chess free

Chess Free

Share and you will discover!
Wild name, real passions
[ Sign up | Log in | Guest ] (beta)
alberlie 89 ( +1 | -1 )
what qualifies as a good move? Sometimes I let Fritz analyze a game of mine. I usually give the machine quite some time (total usually about an hour per game), so I guess the resulting analysis is more or less correct. However, I don't really know how to adjust the "threshold" on which it depends whether fritz gives an alternative line or not. If I use a very low value (like 0,05 pawn-units), I usually end up with an alternative for just about every move :o). But surely, an advantage of 0,05 in terms of Fritz-Evaluation would seem to the human player as just more or less the same.
So that's my question - what would be a value that to give me only alternative lines that were a real, recognizable and understandable improvement of the position? 0,2 pawns? 0,3? 0,5?

Or, asked the other way around: Is a move, that is - according to Fritz - just 0,05 pawns off the mark a good move?
wschmidt 172 ( +1 | -1 )
alberlie, I've experimented with this quite a lot lately on Gameknot games that I've just finished. Clearly, unless you're playing exactly like Fritz, something like 0.05 is too low as you end up with variations on nearly everything that isn't forced. For quite awhile I was using 0.30 as my threshold. This represents about 1/3 of a pawn, or the equivalent of one tempo as some folks think of it. This provides a pretty good balance between too much information for meaningful study while still hitting most of the main points that are useful at our level.

Using 0.30 I still found myself going through the analyzed game and hitting points that weren't notated and wondering what Fritz thought was appropriate there. I would then take the time to do an analysis at that point for a couple of minutes and paste it as a variation into the game. So, in the last few weeks, to lessen the need for that, I've scaled the threshold back to 0.20. That picks up a few of the positional niceties that I was otherwise missing and I figure that, at my level, if I'm playing with 0.20 points of Fritz, I'm doing OK and I don't need to see the rest.

Another thing I've started doing is using the Blundercheck analysis mode rather than the Full Analysis mode. Set at the same threshhold, it seems to give the same analysis, but in the variation mode it gives both the score for the move you actually made and the score for the end of the line it is suggesting. Full analysis just gives the end of the line score and some bells and whistles I don't use.

I've never done the "one hour" type analysis. I tell it to analyze each move for 2 minutes and then leave it on overnight. Have you had any experience comparing the results of those two different approaches? ws
dysfl 144 ( +1 | -1 )
My usage of Fritz analysis Iím just another patzer (around 1500 in GK), so below is my personal opinion only, mainly based on subjective observation.

I use 10-second per move at threshold of 0.50. I run first with full analysis and then using the blunder check. The first is to get opening references and nice comments, and second is to get the score. Maybe I should set the full analysis to 3-second pre move. Anyway, 10 second per move will analyze one game in about 30-50 minutes. If there is a specific position I want to see alternate moves, I run depth analysis (?) using alt+F2 from that position going to level 12 Ė 14.

The main reason I use 0.50 is that I still find so many blunders in my games either my opponents or mine. The second reason is that I cannot understand the differences when I set it to below 0.15. Fritz does not give me any explanation why a move is worth 0.10 more than the other. I think 0.20 differences in the opening stage is negligible because Fritz puts too much stress to centre control and early exchanges. Also, the score in the endgame phase is sometimes just funny. When you have a pawn and a Knight and the opponent has a Bishop, it is either a win or a draw. However, Fritz tries to give different score at every move.

I donít think setting the threshold below 0.10 has any value at all, at least at around my level. I really want to see a conditional analysis in the future version of Fritz. Like analysis from move 15 to 20 only.
azaris 103 ( +1 | -1 )
Alternative approach Numerical analysis of chess games is rarely useful. When playing a game, write down all variations you calculated in-depth but didn't play in the end. It helps to add some comments about the position in the lines (like "weak light squares", "bishop pair is strong", "knight might get trapped") so you remember what the points were after the game.

Then once the game is over, go over your notes with the engine and see where your analysis had holes. Most likely it will suggest variations that you didn't see during the course of the game. It will also point out moves in your lines that would have changed the evaluation. Sometimes you will also see the engine stubbornly argue against your line until after a few moves it changes its mind and starts to agree with you.

Try to refine your thinking process and go through games as a collection of ideas, rather than trying to figure out why Be6 is 0.19 better than Nf5+.
magna68 2 ( +1 | -1 )
what qualifies as a good move? know about one: 1.e4 :-)
ccmcacollister 49 ( +1 | -1 )
IMO It doesnt lose or result in a lesser position than the potential of the prior move allowed for. Additionally, all the better if it can attack and defend at the same time ... follows the theme of the plan ... aor provides the opponent with options that include chances to err ... and none that give him the superior position, nor allow him to equalize unless it was unpreventable.
Actually, the Additions there probably represent the character of a Superior move. And the first sentence is that of a simply Good Move.