Chess for Beginners

Tag: evaluation

  • Think Like a Chess Engine

    In Kotov’s great book “Think Like a Grandmaster” he taught the average chess player the inner workings of how a chess master thinks about a position by creating a tree of candidates and then proceeding along the tree. While there is much to learn from Kotov’s work, I have always been facinated how computers can evaluate positions and how their positional play is derived from these material evaluation algorithms. I believe that the way chess engines derive material evaluations might help the lower rated amateur improve their evaluation skills.

    Below is a combination of the Crafty chess engine algorithms as well as Larry Kaufman’s material evaluation processes which is used by the Rybka engine.

    Pawns

    Pawn = 1

    isolated pawn penalty based on file:
    a-pawn : -.12
    b-pawn : -.14
    c-pawn : -.16
    d-pawn : -.20
    e-pawn : -.20
    f-pawn : -.16
    g-pawn : -.14
    h-pawn : -.12

    Doubled pawn (and not also isolated) penalty of -.12 plus add isolated pawn penalty

    Backward pawns -.06 penalty +.04 bonus for attack on backward pawn on semi-open file.

    Pawn advance in center +.04 / rank increasing to about +.08 / rank in the endgame
    Development

    Rook pawn is worth 15% less than a regular pawn (.85 of a pawn) (L.Kaufman)

    Knights

    Knights = 3.25
    Centralized knight: +.30
    In outpost : +.08

    Unpaired knight is worth approximately 3.14 pawns (worth less) (L.Kaufman)

    Bishops

    Bishops = 3.25
    Bishop in endgame +.10
    Bad bishop : -.04
    Corner bishop : 0.0
    Center bishop: +.3
    Attacking / good bishop: +.18
    Attack against King: +.05

    The Bishop Pair

    Bishop pair: +.50  Bishop pair is worth .50 more if not pieces exist to exchange them (L.Kaufman)

    Bishop pair is worth less than .50 a pawn when most or all of the pawns are on the board, and more than .50 pawns when half or more of the pawns are gone (L.Kaufman)

    If you have the bishop pair, and your opponent’s single bishop is a bad bishop (hemmed in by his own pawns), you already have full compensation for 1 pawn (L.Kaufman)

    Unpaired bishop is worth approximately 3.14 pawns (worth less) (L.Kaufman)

    Rooks

    Rooks = 5
    Attacking on open-file: +.20
    One semi-open file: +.10
    On open-file: +.14

    Rook and Bishop is better than the Rook and Knight. (L.Kaufman)

    Queen

    Queen = 9.75

    Queen and Knight is slightly better than Queen and Bishop

    King
    Centralized in opening: -.24
    Centralized in endgame: +.36
    On open file in opening: -.23
    No adjacent pawns: -.08

    Special Cases

    Hanging pieces: -.10

    Exchanges favor the side with a material advantage

    Minor pieces lose their value as the endgame approaches

    In endgames with no other piece the bishop is worth about 2.5 and the knight 2.25 with other pieces the minors are worth about 3.25

    Rooks gain value as the endgame approaches

    Note: I have modified the Crafty point values to reflect the Kaufman values.

    Resources

    Evaluation of Material Imbalances in Chess

    GNU Chess Positional Heuristics

    http://chessprogramming.wikispaces.com/Evaluation

    http://www.chessvariants.com/d.betza/pieceval/index.html

  • Karpov & Mazukevich's Chess Thought Process

    (adapted by Herman Grooten)

    The following thought process which is recommended by Karpov and Mazukevich is used to detect the important features of an arbitrary position in order to devise a plan that conforms to the positional assessment derived by using this thought process.

    This thinking process is based on 7 criteria in which the White position is compared with the Black position.

    1. What is the material balance?

    2. Are there any (direct) threats?

    3. How is the safety of both Kings?

    4. Pawn structure questions:

       a. Where are the open lines and diagonals?

       b. Are there many strong squares?

       c. Who is controlling the center

       d. Who has more space and where on the board do they have it?

    5. Which pieces are active and which are not?

    Training applications:

    Go through the questions out loud while doing middlegame positions based on  Stoyko Exercises. The idea is to practice “talking out loud” with the list in front of you so that when you are in real game mode the questions will be asked subconsiously.

    Source: Chess Strategy for Club Players pg. 37.

  • Analysis Exercise #1


    Black to move (Stefanova – Shirov 2005)
    r3rq1k/8/p2p3p/P1pPnp1P/1p3Q2/5PNR/1PP2K2/R7 b – – 0 29
    Download game

    Evaluation
    The first thing I did was to evaluate this position, and I gave it the following evaluation: Material =, King Safety +/=, Activity =/+, Pawn Structure =. So the game in my opinon is even. (When I gave it to Fritz 10 he sees it as -+ (- 1.61 with Black winning) and Rybka sees it as -1.19).

    Threats and Opportunities
    I then looked for threats and opportunities in this position, and I found that White was threatening to win Black’s f5 pawn with Qxf5.

    Candidate Move Selection
    After performing the threat analysis I looked for candidate moves, and this is the order in which I found my 4 candidates: 1…Re7, 1…Kh7, 1…b3, 1…Ng4+. The first two candidates are defensive, trying to deal with the potential threat of Qxf5, and the last two are aggressive candidates.

    Analysis
    I then began analyzing the most aggressive candidate based on CCT (checks, captures and threats), in this case …Ng4 would be the most aggressive since it starts with check, below is my analysis:

    Candidate 1 …Ng4+

    29…Ng4+ 30.fxg4 30…fxg4 31.Qxf8+ Rxf8+ 32.Kg2 gxh3+ 33.Kxh3 This line is good for black since he wins the exchange, so taking with 2.cxb3 is not good for White so I analyzed another line in which White does not take to see if …Ng4+ stands up. [30.Kg2 Not taking keeps White in the game. 30…Ne5 coming back to the same position is not good for Black, since White’s King is now safer on g2. At this point I figured that if White didn’t take I would wind up in this position, and I stoped analyzing this line, and I jumped to the …b3 line which was the 2nd most aggressive candidate I had selected. The problem is that 30…Ne3+ should have been considered as a mainline, and I did not see this knight check, with the rook defending it. (30…Ne3+ I did not analyze this subvariation, but it was much better than 30…Ne5 which I did analyze and caused me to jump to another line. 31.Kh1 Nxc2 32.Rc1 Nd4 Black wins a pawn, and defends g5.) ; Rybka 2.3.1 32-bit : 30.Kg1 Ne3 31.Re1 Nxd5 32.Rxe8 Nxf4 33.Rxf8+ Rxf8 34.Rh2 d5 35.Kf2 Kg7 -0.91/12 ]

    Candidate # 2 …b3

    [29…b3 This is the second line I analyzed. 30.Qxf5 So, I analyzed another line other than 30. cxb3, the problem is that this line is not the best line for White. (30.cxb3 Nd3+ Taking my b3 pawn is bad because of this fork.) 30…Qxf5 31.Nxf5 bxc2 32.Rc1 (32.Ne3 Nd3+ At this point I decided to chose this line.) 32…Nd3+ ; Rybka 2.3.1 32-bit : 29…b3 30.Qd2 bxc2 31.Qxc2 Ng4+ 32.Kg1 Ne3 33.Qd2 f4 34.Ne4 Re5 -1.27/11 ]

    Lessons Learned
    My evaluation was off by a pawn, Rybka and Fritz think that Black is at least ahead one pawn, while I thought the position was even, so I need to work on my evaluation skills. I was happy with my candidate move selection, and the order in which I began analyzing (the most aggressive moves first). Out of the 4 candidates I only had to analyze 2 and both were the best moves. In the analysis department, I am missing the best response when I calculate variations or I am not analyzing all possible oppnent candidate moves. I need to use the same candidate move criteria fwhen looking for or my opponent responses, but I have a feeling it also has to do with board visualization. Even though I picked an appropriate candidate, the logic for selecting it was flawed and incomplete. The analysis for …b3 did not take into account White’s best responses and much was left to chance by selecting it. I need to be more thorough when analyzing and I also have to attempt to evaluate the position at the end of the my analysis (which I did not do). The line played in the game was …Ng4+!, while Fritz 10 and Rybka prefer …b3.

  • Faulty Thought Process: Missing the Obvious

    Why is it that beginners fail to choose simple plans that are right before their eyes? One answer may be that beginners do not evaluate the position before choosing a candidate move. Beginners briefly scan the board, choose a move they like and quickly analyze and play this move. The problem with this thought process is that 9 out of 10 times the move they have selected is not the best move, since it is not a move that follows a plan based on the needs of the position. This thought process error occurs during candidate move selection and is a very common mistake that beginner’s must overcome in order to reach the next level. The difference between a weak player and a strong player is that the strong player evaluates the position and they then choose a plan based on this evaluation.

    Dan Heisman in a Novice Nook titled ‘Evaulation Criteria’, uses the following criteria to evaluate a position (in order of importance):

    1. Material
    2. King Safety
    3. Activity
    4. Pawn Structure

    Based on these evaulation criteria, not only will we know which side stands better, but why they stand better and what our plan should be. The candidate moves will show itself based on the plan we have selected.

    In order to improve we need to evaluate the position during critical junctures of the game (after the opening is over, after a series of exchanges, whenever the position changes substantially). We should get in the habit of re-evaluating the position every couple of moves to ensure that the needs of the position have not changed. We then need to choose a plan based on the evaluation and select candidate moves that help us meet the goals of our plan. We should then take a close and honest look at our candidate moves and keep looking for the best move that will improve our position on the board.

    Hope this helps, and I would love to hear your thought process for evaluating the position and selecting candidate moves.