Upgrade to Chess.com Premium!

When an engine says +1.00 is that the equivalent of a pawn?


  • 19 months ago · Quote · #1

    chessteenager

    When an engine tells me +3.00 or +1.00 does that mean the equvilant of a pawn or what? What is a big advantage? When the engine says -1.00 is that negative for the person who is about to move or negative for black?

  • 19 months ago · Quote · #2

    chessteenager

    Also, i have fritz 12 does that have an engine? how do i use it? are there any good online free engines?

  • 19 months ago · Quote · #3

    Kingpatzer

    Depends on the engine, but in general, yes.

  • 19 months ago · Quote · #4

    SmyslovFan

    In general, yes. the base unit of chess evaluations is the pawn. +1 means 1 pawn. But computers and humans evaluate the same positions differently. 

    You can look up free online chess engines just as you can look anything else up. The short answer is there are plenty of world-class analysis engines available for free. And by world class, I mean used by world class GMs.

  • 19 months ago · Quote · #5

    TheGreatOogieBoogie

    No, it's the equivalent of a queen.  In fact, when it says + or - 1.00 you should feel bad for not resigning. 

  • 19 months ago · Quote · #6

    LikeTheLake

    Hi chessteenager. a)  +1.00 is exactly the equivalent of one pawn. b) For GM +1.00 is already a big advantage enough to resign in the endgame, for me even at +3.00 I sometimes blunder it. c) it means for the person is about to move.  d) Fritz 12 has an engine. e) I do not know how to use it. f) yes, several, check other posts on that, there is plenty of conversation about it.  Cheers.

  • 19 months ago · Quote · #7

    TheGreatOogieBoogie

    But seriously, yes, the pawn is the unit and -1.00 means that black is a pawn better and +1.00 means that white is.  It doesn't even have to be a material count per se but positional factors too.  Some openings like the Dutch or Najdorf even give white a whole pawn so don't trust the evaluation too much in the opening.

  • 19 months ago · Quote · #8

    Kingpatzer

    LikeTheLake wrote:

    Hi chessteenager. a)  +1.00 is exactly the equivalent of one pawn. 

    That's actually not true at all. It is an engine evaulation that says it is the equivilant of a pawn advantage in an "average" position. But what that means to each engine is not at all clear. Houdini 3 for example specifically notes :

    The engine evaluations have been carefully recalibrated so that +1.00 pawn advantage gives a 80% chance of winning the game against an equal opponent at blitz time control. At +2.00 the engine will win 95% of the time, and at +3.00 about 99% of the time. If the advantage is +0.50, expect to win nearly 50% of the time.

  • 19 months ago · Quote · #9

    LikeTheLake

    Hey Kingpatzer.  Interesting what you say.  However, first saying my comment is not true at all (--- exposing your certainty ---)  and then saying what engine does is not clear at all (---exposing your uncertainty---).  Rather than copy-pasting info I much preffer to know your opinion about it.  Mine is stated above; +1.00 is exactly the equivalent of one pawn.

  • 19 months ago · Quote · #10

    waffllemaster

    Kingpatzer wrote:
    LikeTheLake wrote:

    Hi chessteenager. a)  +1.00 is exactly the equivalent of one pawn

    That's actually not true at all. It is an engine evaulation that says it is the equivilant of a pawn advantage in an "average" position. But what that means to each engine is not at all clear. Houdini 3 for example specifically notes :

    The engine evaluations have been carefully recalibrated so that +1.00 pawn advantage gives a 80% chance of winning the game against an equal opponent at blitz time control. At +2.00 the engine will win 95% of the time, and at +3.00 about 99% of the time. If the advantage is +0.50, expect to win nearly 50% of the time.

    Read highlighted parts in sequence Tongue Out

  • 19 months ago · Quote · #11

    eboix

    LOLOLOLOLOL!

    :P

  • 19 months ago · Quote · #12

    waffllemaster

    chessteenager wrote:

    When an engine tells me +3.00 or +1.00 does that mean the equvilant of a pawn or what? What is a big advantage? When the engine says -1.00 is that negative for the person who is about to move or negative for black?

    Negative numbers mean black is ahead, positive are for white. 

    It depends on the players and type of position when considering what a "big" advantage is.  Even when the same computers play each other, the evaluation can fluctuate.  And as noted the evaluations aren't tailed for openings or endgames.  Engines work best in middlegames and they can be supplemented with an opening book and endgame tablebase to shore up those phases of the game.

    That said, you woudln't want to play an opening that's handing your opponent +1.00 as white or -1.00 as black in a serious game.

  • 19 months ago · Quote · #13

    Kingpatzer

    Clearly and obviously not the case. 

    How do I know this? Easy experimentation.

    I fire up Stockfish 2.3.1 64-bit and note that in the starting position Stockfish considers white to have a 0.24 advantage at  a depth of 22 ply.

    Now I remove the a2 pawn.

    Now, black has a 0.48 advantage at a depth of 22 ply.

    0.48 + 0.24 does not equal 1.0

    I put the a2 pawn back and remove the a7 pawn.

    Now, at 22 ply, white has a 0.88 advantage.

    Again, 0.24 + 1.0 does not equal 0.88. 

    I continue like this through the entire position and I note that there is not a single pawn one can remove from the position and have the evaluation at 22 ply be exactly 1.0 from the original starting position evaluation.

    Moreover, each individual pawn changes the evaluation at 22 ply by different amounts. Inidicating that no two pawns have the same value.

    Since no two pawns in the starting position have the same value, and no specific pawn in the starting position has the value of 1.0, it is obviously incorrect to say that "+1.0 is exactly the equivilant of one pawn."

    Indeed, my experiment is enough to show empirically that the statement is symantically flawed as pawns neither have an exact nor a static value.

    But just to further demonstrate that your point is incorrect, I further fired up Houdini 3.0 on the basis that if the pawns at least have the SAME relative value, then we could say that an evaluation of +1.0 is at least constant between engines and is in some way related to material imbalances. However, when I fire up Houdini 3.0 I see that the original position at 22 ply is  0.19 at 22 ply. This is considerably different than 0.24 at 22 ply of Stockfish. When I remove the a2 pawn for Houdini, the new evaluation is 0.05 for black compared to 0.48 for black of Stockfish. 

    This shows that not only do engines value different pawns differently, but that different engines value the same pawns differently.

    Ergo, any claim of 1.0 being "exactly equivilant of one pawn" is not only demonstrably not true with a simple experiment; it is also, in teh case of Houdini 3's team, explicitely not the case as they tuned the engine for +1.0 to relate to the winning expectation against equal opposition. 

  • 19 months ago · Quote · #14

    BabyRhinoRainbow

    All that it shows is that one engine is more accurate than another c:. This is called "the argument from disagreement" in da moral philosophies.

  • 19 months ago · Quote · #15

    Kingpatzer

    There is no pawn that has a value of 1.0, in any engine I own. 
    I've tried it now with Stockfish, Houdini, Fritz, Shredder, Fruit, and Komodo. It shows more than one engine being more accurate than another. 

    It shows that the claim of a generic pawn having a single value to engines is itself flawed. 

    Yet, somehow, we're going to say that 1.0 means "exactly 1 pawn?" Given that no engine I have access to evaluates any pawn as being worth 1.0, and no engine I have evaluates any two pawns as being equivilant, I contend that is empirical evidence that the claim is semantically flawed. 

  • 19 months ago · Quote · #16

    LikeTheLake

    Hi Kingpatzer.  The choice of removing one pawn from the starting position is arbitrary together with the choice of 22 ply analysis.  Even though you could manage by removing a bishop in some middle game position to obtain a +1.00 for the removal of a pawn would not prove I am right but that you are wrong, again, because any position of your choice would be arbitrary.  Your thorough analysis does show that different engines evaluate same positions with different values which is expected due to the different coding done by programers.  Considering the arbitrary choices in your analysis my opinion stays the same.  One pawn is exactly 1.00.  Thanks for the conversation.  Cheers.

  • 19 months ago · Quote · #17

    EscherehcsE

    My 2 cents...

    Chess engine programmers typically assign material values to the chess pieces, usually based close to either the Reinfeld or Kaufman values. So a pawn would normally have a material value of 1, a knight and bishop of 3 to 3.25, a rook of 5, and the queen 9 to 9.75. However, this isn't always the case . I've seen some engines with pawn values of 0.8, etc.

    Also, the engine evaluation you see in your GUI is not just the material value. It also includes some evaluation values based on other positional or chess knowledge factors.

     

    And I wasn't aware of Houdart's statement concerning pawn valuation. My interpretation of his statement is that he's added some (positive or negative) offset constant to the evaluation to achieve certain winning chances at certain evaluation values. So in Houdini's case, his evaluation term probably includes material values (maybe with a pawn value of one, maybe not), positional values, and an offset constant.

  • 19 months ago · Quote · #18

    TBentley

    For some engines or GUIs (Arena allows you to choose), +1 means white has an advantage, and for some, +1 means the player to move has an advantage. Just make a move and see if it changes to find out which.

  • 19 months ago · Quote · #19

    waffllemaster

    Kingpatzer wrote:

    Clearly and obviously not the case. 

    How do I know this? Easy experimentation.

    I fire up Stockfish 2.3.1 64-bit and note that in the starting position Stockfish considers white to have a 0.24 advantage at  a depth of 22 ply.

    Now I remove the a2 pawn.

    Now, black has a 0.48 advantage at a depth of 22 ply.

    0.48 + 0.24 does not equal 1.0

    I put the a2 pawn back and remove the a7 pawn.

    Now, at 22 ply, white has a 0.88 advantage.

    Again, 0.24 + 1.0 does not equal 0.88. 

    I continue like this through the entire position and I note that there is not a single pawn one can remove from the position and have the evaluation at 22 ply be exactly 1.0 from the original starting position evaluation.

    Moreover, each individual pawn changes the evaluation at 22 ply by different amounts. Inidicating that no two pawns have the same value.

    Since no two pawns in the starting position have the same value, and no specific pawn in the starting position has the value of 1.0, it is obviously incorrect to say that "+1.0 is exactly the equivilant of one pawn."

    Indeed, my experiment is enough to show empirically that the statement is symantically flawed as pawns neither have an exact nor a static value.

    But just to further demonstrate that your point is incorrect, I further fired up Houdini 3.0 on the basis that if the pawns at least have the SAME relative value, then we could say that an evaluation of +1.0 is at least constant between engines and is in some way related to material imbalances. However, when I fire up Houdini 3.0 I see that the original position at 22 ply is  0.19 at 22 ply. This is considerably different than 0.24 at 22 ply of Stockfish. When I remove the a2 pawn for Houdini, the new evaluation is 0.05 for black compared to 0.48 for black of Stockfish. 

    This shows that not only do engines value different pawns differently, but that different engines value the same pawns differently.

    Ergo, any claim of 1.0 being "exactly equivilant of one pawn" is not only demonstrably not true with a simple experiment; it is also, in teh case of Houdini 3's team, explicitely not the case as they tuned the engine for +1.0 to relate to the winning expectation against equal opposition. 

    This is not much of a surprise.  We see evaluations fluctuate every day where no material is taken off.  If all engines did was add up the pieces then no one would bother using them.  Even my human evaluation doesn't assign each pawn the same value and for example a rook in the starting position certainly isn't 5 in my mind.

    It means "everything else being equal, 1.00 is the value of 1 pawn"

    But "everything else being equal" rarely (if ever) happens.  For example in almost any position removing a pawn will gain you some mobility.  We use engines because they quickly add all these other factors (and augment them by calculating tons of lines).  The standard of 1 = a pawn equivalent is used everywhere.

    However I will agree that this and the relative values are an arbitrary evaluation.  The only real evaluations are winning for white, winning for black, and draw.  Houdini trying to define it more clearly as 1.00 gives a certain chance of winning is better... but still how did they calculate this percentage?  And how useful can this information really be?  Different people play different advantages better or worse.  They best they could give are computer statistics.

    In the end, it's easier to tell a novice that an evaluation of +1.00 means white is better by the equivalent of a pawn.  You could get into how even pawns aren't worth 1 pawn because chess is more than an "add up the pieces" game... but that's probably more information than the novice was asking for Tongue Out

  • 19 months ago · Quote · #20

    TBentley

    According to http://www.open-chess.org/viewtopic.php?f=3&t=1667, in Houdini 1.5, the base value of a pawn is 0.93, with quite a few other factors.


Back to Top

Post your reply: