Calculating the sum of perft on an epd file

[Uri Blass]

I think that calculating the sum of perft on an epd file may be more effective to detect bugs relative to calculating perft on a single position.

I wonder if there is some program that can calculate some of perft on an epd file in order to compare results.

Movei until today had no function to read FEN with enpassant information and always assumed that enpassant is illegal.

I decided today to implement reading FEN with the enpassant information and I hope that I have no bugs and the question is if there is some program that I can download to compare my results with it.

[Dan Honeycutt]

Hi Uri

If I understand, you're looking for something to read an epd file and calculate the perft. While not exactly that, Andrew Wagner's move generator test might be of use. It's an epd file with 100+ positions with the perft values to depth 5 for each position. You can probably find it if the CCC search engine is working, or if not I can email it to you when I get home this evening.

Dan H.

[Sune Fischer]

Hi Uri,

I sent you a tool to do something like that, it outputs like this:

Code: Select all

 line        perft(5)             sum     fen
    1       105159472       105159472     r1b1r1k1/p1p3pp/2p2n2/2bp4/5P2/3BBQPq/PPPK3P/R4N1R b - - bm Bg4; id "ECM.946";
    2       261536154       366695626     1r5r/3b1pk1/3p1np1/p1qPp3/p1N1PbP1/2P2PN1/1PB1Q1K1/R3R3 b - - bm Nxg4; id "ECM.949";
    3       191556672       558252298     2k4r/1pp2ppp/p1p1bn2/4N3/1q1rP3/2N1Q3/PPP2PPP/R4RK1 w - - bm Nd5; id "ECM.952";
    4        50921405       609173703     rn3rk1/4bppp/1q2p3/p2pP3/8/1PN2B1P/P4PP1/2RQ1RK1 w - - bm Bxd5; id "ECM.961";
    5       198670058       807843761     2r2bk1/4qp2/3n2p1/2R1p1Np/2p1N3/r6P/1Q3PP1/3R2K1 w - - bm Rxc8; id "ECM.973";
......
  183       121907403     18748682400     2rq1bk1/1br2p1p/p2p2p1/1p1P4/4Q3/PP3N2/1BP5/1K1R3R w - - bm Rxh7; id "ECM.1623";


I sent it to the email listed in your profile on CCC.

-S.

[Uri Blass]

Thanks
I got it

I still did not use it and I have problems of bugs in the tool that read pgn
(correction I mean in the tool that I wrote to read epd and the problem is possible crash at illegal FEN or at the end of the file).

I will compare results inspite of the fact that it has bugs and can crush at the end of the file

I have the following output at the time of posting(I use relatively slow hardware but I still expect movei to get 18,748,682,400 nodes fast enough

So far I got the following output in the screen

105159472 1 366695626 2 558252298 3 609173703 4 807843761 5 982917177
6 1089988524 7 1176059100 8 1279555611 9 1412214565 10 1583859342 11
1745578164 12 1844112796 13 1939626437 14 2061665341 15
2206909154 16

2552824780 17 2669169673 18 2788859280 19 2815890833 20 2965619332 21
3070131691 22 3252684270 23 3399989461 24 3547535125 25 3638383022 26
3728231577 27 3830428951 28 3972119075 29 4034710064 30 4088635726 31
4104096964 32 4151388663 33 4426563485 34 4442282619 35 4540782132 36
4763386941 37 4899683307 38 4949465924 39 5017570596 40 5086522975 41
5198022522 42 5306981284 43 5421245688 44 5629969193 45 5649929707 46
5816941940 47 5889079698 48 5968844745 49 5976849783 50 6039631629 51
6082839796 52 6154816687 53 6228461464 54 6407205202 55 6448707977 56
6587696077 57 6824050796 58 6917673308 59 7020011707 60 7153427921 61 7244329900 62 7305499773 63 7356634664 64 7467788335 65 7573698882 66
7632147806 67 7824162502 68 8035708980 69 8130847062 70 8233250898 71
8392484660 72 8422911538 73 8568181354 74 8572028824 75 8781241541 76
8874632908 77 8970850068 78 9050678883 79 9139641193 80 9273756849 81
9538999355 82 9554027428 83 9672012112 84 9729425971 85 9830774102 86
9903896483 87 9987173804 88 10095439292 89 10199725733 90 10299214674
91 10485194231 92 10520018272 93 10647622948 94 10757478424 95 1080031
0419 96 10864698060 97 10913869986 98 11086334448 99 11187829406 100 1
1246381270 101 11332318963 102 11445006971 103 11508426393 104 115384068
07 105 11628168022 106 11704303670 107 11762069744 108
11916220952 109 12142841518 110 12254352905 111 12425057892 112 12510350418 113 1264725
4559 114 12747100973 115 12780107127 116 12906651887 117 13025641789 11
8 13067363672 119 13101700374 120 13138914177 121 13306709128 122 13415
620652 123 13458728036 124 13537763673 125 13689962512 126 13804006261
127 13876945285 128 13985690646 129 14046767371 130 14113529386 131 142
96027625 132 14351005176 133 14449735480 134 14560982959 135 14601365624
136 14695536775 137 14889613409 138 14999001782 139 15131277402 140 1
5248887851 141 15314141344 142 15387722711 143 15648767464 144 157088767
35 145 15740647082 146 15848808679 147 16038442212 148 16165440196 149
16242119235 150 16242240444 151 16242582550 152 16340202768 153 1635142
6043 154 16401353654 155 16420052597 156 16428390274 157 16542590465 15
8 16595059722 159 16673164789 160 16676989357 161 16716268361 162 16717
789525 163 16815875857 164 16874650395 165 16993785659 166 17117362519
167 17217661767 168 17260592243 169 17397985600 170 17487689885 171 176
93152696 172 17862382726 173 17911937911 174 18030317659 175 18187002970
176 18297672867 177 18454220923 178 18543311173 179 18608189781 180 1
8638779133 181 18737708800 182 18859616203 183

Unfortunately the final numer is not the same but maybe we have different positions so I will run your tool to check it.


Uri

[Uri Blass]

To be more correct practically there is still no program to read general epd
and I only changed the program that I have to read epd by adding
perft_epd(const char* epdfile) and calling the right epd file from main.

Uri

[Uri Blass]

Frenzee agrees about the numbers and I guess that we have different files
It needs only 22:24 minutes on my slow A1000 without hash to find the result.

Movei did not use perft mode but frenzee is faster than movei in game mode without hash) and it is something that not a lot of programs can do.

Uri