leavenfish wrote:Any 'book' that can't include my favorite Bent Larsen contribution: 1. g3 e5, 2. Bg2 d5, 3. Nf3 (often aiming for a reversed Alekhine...tempo up) doesn't get the time of day from me!
But seriously (?) very interesting project.
Wasn't Bent Larson, buddy.
That system was first (to my mind) played, via transposition by Reti against Alekhine in Baden Baden 1925 in what become one of the most beautiful games ever played.
1. If we have later more engine analysis we should add all in one Excel file. Much more easy to compare the 0,00 evals. I am thinking this will be the last step for the database, means to reject 0.0 positions from the database.
2. If you will help with analysis, it's very important that you are using the file: beta-v03_endpos-n.epd.
At first Stockfish and Komodo are cleaning the beta-v03 main database only. The beta-v01 and beta-v02 files are not longer important.
If different engines later analyzing the beta-v03_endpos-n.epd we can do a lot with all the statistics.
If you like I can added the files you created to the project files later.
Very nice that I have Dann Corbit in the boat.
1. If we have later more engine analysis we should add all in one Excel file. Much more easy to compare the 0,00 evals. I am thinking this will be the last step for the database, means to reject 0.0 positions from the database.
2. If you will help with analysis, it's very important that you are using the file: beta-v03_endpos-n.epd.
At first Stockfish and Komodo are cleaning the beta-v03 main database only. The beta-v01 and beta-v02 files are not longer important.
If different engines later analyzing the beta-v03_endpos-n.epd we can do a lot with all the statistics.
If you like I can added the files you created to the project files later.
Very nice that I have Dann Corbit in the boat.
Best
Frank
I used beta-v03_endpos-n.epd
I have every position analyzed. Most are 30 plies deep or more, but a few are less than 30 plies.
90% of the analysis is Stockfish.
8% of what is left is Komodo
The last 2% is Houdini.
Unfortunately, I cannot tell you which is which for the EPD records because I did not design the database to store the engine details.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Tdunbug wrote:How the hell did you get through the database so quickly! I am not even a quarter the way through!
I have an army of machine.
I have most commonly played positions already analyzed (I have about a billion analyzed positions).
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Tdunbug wrote:How the hell did you get through the database so quickly! I am not even a quarter the way through!
I have couple of machines but only analyse it on 16 core server. If I left it 1min per thread, 32 threads, 26k positions, that would be less than 820 minutes or less than 14h. There were like 56 hours (4 times 14h) between Frank's and Dann's post.
Tdunbug wrote:so is it recommended to analysis pre thread or full cpu power?
If I am going to use a position for an opening book, I analyze to 36 plies (at least).
The more threads you use, the faster you get there.
It is sometimes good to do it in layers, so that you find the really awful ones more quickly and do not waste time on them.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
