how learning will solve 1.g4 as black ?

[deeds]

After 2 000 games :



Draw rate : 18%
Black win rate : 90.8%

[deeds]

After 12 000 games :



Draw rate : 15% (-3%)
Black win rate : 92.6% (+1.8%)

[deeds]

After 46 000 games :



Draw rate : 10% (-5%)
Black win rate : 95.2% (+2.6%)

[deeds]

After 88 000 games :



Draw rate : 9% (-1%)
Black win rate : 95.4% (+0.2%)

[dorsz]

Interesting. How long did it take?

Would learning for white increase draws significantly?

[deeds]

Interesting. How long did it take ?

I confess that I didn't try too hard to move forward quickly in this project because Eman had other openings to learn.

Since I want to get a (perfectly) balanced experience file (distribution of trained openings), from "1.g4" it produced too much "ECO A" openings...

But otherwise the games were mainly played at TC 1m+1s @ 7-8 threads so depending on how many openings were trained at the same time, it can go quickly :

Would learning for white increase draws significantly?

I doubt !
As you know, the "key" positions with a "high draw rate for white" are very rare after "1.g4".
Each time I found one of them, i trained Eman and after 2000 games the "draw rate for white" had always dropped.
Even with an odd tc like 1m+1s vs 5m+5s, "Eman 1m+1s" managed to win some games after "1.g4".

[dorsz]

A lot of work.