Great responses, thanks. Of course I agree completely -- my earlier comment about sitting back and watching the opera was just facetious! From the point of view of chess principle and of learning as well as of analytical accuracy it is surely foolish to remain passive when analysing with engines, rather than forcing them down key lines which suggest themselves to our human intelligence, and also ones which the engines themselves suggest but don't necessarily 'play'.

Another thing that's surely important, though, is to try to be objective. I think it's very easy sometimes to slip in this respect, especially when we're looking with an engine at some lines of an opening that we're keen on. The temptation is great to look at all sorts of 'reasonable tries' for one colour in such an opening hoping to bolster it against engine 'assaults', while failing to look equally conscientiously at tries for the 'opponent' ...

Smyslov_Fan wrote on 11/26/12 at 05:26:05:


IMO this is an extremely important technique...going down the "critical line" with the engine.

Thanks again. Perhaps I should treat myself to a Christmas present ...

Thank you Vass! This is very interesting ...

The question I guess is, what should we do in consequence of it? Perhaps both kinds of analysis!? -- perhaps we should do the infinite analysis, then switch to 'demo' mode, make a cuppa and sit back and enjoy the opera!

Or maybe best of all is to choose infinite analysis, then, after the engine has had a very long calculate, get it to make its first move on the board and then set up infinite analysis again, and so on!

As a refinement of (or sub-question of) this question:

If I am analysing with say Houdini or some other engine in a GUI like Arena or WinBoard, is there any difference in quality of analysis between an engine vs engine setting (where the engine plays itself at a given time limit and so moves occur on the board) and an 'infinite' mode where no moves happen on the board and one draws conclusions from the engine's top analysis line? Of course time must be a factor, so just for a silly example to make the point: suppose I have 60 minutes and I want to analyse a position 10 moves (20 half moves) deep. If I get the engine to play itself @ 3 minutes a move, should I expect to see the same position on the GUI's chessboard as I would if I chose 'infinite' mode and then read off the top line of analysis after 60 minutes? If not, which process if either would produce the superior analysis, and why?

Maybe you need this one: (I don't have it and I didn't read it, but I looked at some reviews..)

"Secrets of Practical Chess (New Enlarged Edition)" - John Nunn
Paperback: 254 pages
Publisher: Gambit Publications; 2nd Enlarged edition (April 30, 2007)
Language: English
ISBN-10: 1904600700
ISBN-13: 978-1904600701
This new “enlarged” edition of Secrets of Practical Chess contains 80 new pages. While the first edition came in at a svelte 176 pages, this one is a solid 256. Here are the contents:

1 At the Board
2 The Opening
3 The Middlegame
4 The Endgame
5 Using a Computer
6 Chess Literature

"You should know that the first four chapters in both books are identical. The added material is in a greatly expanded chapter five on Using a Computer and a new chapter six on Chess Literature."

Well, so...Chapter 5.. Have a good luck! 
And we can answer your questions...if any left.. 

Edit: Ups! I've found another Gambit piblication:
Christian Kongsted - "How To Use Computers To Improve Your Chess" 2003 (a bit outdated as it seems, but no doubt it can be useful)

Hello
      In the 1980's I had a Novag Primo. Now I have Fritz 8 and Shredder classic 4. As a serious player who is 1800ish and trying to get to the next rung of expert/2000, I use Shredder to analyse my games. There is only one problem: I don't really know how I should go about it. Sure. I put the game in, press full analysis, and get the engine's assessments and the bar graph of who was doing better/worse each move and by how many units. 

But then things get messy. How many moves should I get Shredder to analyse ahead? For how long? Does that matter? Then there is a "moving horizon" effect: the engine will say +0.90 for this move in full analysis but if I go to the move at the end of its analysis line I often get a very different assessment. 

I would like a book like that one that David Levy wrote in the 1980's - how to get the most from your chess computer. But written in a way that reflects the way that chess engines have gone and the opportunities to use them to improve. It astonishes me that I can't find such a book. Especially when correspondence chess events [ not that I play in them] often allow legal engine use. 

In my opinion there is too much "macho hairy chested my engine is 20 rating points bigger and badder than the next engine" stuff. I don't care about the chess strength. For me 2400 is as good as 3000. I just want an engine that I can USE to get better. Maybe computer AI isn't there yet. But I would like a feature that analysed all my games and created "test puzzle sets" out of them based upon my deficiencies as a player. 

David Levy needs to write a new book - "a 21st century guide to getting the most from your chess computer"