns with hashtables - I don't want to go there).

    That's about as much as I have been able to grasp, but I have tried setting Stockfish 9's comtempt factor to 0, to see if it makes a noticeable difference. As long as it doesn't start singing "Daisy, Daisy"...

    Rick

    ReplyDelete