Post
by BertTuyt » Sat May 02, 2020 02:46
As shared before it is not easy to quickly close the gap with Scan 3.1 and Kingsrow.
As an example some recent standard match results (from yesterday and today), based upon 158 games at 1 min for 75 moves, all against Scan 3.1
* Both engines 1 core and 6p DB, Scan results 8W, 0L, 150D, 0U, ELO = 18
* Both engines 6 cores and 6p DB, Scan results 4W, 0L, 153D, 1U (did not analyze yet the Unknown), ELO = 9
* Both engines 6 cores, Damage 7p DB, Scan 6p DB, Scan results 2W, 0L, 156D, 0U, ELO = 4
So as expected (within the limited statistics), some improvement when going to higher core count and a 7P DB.
With standard Tournament settings (so > 5 min Game) and 16 Thread-ripper cores, I would be not surprised when all games (within the 158 Games set) are draw.
Bert