HAL9000

HAL9000
"It just isn't conceivable that you can design a program strong enough to beat players like me."

February 5, 2015

RAPIDROID: Restarting the whole thing

I've just tried to find a reasonable and ergonomic method to embed all those Jim Ablett compiles, updated and new engines, into my existing database of games.

Unfortunately, rewinding all the film back and forward seems to be much more painful and time consuming than repeating all the rounds.

Last published list had 56 engines of which today:
* 16 need to be replaced by their newer versions
* 21 newcomers need to be introduced.

Looking at the amount of change, it's better to start from scratch. That would take less time than reworking a lot of past records. It's also an opportunity to improve the numbers of played games, keeping score % closer to 50%, by recovering the bias caused by unbalanced groupings and previous gauntlets.

By descending to my fastest acceptable time control of 10 sec/move from 15 sec/move, 77 engines x 100 games = 3500 games should cost approximately (7000 / 2 x 70 moves x 2 x 10 sec) / (60  x 60 x 24) = 56 full days with non-stop tournaments.

With 80% time efficiency and two devices, i expect a new release in 35 days, on 12-March-2015.

So I sing "Don't cry for me Enginetina !!!"

2 comments:

AAR said...

Great idea my friend.....good luck!
No telling how many engines Jim Ablett will push out...when he released Akkad 0.44b he had me scratching my head and searching the archives!

Unknown said...

First round of 450 games by 77 engines almost coming to an end after many interrupted tournaments, lots of replayed and cancelled games, regroupings, banned engines due to hanging the gui. All of this happening under heavy engine storm by JA. But i can cope with him so far :-)