ThatsIt wrote:Hi to all !
Currently i've running some test tournaments from the starting position using different time controls.
A wonderful idea. All the other tests like IPON or others are wrong.
For what reason ?
Best wishes,
G.S.
IMO only these conditions are full honest for rating lists:
- without books, tablebases
- startposition (from zero)
- time per move, changed after each pair of games
- 1CPU
But the main problem will be, that you only get 2 games per match !
Far too little for a rating list.
Best wishes,
G.S.
I mean each pair of games for the same opponents. Time is changed (to avoid same games) after cycle. It is rather a technical problem.
ThatsIt wrote:Hi to all !
Currently i've running some test tournaments from the starting position using different time controls.
A wonderful idea. All the other tests like IPON or others are wrong.
For what reason ?
Best wishes,
G.S.
IMO only these conditions are full honest for rating lists:
- without books, tablebases
- startposition (from zero)
- time per move, changed after each pair of games
- 1CPU
But the main problem will be, that you only get 2 games per match !
Far too little for a rating list.
Best wishes,
G.S.
I mean each pair of games for the same opponents. Time is changed (to avoid same games) after cycle. It is rather a technical problem.
Currently i've running some test tournaments from the starting position using different time controls.
The first tournament is completed, the second is in process.
Conditions:
Startingposition with color change = 2 games per match
Books = off
Ponder = on
Core(s) = one for each engine
Intel i5-2400 @ 3.1GHz
OS = Windows 7 Professional
GUI = Shredder Classic 3
Bases = four man
Learning = disabled
That's a bizarre idea. As you mentionned, each pair would only play 2 games (due to the deterministic nature of the search algorithm). It would be very hard to produce any meaningful statistics with that.
What about reducing the bookdepth drastically instead. Just get a book that stops at depth 5, lets say, and make sure it has enough lines. That would be a reasonable compromise, and help account for the various engine's ability at handling an opening by themselves.
Theory and practice sometimes clash. And when that happens, theory loses. Every single time.
lucasart wrote:
[...snip...]
That's a bizarre idea. As you mentionned, each pair would only play 2 games (due to the deterministic nature of the search algorithm). It would be very hard to produce any meaningful statistics with that.
[...snip...]
That was never the (an) intention !
It's just for fun.
