Testing engines at fixed time per move

[gordonr]

I guess that running engine matches/tournaments with a fixed time per move isn't a common approach. And I appreciate that this doesn't test an engine's time management, which can be an important factor.

But my main interest is using an engine to analyse my completed games, usually by stepping through with interactive analysis and I'm always deciding how long to give the engine at each position. So if I'm used to pausing for approx. 30 seconds at positions I find interesting, maybe I should test engines at a fixed timed of 30 seconds per move?!

Does this sound more reasonable than using, say 40 moves in 20 mins, for my purposes? Or am I overlooking something?

[Dann Corbit]

I guess that running engine matches/tournaments with a fixed time per move isn't a common approach. And I appreciate that this doesn't test an engine's time management, which can be an important factor.

But my main interest is using an engine to analyse my completed games, usually by stepping through with interactive analysis and I'm always deciding how long to give the engine at each position. So if I'm used to pausing for approx. 30 seconds at positions I find interesting, maybe I should test engines at a fixed timed of 30 seconds per move?!

Does this sound more reasonable than using, say 40 moves in 20 mins, for my purposes? Or am I overlooking something?

40 moves in 20 minutes will play more like you do. It will spend a long time thinking on the tough problems, and move instantly on easy moves (like necessary recaptures).

I find that Nmoves/Xminutes is *much* more appealing to watch. But that's just me. Now, for analyzing a set of positions, I usually used fixed time scales.

As a suggestion, try both and do the one you like more.

[gordonr]

40 moves in 20 minutes will play more like you do.

Yes, agreed. But in terms of finding the strongest engine for analysis (not playing), isn't it desirable to remove the time management factor? I see other threads mentioning the time management of particular engines and it may have a significant effect. But for interactive analysis, it has no effect, so why test with this factor involved? I guess the majority of engine users (beyond this forum) are interested in mainly analysis capability.

And I also agree, watching "fixed time per move" is less appealing but it's the results I'm most interested in.

[CRoberson]

For me, analysis mode is used in different ways for different situations.

When I am at a chess tournament and analyzing my games between rounds, I use fixed depth. It will breeze through the easy positions and slow down on a tough one. During those analysis sessions, I am looking for in fairly shallow depth counter move sequences.