COMPLETED: The Alternative Unofficial UHO 119th Amateur Series Division 1

[Modern Times]

Graham's tournament is currently sitting at a draw rate of about 93%. I thought I'd run this as a comparison, to see what effect UHO openings would have. I ran reversed sides per pairing so there are double the number of games here. That was probably not strictly necessary.

Whether this is better, worse, or just different is personal choice.

Results with UHO were:

Total: 728 games played by 14 programs

White wins: 303 (41.6%)
Black wins: 1 (0.1%)
Draws: 424 (58.2%)
White score: 70.7%

Engine Scores:

Code: Select all

   # PLAYER                  :  POINTS  PLAYED   (%)    W    D    L  D(%)  
   1 Stockfish 17.1          :    72.0     104  69.2   46   52    6  50.0  
   2 Obsidian 16             :    66.5     104  63.9   36   61    7  58.7  
   3 PlentyChess 6.0.0       :    62.0     104  59.6   32   60   12  57.7  
   4 Berserk 13              :    58.5     104  56.3   27   63   14  60.6  
   5 Integral V7             :    55.5     104  53.4   28   55   21  52.9  
   6 Alexandria 8.0.0        :    54.0     104  51.9   21   66   17  63.5  
   7 Caissa 1.22             :    54.0     104  51.9   21   66   17  63.5  
   8 Stormphrax 7.0.0        :    49.5     104  47.6   17   65   22  62.5  
   9 Viridithas 17.0.0       :    48.0     104  46.2   16   64   24  61.5  
  10 Horsie 1.1              :    47.0     104  45.2   16   62   26  59.6  
  11 Starzix 6.0             :    44.0     104  42.3   13   62   29  59.6  
  12 Motor 0.9.0             :    42.5     104  40.9   12   61   31  58.7  
  13 Clover 8.1              :    38.5     104  37.0   10   57   37  54.8  
  14 Quanticade Orion 2.0    :    36.0     104  34.6    9   54   41  51.9  

Conditions:
Hardware: Intel i9 10900 and AMD Ryzen 5900X
GUI: CuteChess, concurrency limited to number of actual cores, Ponder Off
Tournament Format: Round-robin, 4 cycles 52 rounds, each round with reversed sides, so 364 x 2 = 728 games
Book: Stefan Pohl 6-move UHO_2024_6mvs_+085_+094.pgn
Time Control: game in 15 minutes plus 10 second increment, 15'+10" (actual)
Hash: All engine 512MB
Threads: All engines 1 thread
Endgame tablebases: Syzygy 5-men
Adjudication: GUI ends game for 5-piece endgames, otherwise games are played until mate or draw by 3-fold rep, 50-moves, stalemate

[Graham Banks]

You could always come back to CCRL and run a separate UHO list for the top 15 engines.

[pohl4711]

Graham's tournament is currently sitting at a draw rate of about 93%. I thought I'd run this as a comparison, to see what effect UHO openings would have. I ran reversed sides per pairing so there are double the number of games here. That was probably not strictly necessary.

Whether this is better, worse, or just different is personal choice.

Results with UHO were:

Total: 728 games played by 14 programs

White wins: 303 (41.6%)
Black wins: 1 (0.1%)
Draws: 424 (58.2%)
White score: 70.7%

Quite good stats, but not perfect: My UHO openings allow it to adjust the stats:

Best UHO-stats mean (this is the UHO-concept): Get whites wins as close as possible to 50% and the draw-ratio as close as possible to 50%, too.
Here we see a draw-ratio of 58.2% and a white-wins-ratio of 41.6%. So, to improve the stats, UHO offers the opportunity to choose an UHO-opening-set with a higher eval-advantage for white in the endpositions of the opening-lines.
So, if you choose the next UHO-eval-stage: UHO_2024_6mvs_+090_+099.pgn instead of UHO_2024_6mvs_+085_+094.pgn. This should lower the draw-ratio around -5%.

Following my UHO-concept, a perfect stats ratinglist should look like my UHO-Top15 Ratinglist:
Games : 120000 (finished)
White Wins : 59273 (49.4%)
Black Wins : 1160 (1.0%)
Draws : 59567 (49.6%)

[Modern Times]

Yes you provide a lot of options with your UHO books. Great work, thanks. The longer time control may or may not be influencing the draw rate as well. This is likely just a one-off, but if I did something similar again I'll give the choice of book some further though.

[pohl4711]

Yes, draw-ratio depends on a lot of things: Participating engines, thinking-time, hardware-speed.

But my UHO 2024 openings offer 8 layers, means 8 increasing eval-advantage "levels" or tiers for white. So, I am convinced, UHO 2024 will offer a proper working opening-set (giving draw-ratio around 50%) now and in the future for all computerchess-events and -tests.

And my other AntiDraw openings (and older UHO releases) were done in the same way: Several opening-sets with increasing eval-advantage for white. So, these openings should work fine now and in the future, too.

https://www.sp-cc.de/anti-draw-openings.htm

[Modern Times]

You could always come back to CCRL and run a separate UHO list for the top 15 engines.

Even if I was interested in doing another Top15 revolving round-robin, I don't need CCRL for that. I can do it myself as I am doing quite successfully with Chess324.