EAS - for Stefan

The new exe works, regardless of name. You can call it somu.

When you said "Added game number in the display as progress indicator, example game 33 of 10000" I thought it was in substitution of the detailed output, my wrong interpretation. Still, there should be a verbose/silent option, where the former does what does now, and the later just have a progress indicator.

My intention is to use a short book (2-4 moves) where the lines are thoroughly checked. Better to spend one minute analyzing a position once, than a fraction of a second thousands of times. That book won't be the result of any blunder check, but a self-generated, Lc0 minimaxed tree,

OK, speed test finished. I analyzed with Aquarium and in 3 hours it had only gone trough the first 538 games. Then I took 544, split them among 16 files and run them concurrently in somu. Time to completion? 3-4 minutes, closer to 3. Yep, that's right, almost 60 to 1.

Ozymandias wrote:

The new exe works, regardless of name. You can call it somu.

When you said "Added game number in the display as progress indicator, example game 33 of 10000" I thought it was in substitution of the detailed output, my wrong interpretation. Still, there should be a verbose/silent option, where the former does what does now, and the later just have a progress indicator.

I don't understand, maybe you can rephrase. As it is now you can press escape or click with the mouse in the window and the process will pause and you can see how far the process is and press enter to continue.

Quote :

My intention is to use a short book (2-4 moves) where the lines are thoroughly checked. Better to spend one minute analyzing a position once, than a fraction of a second thousands of times. That book won't be the result of any blunder check, but a self-generated, Lc0 minimaxed tree,

I think I have a better idea, but you decide. I remember you said you run the analysis on "depth". If you tell me the depth + analysis engine I can analyze with that depth and engine a book of your choice and everything is in sync.

Also I think it's not wise to compare Lc0 scores with SF scores. Which is more or less the current problem, books I analyzed at the time with SF12 (likely HCE!) at depths >20, meaning incompatible.

I mean that if you don't minimize the somu window, the lines scroll like crazy, that's too much overhead. If, instead, you had a black window, where the only thing that changes is the game counter, it would run smoother.

For earlier moves, Lc0 is better. At least the bigger, newer nets at a high depth. Of course, if you only analyze the moves in the book, you'll finish earlier, but then you may miss playable moves which have never been tried. Hence a self-generated tree.

Ozymandias wrote:

I mean that if you don't minimize the somu window, the lines scroll like crazy, that's too much overhead. If, instead, you had a black window, where the only thing that changes is the game counter, it would run smoother.

Nothing I can do about that.

Ozymandias wrote:

For earlier moves, Lc0 is better. At least the bigger, newer nets at a high depth. Of course, if you only analyze the moves in the book, you'll finish earlier, but then you may miss playable moves which have never been tried. Hence a self-generated tree.

Analyzed ?

And how?

I don't know what tool you use, but I favor Aquarium's IdeA.

I use POLY.
https://web.archive.org/web/20230622210013/https://rebel13.nl/download/books.html

I doubt Aquarium is compatible with POLY.

For the record, I created an analyzed book (depth=11) from 2210 Kasparov games with POLY and it works fine with somu [F5] and blunder margin 50 and depth=11. On github, see for yourself.

Strictly speaking, it's not compatible with anything. I will have to look at the epd it outputs (as a byproduct) and see if I can convert it to bin, via PGN. Otherwise I'll have to go manual. Let's hope it doesn't come to that, even with a 2 move book, we're talking thousands of positions/lines,