Sorry to answer your question by another question but I'm new and don't understand this board of results of the site's computer analysis function : what do you mean by how much value you are giving away on average per move, using the value of a pawn to equal "1." At the end of the game how much would be a perfect game: 0? And every point is less good? 1 would be what? that per move you "losing" one pawn What means throwing away? Do less than the perfection?
Thanks by advance
Hermand:
If I understand your question correction, you are basically asking if 0 is perfection, or if the score can ever be negative? (In other words if you add value to a position equal to a pawn per five moves, you'd have a negative 0.20.) I do not know the answer, but my guess is that it does not go into the negative range for moves that add value. In fact, I guess since the score itself is an average, your good moves are naturally weighted against the bad so they are counted to some extent. I would guess the best you can get is zero on this. But I'm curious to know if that is right or wrong.
Maybe if you blunder a piece every 10 moves on average then the average difference would be 3/10 = 0.3. Mine was 0.17 which is equivalent to blundering a minor piece every 18 moves.
To me, an avg. dif of 0.5 per move does not necessarily mean you are giving 0.5 pieces per move, it could mean I missed taking a pawn or improving my board position. For example, in one of the games I played where it was a quite open, sophisticated, and calculation heavy board position, both me and my opponent ended up with an average difference of around 0.6. Usually I play games with around a 0.3 average difference, but I realized the reason why the average difference was so high was not because we were blundering pawns, it was because there were many times where we missed the "best move", but at a 1500-1600, especially in open positions, this can be hard to find. In closed pawn structure games, where the "best move" comes naturally since there is usually only one or two playable moves, I find that the average difference for both me and my opponent is reduced dramatically.
In conclusion, I find that average difference is relative based on the difficulty of the position, and can indicate both the loss of your own position, or missed opportunities to gain the slight upper hand, which is why this number can vary drastically from game to game.
My advice if you want to get serious about analysis - don't use that method, use the "self analysis" method where you can step thru each move and Stockfish will give evaluations of the position, with lines. You will learn so much more this way. You will see exactly what moves were mistakes (or blunders) and where you went from winning to losing or vice-versa. Seeing the lines is also good because it lets you see why Stockfish thinks a particular move is good.
It's the average loss from the best engine move. For example, if you score (relative difference to the best engine move) the following: 0.03, 0.00, 0.00, 0.04, 0.38, 0.05, 0.00, 0.00, 1.15, 0.14
then your avg is 0.18, if i understand correctly.
Off the top of my head, a consistently good player (1800+) can be expected to miss a chance for a pawn or two over the course of the whole game, but the average will be thrown off if he blunders repeatedly in time trouble. Ordinary not-quite-perfect play => -2 over 40 moves = -0.05 average per move. If he loses every third game badly ("badly" meaning three minor-piece blunders in one game) => -11 over 40 moves = -0.28 per move for those games, => -0.13 per move over all games.
If you're an 1100 player, then expect to make multiple serious blunders most games, say -15 over 40 moves => -0.375 per move average.
I haven't bothered to check these numbers against anybody's actual rating or games, because as PsYchHo_ChEsS says, this doesn't tell you much about your play and doesn't offer any practical path to improvement. Look at the actual lines to see what you missed.
Off the top of my head, a consistently good player (1800+) can be expected to miss a chance for a pawn or two over the course of the whole game, but the average will be thrown off if he blunders repeatedly in time trouble. Ordinary not-quite-perfect play => -2 over 40 moves = -0.05 average per move. If he loses every third game badly ("badly" meaning three minor-piece blunders in one game) => -11 over 40 moves = -0.28 per move for those games, => -0.13 per move over all games.
If you're an 1100 player, then expect to make multiple serious blunders most games, say -15 over 40 moves => -0.375 per move average.
I haven't bothered to check these numbers against anybody's actual rating or games, because as PsYchHo_ChEsS says, this doesn't tell you much about your play and doesn't offer any practical path to improvement. Look at the actual lines to see what you missed.
I don't think the avg. dif is based on solely on blundered material, it is also based on missed opportunities to gain material FROM your opponent.
Seeing the lines is also good because it lets you see why Stockfish thinks a particular move is good.
Most of the time when the computer points out a move, I can figure out why the move is good. The computer analysis also lets you see the lines of a good move, but it only does this sometimes...
I have been using the site's computer analysis function to review some of my games. It has a category for "avg. diff." I understand what this is, in that it tells you how much value you are giving away on average per move, using the value of a pawn to equal "1." So an avg. diff. of 0.50 would mean you are basically throwing away half a pawn per move on average through less than optimal play.
What I do not understand is what is a "good" score for "avg. diff." 0.75? 0.5? I would assume this is related somewhat to your rating, so that players rating around 1800 would tend to have an "avg. diff." in the same range, which would be somewhat lower than the expected "avg. diff." for players who rate around 1500.
Does anyone know if there is a generally recognized correlation between avg. diff and rating? Or what the expected/normal avg. diff. is for players at different ratings levels?
Thanks,
Ken