Houdini 2.0 running for the IPON

[IWB]

Ingo does not make the games available which is his right. I assume that a big part of the reason for this is that he only uses a small number of openings and does not want players tuning for that, but this is just a guess.

If you look on the site you will see that he makes a clear statement that the games are NOT available for downloading.

You are absolutly right. When playing with a book it doesnt matter much if the games are available. When using a fixed set it is a bit risky to publish them.
Some might think it is pranoid not to publish the games but ... have a look at the recent events - everything is possible in computer chess!

Another reason is, that I do some betatesting from time to time, I dont want this to be too public in advance.

Bye
Ingo

[lkaufman]

Using the Elostat averaging causes all the ratings to "contract" towards their average value, with the percentage contraction depending on the spread of the ratings of the players. This is what we observe in the present example.
I believe BayesElo handles this issue properly. However due to the use of a "prior" assumed result and perhaps also to the special treatment of draws, their ratings also contract towards the mean for entirely different reasons.

Hi Larry,

I understand very well what you say.
But can this "contraction" go as far as to explain the difference between 3016 and 3045, which is what I asked in
http://www.talkchess.com/forum/viewtopi ... 196#422196
that is:
If you look at Ingo's result
http://forum.computerschach.de/cgi-bin/ ... 1#pid41321
can you explain why Houdini's calculated Elo (3016, resulting from Elostat or Bayeselo) differs so much from the average of the individual matches Elo (the so called Perfs, at right), which I calculated to be 3045 ?

Yes, the contraction seems about right from my experience with Elostat and Bayeselo. In the case of Elostat, I fully understand why this large (maybe 15% or so) contraction happens. In the case of Bayeselo, I don't fully understand why the contraction is that large; I haven't taken the time to really fully understand all the details of Bayeselo. But since the two usually do come out very close (at least with Ingo's data and sample sizes), there is nothing suspicious here.

[Rémi Coulom]

Hi Kai,

I figure maybe you have an answer to this question, pertaining to Elo calculation:

If you look at Ingo's result
http://forum.computerschach.de/cgi-bin/ ... 1#pid41321
can you explain why Houdini's calculated Elo (3016, resulting from Elostat or Bayeselo) differs so much from the average of the individual matches Elo (the so called Perfs, at right), which I calculated to be 3045 ?

Hi took a very quick look, as Ernest asked me in another thread. Without having the exact data, and the method for producing these tables, it is difficult to say. My feeling is that the main reason is the strong non-linearity of elo as a function of win rate when the win rate gets close to 100%. For instance 95% win rate is +500 elo or so, and 50% is 0 Elo. On average, that's (95+50)/2=72.5%, so +170 points. So 500 and 0 points average to 170, not 250.

As Larry wrote, prior may have a role, too. Difficult to say. Nothing to worry about, IMO.

Rémi

[ernest]

So 500 and 0 points average to 170, not 250.

Merci Rémi,

I think I am convinced and satisfied now:
the reason of the large discrepancy in Ingo's result for Houdini (3016 vs. 3045) is due to the large number of competitors in the list, which are outclassed in the individual matches,
13 out of 21 are beaten by at least 80% score,
9 out of 21 are beaten by at least 85% score,
4 out of 21 are beaten by at least 90% score.

[michiguel]

Ingo does not make the games available which is his right. I assume that a big part of the reason for this is that he only uses a small number of openings and does not want players tuning for that, but this is just a guess.

If you look on the site you will see that he makes a clear statement that the games are NOT available for downloading.

You are absolutly right. When playing with a book it doesnt matter much if the games are available. When using a fixed set it is a bit risky to publish them.
Some might think it is pranoid not to publish the games but ... have a look at the recent events - everything is possible in computer chess!

Another reason is, that I do some betatesting from time to time, I dont want this to be too public in advance.

Bye
Ingo

Could you send me a pgn with only the results? i.e. games truncated after move one?

Miguel

[Laskos]

Hi Kai,

I figure maybe you have an answer to this question, pertaining to Elo calculation:

If you look at Ingo's result
http://forum.computerschach.de/cgi-bin/ ... 1#pid41321
can you explain why Houdini's calculated Elo (3016, resulting from Elostat or Bayeselo) differs so much from the average of the individual matches Elo (the so called Perfs, at right), which I calculated to be 3045 ?

Hi took a very quick look, as Ernest asked me in another thread. Without having the exact data, and the method for producing these tables, it is difficult to say. My feeling is that the main reason is the strong non-linearity of elo as a function of win rate when the win rate gets close to 100%. For instance 95% win rate is +500 elo or so, and 50% is 0 Elo. On average, that's (95+50)/2=72.5%, so +170 points. So 500 and 0 points average to 170, not 250.

As Larry wrote, prior may have a role, too. Difficult to say. Nothing to worry about, IMO.

Rémi

Do you take the same average scoring against the opposition calculating the performance in a gauntlet? Isn't that too simplistic, treating every match and game on equal footing? The primary values of a match are its error margins, not the number of games. Is Bayeselo the same as Elostat calculating the performance when using a general offset (I am not sure)?

Kai

[IWB]

Hi

Could you send me a pgn with only the results? i.e. games truncated after move one?

Miguel

I wrote a PM but it was easier than I thought. I added a file called "games.pgn" in my currennt download. All pairings and results. The run in question is included but will be replaced by the current one when finished.
I will add these games.pgn in the future in all downloads!

Bye
Ingo

[IWB]

Hello Ernest,

I added a file "games.pgn" with all pairings and results in the download in case someone wants to do some statistics.

Bye
Ingo

[michiguel]

Hello Ernest,

I added a file "games.pgn" with all pairings and results in the download in case someone wants to do some statistics.

Bye
Ingo

After normalizing with a spreadsheet to DS12 to 2800 as in the website , and rounding, this is the ranking with my program "ordo". The original output is at the end (normalized to an average of 2300).

Miguel

Code: Select all

    Houdini 2.0 Pro x64:	3036
           Houdini 1.5a:	3030
       Komodo64 3 SSE42:	2983
   Deep Rybka 4.1 SSE42:	2977
            Critter 1.2:	2976
           Deep Rybka 4:	2973
          Houdini 1.03a:	2971
   Komodo 2.03 DC SSE42:	2967
     Stockfish 2.1.1 JA:	2960
     Critter 1.01 SSE42:	2938
      Stockfish 2.01 JA:	2938
     Stockfish 1.9.1 JA:	2916
             Rybka 3 mp:	2916
     Critter 0.90 SSE42:	2908
     Stockfish 1.7.1 JA:	2901
            Rybka 3 32b:	2857
     Stockfish 1.6.x JA:	2842
        Komodo64 1.3 JA:	2837
               Naum 4.2:	2831
           Critter 0.80:	2823
          Komodo 1.2 JA:	2808
        Rybka 2.3.2a mp:	2804
       Deep Shredder 12:	2800
               Gull 1.2:	2794
           Critter 0.70:	2790
               Gull 1.1:	2790
               Naum 4.1:	2788
Deep Sjeng c't 2010 32b:	2785
          Komodo 1.0 JA:	2784
          Spike 1.4 32b:	2779
      Deep Fritz 12 32b:	2777
                 Naum 4:	2775
         Rybka 2.2n2 mp:	2774
              Gull 1.0a:	2766
     Stockfish 1.5.1 JA:	2761
             Rybka 1.2f:	2760
    Protector 1.4.0 x64:	2756
        spark-1.0 SSE42:	2752
           Hannibal 1.1:	2751
     HIARCS 13.2 MP 32b:	2747
           Fritz 12 32b:	2740
     HIARCS 13.1 MP 32b:	2727
       Deep Junior 12.5:	2725
      Deep Fritz 11 32b:	2720
          Doch64 1.2 JA:	2710
              spark-0.4:	2709
       Stockfish 1.4 JA:	2708
        Protector 1.4.0:	2708
        Zappa Mexico II:	2707
      Shredder Bonn 32b:	2705
           Critter 0.60:	2694
     Protector 1.3.2 JA:	2694
       Deep Shredder 11:	2685
       Doch64 09.980 JA:	2682
         Deep Junior 12:	2675
             Onno-1-1-1:	2675 
          Hannibal 1.0a:	2674
       Deep Onno 1-2-70:	2674
               Naum 3.1:	2673
         Zappa Mexico I:	2672
         Rybka 1.0 Beta:	2671
        Spark-0.3 VC(a):	2668
             Onno-1-0-0:	2666
      Deep Sjeng WC2008:	2663
  Toga II 1.4 beta5c BB:	2660
          Strelka 2.0 B:	2658
       Deep Junior 11.2:	2658
     Hiarcs 12.1 MP 32b:	2651
         Umko 1.2 SSE42:	2650
         Deep Sjeng 3.0:	2648
          Critter 0.52b:	2637
 Shredder Classic 4 32b:	2637
      Deep Junior 11.1a:	2627
           Naum 2.2 32b:	2625
         Umko 1.1 SSE42:	2621
       Deep Junior 2010:	2618
        Glaurung 2.2 JA:	2618
     Rybka 1.0 Beta 32b:	2617
        HIARCS 11.2 32b:	2612
     Fruit 05/11/03 32b:	2610
         Loop 13.6/2007:	2601
         Toga II 1.2.1a:	2600
         Jonny 4.00 32b:	2600
              ListMP 11:	2596
          LoopMP 12 32b:	2594
       Deep Shredder 10:	2590
Twisted Logic 20100131x:	2585
         Crafty 23.3 JA:	2580
    Spike 1.2 Turin 32b:	2563
     Deep Sjeng 2.7 32b:	2539
         Crafty 23.1 JA:	2528

Code: Select all

           Houdini 2.0 Pro x64: 1694.0 /  2100, 80.7%   rating = 2597.0
                  Houdini 1.5a: 3162.5 /  4000, 79.1%   rating = 2590.8
              Komodo64 3 SSE42: 1843.5 /  2500, 73.7%   rating = 2543.3
          Deep Rybka 4.1 SSE42: 2234.5 /  3100, 72.1%   rating = 2537.9
                   Critter 1.2: 1961.0 /  2700, 72.6%   rating = 2536.4
                  Deep Rybka 4: 3584.5 /  4800, 74.7%   rating = 2533.7
                 Houdini 1.03a: 2520.0 /  3200, 78.8%   rating = 2531.9
          Komodo 2.03 DC SSE42: 1985.5 /  2700, 73.5%   rating = 2527.4
            Stockfish 2.1.1 JA: 2119.5 /  3000, 70.7%   rating = 2520.8
            Critter 1.01 SSE42: 1970.0 /  2800, 70.4%   rating = 2499.1
             Stockfish 2.01 JA: 2246.0 /  3100, 72.5%   rating = 2498.4
            Stockfish 1.9.1 JA: 2131.0 /  3000, 71.0%   rating = 2476.7
                    Rybka 3 mp: 3228.0 /  4200, 76.9%   rating = 2476.4
            Critter 0.90 SSE42: 2327.5 /  3400, 68.5%   rating = 2469.0
            Stockfish 1.7.1 JA: 2131.0 /  2900, 73.5%   rating = 2461.5
                   Rybka 3 32b: 1191.5 /  1700, 70.1%   rating = 2418.2
            Stockfish 1.6.x JA: 1792.5 /  2600, 68.9%   rating = 2402.5
               Komodo64 1.3 JA: 1946.0 /  3300, 59.0%   rating = 2398.1
                      Naum 4.2: 3682.5 /  6300, 58.5%   rating = 2391.3
                  Critter 0.80: 1795.5 /  2800, 64.1%   rating = 2383.8
                 Komodo 1.2 JA: 2175.0 /  3700, 58.8%   rating = 2368.6
               Rybka 2.3.2a mp: 2172.5 /  3500, 62.1%   rating = 2364.3
              Deep Shredder 12: 4130.0 /  7400, 55.8%   rating = 2360.8
                      Gull 1.2: 1639.5 /  3300, 49.7%   rating = 2354.5
                  Critter 0.70: 1107.0 /  1900, 58.3%   rating = 2351.1
                      Gull 1.1: 1675.5 /  3100, 54.0%   rating = 2350.7
                      Naum 4.1: 1465.0 /  2300, 63.7%   rating = 2349.2
       Deep Sjeng c't 2010 32b: 2116.0 /  4300, 49.2%   rating = 2346.2
                 Komodo 1.0 JA: 1756.5 /  2900, 60.6%   rating = 2344.6
                 Spike 1.4 32b: 1626.5 /  3400, 47.8%   rating = 2340.1
             Deep Fritz 12 32b: 3251.5 /  6200, 52.4%   rating = 2338.1
                        Naum 4: 1628.5 /  2700, 60.3%   rating = 2335.7
                Rybka 2.2n2 mp: 1311.5 /  2100, 62.5%   rating = 2335.1
                     Gull 1.0a: 1254.0 /  2300, 54.5%   rating = 2326.5
            Stockfish 1.5.1 JA: 1128.5 /  1900, 59.4%   rating = 2322.2
                    Rybka 1.2f: 1578.5 /  2400, 65.8%   rating = 2321.0
           Protector 1.4.0 x64: 1562.5 /  3300, 47.3%   rating = 2317.0
               spark-1.0 SSE42: 1785.5 /  4000, 44.6%   rating = 2312.8
                  Hannibal 1.1: 1242.5 /  2800, 44.4%   rating = 2311.3
            HIARCS 13.2 MP 32b: 1683.5 /  3800, 44.3%   rating = 2307.4
                  Fritz 12 32b: 1091.0 /  2000, 54.5%   rating = 2300.7
            HIARCS 13.1 MP 32b: 1734.5 /  3600, 48.2%   rating = 2287.9
              Deep Junior 12.5: 1266.0 /  3100, 40.8%   rating = 2285.9
             Deep Fritz 11 32b:  744.5 /  1300, 57.3%   rating = 2281.2
                 Doch64 1.2 JA:  820.5 /  1600, 51.3%   rating = 2271.2
                     spark-0.4: 1458.0 /  3100, 47.0%   rating = 2269.5
              Stockfish 1.4 JA:  849.0 /  1700, 49.9%   rating = 2269.0
               Protector 1.4.0:   30.0 /   200, 15.0%   rating = 2268.6
               Zappa Mexico II: 3984.0 /  8700, 45.8%   rating = 2267.8
             Shredder Bonn 32b: 1119.0 /  2200, 50.9%   rating = 2265.8
                  Critter 0.60: 1072.0 /  2200, 48.7%   rating = 2254.5
            Protector 1.3.2 JA: 2361.5 /  5300, 44.6%   rating = 2254.3
              Deep Shredder 11: 1412.0 /  2700, 52.3%   rating = 2245.7
              Doch64 09.980 JA:  710.0 /  1500, 47.3%   rating = 2243.2
                Deep Junior 12: 1356.0 /  3600, 37.7%   rating = 2235.6
                    Onno-1-1-1: 1923.0 /  4300, 44.7%   rating = 2235.3
                 Hannibal 1.0a: 1600.0 /  4200, 38.1%   rating = 2235.0
              Deep Onno 1-2-70: 1973.5 /  5300, 37.2%   rating = 2234.8
                      Naum 3.1: 1514.5 /  3000, 50.5%   rating = 2233.7
                Zappa Mexico I: 1221.0 /  2200, 55.5%   rating = 2232.9
                Rybka 1.0 Beta: 1023.5 /  2300, 44.5%   rating = 2232.0
               Spark-0.3 VC(a): 1625.0 /  3600, 45.1%   rating = 2229.1
                    Onno-1-0-0:  594.5 /  1200, 49.5%   rating = 2226.5
             Deep Sjeng WC2008: 2434.5 /  5600, 43.5%   rating = 2224.1
         Toga II 1.4 beta5c BB: 3244.5 /  8200, 39.6%   rating = 2220.4
                 Strelka 2.0 B: 1125.0 /  3400, 33.1%   rating = 2219.2
              Deep Junior 11.2: 1176.0 /  2900, 40.6%   rating = 2219.1
            Hiarcs 12.1 MP 32b: 2427.5 /  5600, 43.3%   rating = 2211.4
                Umko 1.2 SSE42:  825.5 /  2600, 31.8%   rating = 2211.1
                Deep Sjeng 3.0:  601.5 /  1400, 43.0%   rating = 2208.9
                 Critter 0.52b: 1097.0 /  2600, 42.2%   rating = 2197.9
        Shredder Classic 4 32b:  922.5 /  1800, 51.2%   rating = 2197.8
             Deep Junior 11.1a: 1153.0 /  2800, 41.2%   rating = 2187.9
                  Naum 2.2 32b:  614.0 /  1300, 47.2%   rating = 2186.0
                Umko 1.1 SSE42: 1146.0 /  3900, 29.4%   rating = 2181.4
              Deep Junior 2010: 1210.0 /  3100, 39.0%   rating = 2179.1
               Glaurung 2.2 JA: 1027.5 /  2600, 39.5%   rating = 2178.4
            Rybka 1.0 Beta 32b:  506.0 /  1100, 46.0%   rating = 2177.9
               HIARCS 11.2 32b:  827.0 /  1900, 43.5%   rating = 2173.1
            Fruit 05/11/03 32b: 1774.0 /  4400, 40.3%   rating = 2170.7
                Loop 13.6/2007: 2270.0 /  7200, 31.5%   rating = 2161.6
                Toga II 1.2.1a:  716.5 /  1600, 44.8%   rating = 2160.5
                Jonny 4.00 32b: 1225.5 /  4500, 27.2%   rating = 2160.3
                     ListMP 11:  987.5 /  2600, 38.0%   rating = 2156.3
                 LoopMP 12 32b:  635.0 /  1500, 42.3%   rating = 2154.3
              Deep Shredder 10: 1754.0 /  4400, 39.9%   rating = 2150.5
       Twisted Logic 20100131x: 1140.0 /  3500, 32.6%   rating = 2146.2
                Crafty 23.3 JA: 1123.5 /  4400, 25.5%   rating = 2140.8
           Spike 1.2 Turin 32b: 2349.5 /  7700, 30.5%   rating = 2124.0
            Deep Sjeng 2.7 32b:  465.5 /  1400, 33.2%   rating = 2100.2
                Crafty 23.1 JA: 1002.0 /  3800, 26.4%   rating = 2089.0

[Adam Hair]

Here is the Bayeselo output of the games.pgn file that Ingo provided.

The default values for white advantage and draw elo have been replaced by the values computed from the data, and I used the covariance function rather than exact distance.

As you can see, some of the compression has been removed.

Code: Select all

Rank Name                      Elo    +    - games score oppo. draws
   1 Houdini 2.0 Pro x64      3021   14   14  2100   81%  2765   24%
   2 Houdini 1.5a             3015   10   10  4000   79%  2775   26%
   3 Komodo64 3 SSE42         2969   12   12  2500   74%  2781   31%
   4 Deep Rybka 4.1 SSE42     2960   11   11  3100   72%  2792   37%
   5 Critter 1.2              2958   11   11  2700   73%  2784   36%
   6 Deep Rybka 4             2958    9    9  4800   75%  2765   33%
   7 Komodo 2.03 DC SSE42     2956   12   12  2700   74%  2770   30%
   8 Houdini 1.03a            2956   11   11  3200   79%  2728   30%
   9 Stockfish 2.1.1 JA       2947   11   11  3000   71%  2787   35%
  10 Critter 1.01 SSE42       2927   11   11  2800   70%  2773   36%
  11 Stockfish 2.01 JA        2926   11   11  3100   72%  2756   35%
  12 Rybka 3 mp               2906    9    9  4200   77%  2702   31%
  13 Stockfish 1.9.1 JA       2905   11   11  3000   71%  2748   36%
  14 Critter 0.90 SSE42       2899   10   10  3400   68%  2762   36%
  15 Stockfish 1.7.1 JA       2891   11   11  2900   73%  2713   33%
  16 Rybka 3 32b              2854   14   14  1700   70%  2708   35%
  17 Stockfish 1.6.x JA       2836   11   11  2600   69%  2699   37%
  18 Komodo64 1.3 JA          2835   10   10  3300   59%  2769   37%
  19 Naum 4.2                 2826    7    7  6300   58%  2766   40%
  20 Critter 0.80             2822   11   11  2800   64%  2718   36%
  21 Komodo 1.2 JA            2807    9    9  3700   59%  2743   40%
  22 Rybka 2.3.2a mp          2801    9    9  3500   62%  2717   40%
  23 Deep Shredder 12         2800    6    6  7400   56%  2758   38%
  24 Gull 1.2                 2795   10   10  3300   50%  2796   36%
  25 Gull 1.1                 2791   10   10  3100   54%  2761   38%
  26 Critter 0.70             2791   13   13  1900   58%  2730   36%
  27 Naum 4.1                 2788   12   12  2300   64%  2690   40%
  28 Deep Sjeng c't 2010 32b  2787    9    9  4300   49%  2792   39%
  29 Komodo 1.0 JA            2783   10   10  2900   61%  2708   42%
  30 Spike 1.4 32b            2781   10   10  3400   48%  2797   38%
  31 Deep Fritz 12 32b        2778    7    7  6200   52%  2761   38%
  32 Rybka 2.2n2 mp           2776   12   12  2100   62%  2688   40%
  33 Naum 4                   2776   11   11  2700   60%  2703   40%
  34 Gull 1.0a                2768   11   11  2300   55%  2735   39%
  35 Rybka 1.2f               2764   12   12  2400   66%  2652   36%
  36 Stockfish 1.5.1 JA       2763   13   13  1900   59%  2698   38%
  37 Protector 1.4.0 x64      2758   10   10  3300   47%  2778   37%
  38 spark-1.0 SSE42          2755    9    9  4000   45%  2794   39%
  39 Hannibal 1.1             2754   11   11  2800   44%  2796   39%
  40 HIARCS 13.2 MP 32b       2750    9    9  3800   44%  2792   37%
  41 Fritz 12 32b             2744   12   12  2000   55%  2711   40%
  42 HIARCS 13.1 MP 32b       2730    9    9  3600   48%  2745   37%
  43 Deep Junior 12.5         2728   10   10  3100   41%  2799   34%
  44 Deep Fritz 11 32b        2727   15   15  1300   57%  2675   39%
  45 Protector 1.4.0          2716   47   47   200   15%  2995   22%
  46 Doch64 1.2 JA            2715   14   14  1600   51%  2706   41%
  47 spark-0.4                2713   10   10  3100   47%  2735   39%
  48 Stockfish 1.4 JA         2713   13   13  1700   50%  2714   38%
  49 Zappa Mexico II          2712    6    6  8700   46%  2744   37%
  50 Shredder Bonn 32b        2711   12   12  2200   51%  2705   36%
  51 Protector 1.3.2 JA       2699    8    8  5300   45%  2739   38%
  52 Critter 0.60             2699   12   12  2200   49%  2708   37%
  53 Deep Shredder 11         2692   11   11  2700   52%  2676   36%
  54 Doch64 09.980 JA         2688   14   14  1500   47%  2707   38%
  55 Naum 3.1                 2682   10   10  3000   50%  2676   39%
  56 Deep Onno 1-2-70         2681    8    8  5300   37%  2778   36%
  57 Hannibal 1.0a            2680    9    9  4200   38%  2770   33%
  58 Onno-1-1-1               2680    8    8  4300   45%  2719   40%
  59 Deep Junior 12           2679   10   10  3600   38%  2773   30%
  60 Zappa Mexico I           2679   12   12  2200   56%  2641   41%
  61 Rybka 1.0 Beta           2677   12   12  2300   45%  2721   35%
  62 Spark-0.3 VC(a)          2674    9    9  3600   45%  2709   40%
  63 Onno-1-0-0               2673   16   16  1200   50%  2675   41%
  64 Deep Sjeng WC2008        2670    7    7  5600   43%  2717   37%
  65 Strelka 2.0 B            2668   10   10  3400   33%  2797   34%
  66 Toga II 1.4 beta5c BB    2667    6    6  8200   40%  2745   38%
  67 Deep Junior 11.2         2662   11   11  2900   41%  2735   31%
  68 Umko 1.2 SSE42           2660   11   11  2600   32%  2801   36%
  69 Hiarcs 12.1 MP 32b       2656    7    7  5600   43%  2705   38%
  70 Deep Sjeng 3.0           2654   15   15  1400   43%  2706   34%
  71 Shredder Classic 4 32b   2645   13   13  1800   51%  2636   38%
  72 Critter 0.52b            2644   11   11  2600   42%  2701   39%
  73 Naum 2.2 32b             2636   15   15  1300   47%  2653   45%
  74 Deep Junior 11.1a        2635   11   11  2800   41%  2699   34%
  75 Umko 1.1 SSE42           2630    9    9  3900   29%  2789   33%
  76 Rybka 1.0 Beta 32b       2627   17   17  1100   46%  2654   37%
  77 Glaurung 2.2 JA          2627   11   11  2600   40%  2700   38%
  78 Deep Junior 2010         2625   10   10  3100   39%  2706   31%
  79 HIARCS 11.2 32b          2621   13   13  1900   44%  2667   38%
  80 Fruit 05/11/03 32b       2620    8    8  4400   40%  2688   41%
  81 Loop 13.6/2007           2612    7    7  7200   32%  2754   34%
  82 Jonny 4.00 32b           2608    9    9  4500   27%  2790   28%
  83 Toga II 1.2.1a           2608   14   14  1600   45%  2646   41%
  84 ListMP 11                2605   11   11  2600   38%  2693   37%
  85 LoopMP 12 32b            2603   14   14  1500   42%  2656   38%
  86 Deep Shredder 10         2597    9    9  4400   40%  2672   33%
  87 Twisted Logic 20100131x  2593   10   10  3500   33%  2727   30%
  88 Crafty 23.3 JA           2589    9    9  4400   26%  2787   28%
  89 Spike 1.2 Turin 32b      2574    7    7  7700   31%  2722   33%
  90 Deep Sjeng 2.7 32b       2551   15   15  1400   33%  2671   36%
  91 Crafty 23.1 JA           2539   10   10  3800   26%  2724   28%