Install the self-extracting archive into a folder of your choice.
To run Rebel TNG you will need to announce the engine first.
Under Chessbase this means: Engine -> New UCI engine -> Browse -> RebelUCI.exe
Under ARENA do: Engines -> New Engine -> UCI -> RebelUCI.exe
Under ChessPartner do: Extra -> Engine Import Wizard -> Winboard -> Install -> Rebel.exe
ARENA is freeware and can be downloaded here.
REBEL TNG at
A new paradigm in computer chess
Release Notes version 1.0
Since the release of Rebel Million (1997) I got fascinated by the enormous amount of human chess knowledge inside these databases and since then invested massive efforts (and computer time) to extract patterns for the use in the evaluation function. Through the years I tried various systems but no real progress was established.
Instead (as a sort of consolation for the efforts invested) I was only able to gain some 20-30 elo via the EOC approach [ 1 ] [ 2 ] and lately via a giant extented opening book in ProDeo version 1.84.
However because of (or better thanks to) the release of the last ProDeo version (1.88) the light finally broke, the key to pattern recognition is not in human-human games but in commented comp-comp games that contain depths and scores.
From the games of CCRL | CEGT | FCP | SF testing | Gladiators chess and other sources (in total 4.1 million games) I was able to build a database of patterns currently containing 2019 patterns of the opening, middle and endgame that took about 5 days to built.
Rebel TNG is estimated 250+ elo stronger due to this new algorithm (see the result page) and there is still potential. Much will depend on the quality, volume and variation of the comp-comp database. A logical next step (once the system is perfected) is to consider reductions via pattern recognition.
In other words there (still) is lot of work to be done. However, it's not my intention to return into the arena of computer chess, I retired in 2003 and ever since I enjoy my life too much to return into the stressful world of competition.
Instead I plan to document the system and present a paper how it is done including source code.
In the meantime enjoy the first try.