Huge TheKing Test of the performance of different Selectivity settings in the
endgame

Each Selectivity setting (s06, s09, s12, s16 and s20) of TheKing (Normal native
settings (all material- and positional parameters = 100)) played 2100 games 
(300 games vs. each of the 7 opponents).
Used opening-set was endgames.epd, an epd-collection of endgame-positions, which
is used for testing in the Stockfish-Framework.
https://github.com/official-stockfish/books/blob/master/endgames.epd.zip


TheKing 3.50 PC-engine thinkingtime: 15 sec + 1.5 sec (around 20x faster on my PC 
than the TheKing Chesscomputer, so on Chesscomputer around 5 min + 30 sec 
thinkingtime).

Opponents thinkingtime: 10 sec + 1 sec (Because King had only 512 Kbyte Hash 
(because the King-Element has only somewhat over 300 Kbyte Hashtables) and the 
7 opponent engines had 128MByte Hash, the opponent engines got 1/3 less 
thinkingtime, than TheKing.)

     Program                 Elo    +    -   Games   Score   Av.Op.  Draws

   1 Delfi 5.4             : 2581   12   12  1500    57.4 %   2529   46.3 %
   2 TheKing Normal s20    : 2558   10   10  2100    53.4 %   2534   45.3 %
   3 Rebel 13              : 2547   12   12  1500    52.6 %   2529   46.2 %
   4 TheKing Normal s16    : 2545   10   10  2100    51.5 %   2534   46.0 %
   5 K2 0.95               : 2543   11   11  1500    52.0 %   2529   46.7 %
   6 Gandalf 7             : 2542   12   12  1500    51.9 %   2529   48.2 %
   7 TheKing Normal s12    : 2532   10   10  2100    49.8 %   2534   47.7 %
   8 Ruffian Leiden        : 2524   11   11  1500    49.4 %   2529   43.7 %
   9 TheKing Normal s09    : 2513   10   10  2100    47.0 %   2534   44.6 %
  10 RedQueen 1.1.98       : 2507   12   12  1500    46.9 %   2529   42.8 %
  11 TheKing Normal s06    : 2497   10   10  2100    44.7 %   2534   43.9 %
  12 Orion 0.6             : 2494   12   12  1500    45.1 %   2529   44.7 %

Games        : 10500 (finished)

White Wins   : 3181 (30.3 %)
Black Wins   : 2542 (24.2 %)
Draws        : 4777 (45.5 %)


Conclusions:
It is very clear, that the performance of TheKing is better, the higher the 
Selectivity is set. Selectivity 20 scored +61 Elo (!) better in the endgame, 
than Selectivity 6 (which was default in the old TheKing Chesscomputers (Tasc R30,
Montreux, Risc 2500).
So, it is very clear, that the "AUTO" (default) setting of the Selectivity in the
TheKing Chesscomputers is complete nonsense, because Auto-Selectivity means, 
the Selectivity gets lower and lower, the lower the number of pieces on the board 
gets.
So, it is strongly recommended, to set the Selectivity at least to 16, better 20 in 
each user-setting, otherwise a lot of Elo-performance is lost in the endgame!!!

(The played games of this test-tournaments are included in the games-download of the
TrS-settings testruns games download on this site).