|
Engine: Tao |
Author: Bas Hamstra, Netherlands. |
|
|
Move Now |
Edit Mode |
Analyze Mode |
Opening Book |
Create book |
Ponder On/Off |
Hash Tables |
Table Bases |
|
v5.6 |
Yes |
Yes |
Yes |
Yes |
? |
Yes |
Yes |
Nalimov |
|
Draw claim |
Resign |
Force and go |
Program language |
Source available |
Learn |
XBoard |
UCI |
First release |
|
Yes |
No |
Yes |
? |
No |
No |
No |
Yes |
21-10-2000 |
|
WinBoard.ini lines: |
|
Homepage: No |
Tao 5.6 update: - Bugfixes, more stable now - Improved UCI (and Arena) support - Rewritten pawnstructure - Improved search-efficiency - Improved time management moves/time for Winboard - Modified king-safety - Note: if you have an UCI interface you might try this settings in stead of defaults: PreExt=18, CheckExt=24, MateExt=18, OneRep=18, RecapType=Normal, RecapExt=18, NullType=23. Not sure it's better, though. But it seems a good alternative in shorter time controls. Let me know what you think is better at long tc's (bas.hamstra@wxs.nl) or if you have other suggestions.
Bas Hamstra, Tao information: Tao was completely built from scratch and is not based on any other program or modelled to any other program. Apart from the standard techniques, such as the alphabeta algorithm, the use of hastables, tablebases and some other common techniques, it is 100% authentical. It is my second chessprogram (the first one had no name and reached a top rating of 2200 at FICS). Tao is about 3-4 years old and it is a bitboard program that uses PVS search without root (aspiration) window. It does not use futility pruning at any level, but it does use nullmove pruning. The structure of the program is more or less object-oriented. Tao tends to attack and is a tactical program. Weak point is the endgame. Although the endgame graduallly improves, much work still has to be done. Version history: 4.3. Low knowledge version of Tao, tuned for searching deeply, knowledge that let it search less deep was discarded if possible. I tuned on the basis of blitz games. 4.4. Not much different, slighly more efficient search 5.0. Added a lot of knowledge, especially about endgames, more advanced mobility, kingattacks and trapped pieces. This version searches less deep than 4.3/4.4 and it might score somewhat less in blitz, but at longer time controls it is stronger. 5.1 Tried to rewrite eval 5.2 Threw the new eval away and copied the old one back, added TableBase support and UCI support, tuned eval somewhat more aggressive 5.3 There appeared to be many timemanagment bugs in 5.2. Time management fixes for UCI and Winboard and other small fixes, some tuning changes in evaluation 5.4 Last details in time management fixes, I consider this topic solved now. Rewrote one kingsafety term, because I felt it was not accurate. This worked well and I tuned the value up. This worked even excellent and at the Internet server it achieved the highest rating ever. Compiled with MSVC to create a 30% faster executable VCYin54.exe. Results Testsuites: WAC 295/300 at 5 sec position LCT-II 2600 BS2830 2752 (Be welcome to mail me other results, I will put them here) Internet server: ICC standard rating 2583 (highest 2612) FICS standard rating 2453 (highest 2453) Recent computermatches 60/0 time control: Tao-Comet 11,5-9,5 Tao-Pepito 11-9 Tao-Yace 10-10 Questions/Suggestions Mailto: Bas Hamstra Have fun! Bas Hamstra. |
© COPYRIGHT 2001 - 2012 All Rights Reserved wbec-ridderkerk.nl