Protej Chess Engine
Protej Chess Engine
alexNew official release.
It includes all updates released with the dev versions after Protej 0.6.6:
Non-functional changes:
It includes also:
Non-functional changes:
Rank Name Elo + - games score oppo. draws 1 rofChade 2.3 3327 11 11 3200 76% 3139 29% 2 Laser 1.7 3306 10 10 3200 73% 3139 30% 3 Roc 1.9.1 3270 10 10 3200 68% 3139 29% 4 Schooner 2.2-sse 3263 10 10 3200 68% 3139 33% 5 Booot 6.5 popcnt 3249 10 10 3200 66% 3139 35% 6 Wasp 6.50 HCE 3242 10 10 3200 65% 3139 34% 7 Drofa 3.3.22 popcnt 3222 10 10 3200 62% 3139 31% 8 Senpai 2.0 3214 10 10 3200 61% 3139 34% 9 iCE 4.0.853 popcnt 3211 10 10 3200 60% 3139 34% 10 IPPOLIT 0.080a intel JA 3199 10 10 3200 58% 3139 26% 11 SlowChess Blitz Classic 1.8 3196 10 10 3200 58% 3139 35% 12 Pedantic 1.1.0 3182 10 10 3200 56% 3139 36% 13 Hannibal 1.7 3168 10 10 3200 54% 3139 34% 14 Protector 1.9.0 3165 10 10 3200 54% 3139 32% 15 Rybka 3 1-cpu 3148 10 10 3200 51% 3139 31% 16 Cheng 4.43 HCE 3145 10 10 3200 51% 3139 30% 17 Protej 0.6.7 3139 2 2 80000 46% 3163 31% 18 Berserk 4.1.0 popcnt 3135 10 10 3200 49% 3139 38% 19 Mr Bob 1.1.0 popcnt 3125 10 10 3200 48% 3139 31% 20 Monolith 2 popcnt 3107 10 10 3200 46% 3139 33% 21 Polaris 1.8.1 popcnt 3073 10 10 3200 41% 3139 32% 22 Naum 4.6 3051 10 10 3200 38% 3139 28% 23 FabChess 1.16 POPCNT 3046 10 10 3200 37% 3139 33% 24 PeSTO 2.210 3023 10 10 3200 35% 3139 25% 25 Cheese 3.1.1 pop 3020 10 10 3200 34% 3139 29% 26 Crafty 25.3 JA Wb2Uci 3000 10 10 3200 31% 3139 27% Pool of 25 engines, average UBC rating 3165, TC=4"+0.1, Little Blitzer testing framework STC Total: 80000 W: 24575 L: 30242 D: 25183 Elo: +62 vs Protej 0.6.6
Rank Name Elo + - games score oppo. draws 1 Wasp 4.50 73 32 32 250 61% -7 43% 2 Weiss 1.4 popcnt 59 32 32 250 60% -7 45% 3 Smallbrain 5.0 (NNUE) 23 32 32 250 54% -7 45% 4 Gull 3 18 31 31 250 54% -7 49% 5 Princhess 0.17.0 (NNUE) 13 32 32 250 53% -7 38% 6 Critter 1.6a 13 32 32 250 53% -7 44% 7 Avalanche 1.4.0 Haswell (NNUE 5 32 32 250 52% -7 43% 8 Equinox 3.30-x64mp -1 31 31 250 51% -7 49% 9 Altair 5.0.0 v2 -5 31 31 250 50% -7 49% 10 Protej 0.6.7 -7 9 9 3000 49% 1 45% 11 Counter 4.0 -42 31 31 250 45% -7 52% 12 Halogen 9 POPCNT -71 32 32 250 41% -7 42% 13 Pedantic 1.0.0 -78 32 32 250 40% -7 40% Pool of 12 engines, average CCRL rating 3200, TC=2'+1" (CCRL Blitz), Little Blitzer testing framework LTC Total: 3000 W: 794 L: 856 D: 1350 Elo: 3193 (CCRL Blitz scale)
Rank Name Elo + - games score oppo. draws 1 Protej 0.6.7 142 4 4 20000 72% -36 25% 2 Protej 0.6.6 83 8 8 5000 41% 142 42% 3 Protej 0.6.5 17 8 8 5000 33% 142 26% 4 Protej 0.6.4 -60 9 9 5000 25% 142 16% 5 Protej 0.6.3 -182 11 11 5000 13% 142 15% Selfplay vs. previous versions, TC=4"+0.1, Little Blitzer testing framework STC Total: 20000 W: 11957 L: 3136 D: 4907 Elo: +59 vs Protej 0.6.6 Elo: +125 vs Protej 0.6.5 Elo: +202 vs Protej 0.6.4 Elo: +324 vs Protej 0.6.3
alexImproved the search function by adjusting the search parameters and making various refinements, resulting in slightly better move ordering and search efficiency. Non-functional change: removed the graphical embellishments used in console mode, as this mode is virtually unused and the extensive code did not offer any improvement to the playing strength. Rank Name Elo + - games score oppo. draws 1 rofChade 2.3 3338 10 10 3600 78% 3133 27% 2 Laser 1.7 3315 10 10 3600 75% 3133 29% 3 Schooner 2.2-sse 3281 10 10 3600 71% 3133 32% 4 Roc 1.9.1 3257 10 10 3600 67% 3133 29% 5 Booot 6.5 popcnt 3255 9 9 3600 67% 3133 34% 6 Wasp 6.50 HCE 3252 9 9 3600 67% 3133 34% 7 Drofa 3.3.22 popcnt 3222 9 9 3600 62% 3133 30% 8 iCE 4.0.853 popcnt 3213 9 9 3600 62% 3133 34% 9 Senpai 2.0 3212 9 9 3600 61% 3133 35% 10 IPPOLIT 0.080a intel JA 3201 10 10 3600 59% 3133 25% 11 SlowChess Blitz Classic 1.8 3192 9 9 3600 59% 3133 36% 12 Protector 1.9.0 3185 9 9 3600 57% 3133 32% 13 Pedantic 1.1.0 3184 9 9 3600 57% 3133 35% 14 Hannibal 1.7 3173 9 9 3600 56% 3133 33% 15 Cheng 4.43 HCE 3150 9 9 3600 52% 3133 33% 16 Berserk 4.1.0 popcnt 3149 9 9 3600 52% 3133 39% 17 Rybka 3 1-cpu 3143 9 9 3600 51% 3133 30% 18 Protej dev_20240811a 3133 2 2 90000 45% 3169 32% 19 Mr Bob 1.1.0 popcnt 3127 9 9 3600 49% 3133 36% 20 Monolith 2 popcnt 3119 9 9 3600 48% 3133 32% 21 Polaris 1.8.1 popcnt 3088 9 9 3600 44% 3133 33% 22 Naum 4.6 3066 9 9 3600 41% 3133 29% 23 FabChess 1.16 POPCNT 3053 9 9 3600 39% 3133 33% 24 Cheese 3.1.1 pop 3036 10 10 3600 37% 3133 29% 25 PeSTO 2.210 3027 10 10 3600 36% 3133 25% 26 Crafty 25.3 JA Wb2Uci 3000 10 10 3600 32% 3133 26% Pool of 25 engines, average UBC rating 3165, TC=4"+0.1, Little Blitzer testing framework STC Total: 90000 W: 26222 L: 35408 D: 28370 Elo: +13 vs Protej dev_20240807b
alexThe evaluation of passed pawns in specific types of positions has been toned down, as their strength was previously overestimated. Non-functional change: the Online Syzygy feature and its associated UCI options have been removed. The philosophy behind Protej is to be a completely self-sufficient engine, one that "knows what it knows", and thus it does not rely on external resources such as endgame tablebases, opening books, or external support. Furthermore, Protej will not utilize multithreading or NNUE evaluations, staying true to its core principle of independence. Rank Name Elo + - games score oppo. draws 1 rofChade 2.3 3325 12 12 2400 78% 3120 28% 2 Laser 1.7 3307 12 12 2400 76% 3120 29% 3 Roc 1.9.1 3273 12 12 2400 71% 3120 28% 4 Schooner 2.2-sse 3273 12 12 2400 71% 3120 32% 5 Booot 6.5 popcnt 3254 12 12 2400 69% 3120 35% 6 Wasp 6.50 HCE 3245 12 12 2400 68% 3120 33% 7 Drofa 3.3.22 popcnt 3225 12 12 2400 64% 3120 28% 8 Senpai 2.0 3218 11 11 2400 64% 3120 35% 9 iCE 4.0.853 popcnt 3211 11 11 2400 63% 3120 34% 10 IPPOLIT 0.080a intel JA 3206 12 12 2400 61% 3120 23% 11 SlowChess Blitz Classic 1.8 3197 11 11 2400 61% 3120 34% 12 Protector 1.9.0 3178 11 11 2400 58% 3120 32% 13 Pedantic 1.1.0 3177 11 11 2400 58% 3120 36% 14 Cheng 4.43 HCE 3164 11 11 2400 56% 3120 30% 15 Hannibal 1.7 3159 11 11 2400 55% 3120 33% 16 Berserk 4.1.0 popcnt 3146 11 11 2400 54% 3120 41% 17 Rybka 3 1-cpu 3144 11 11 2400 53% 3120 30% 18 Mr Bob 1.1.0 popcnt 3131 11 11 2400 52% 3120 34% 19 Protej dev_20240807b 3120 2 2 60000 43% 3168 32% 20 Monolith 2 popcnt 3112 11 11 2400 49% 3120 35% 21 Polaris 1.8.1 popcnt 3088 11 11 2400 45% 3120 33% 22 FabChess 1.16 POPCNT 3062 11 11 2400 42% 3120 35% 23 Naum 4.6 3060 12 12 2400 42% 3120 30% 24 Cheese 3.1.1 pop 3033 12 12 2400 38% 3120 30% 25 PeSTO 2.210 3024 12 12 2400 37% 3120 26% 26 Crafty 25.3 JA Wb2Uci 3000 12 12 2400 34% 3120 27% Pool of 25 engines, average UBC rating 3165, TC=4"+0.1, Little Blitzer testing framework STC Total: 60000 W: 16434 L: 24533 D: 19033 Elo: +18 vs Protej dev_20240805a
alexThe terms of the pawn evaluation function have been refined, and the level of abstraction, which in Delphi is counterproductive, has also been reduced. Non-functional change: new compiler version, Delphi 12. Rank Name Elo + - games score oppo. draws 1 rofChade 2.3 3337 14 14 2000 81% 3102 25% 2 Laser 1.7 3298 14 14 2000 76% 3102 26% 3 Roc 1.9.1 3263 13 13 2000 72% 3102 28% 4 Booot 6.5 popcnt 3261 13 13 2000 72% 3102 30% 5 Schooner 2.2-sse 3258 13 13 2000 72% 3102 30% 6 Wasp 6.50 HCE 3242 13 13 2000 70% 3102 31% 7 Drofa 3.3.22 popcnt 3233 13 13 2000 68% 3102 28% 8 Senpai 2.0 3208 13 13 2000 65% 3102 33% 9 iCE 4.0.853 popcnt 3203 13 13 2000 64% 3102 31% 10 IPPOLIT 0.080a intel JA 3197 13 13 2000 63% 3102 23% 11 SlowChess Blitz Classic 1.8 3193 13 13 2000 63% 3102 32% 12 Pedantic 1.1.0 3176 13 13 2000 61% 3102 31% 13 Protector 1.9.0 3171 13 13 2000 60% 3102 31% 14 Hannibal 1.7 3164 13 13 2000 59% 3102 31% 15 Berserk 4.1.0 popcnt 3147 12 12 2000 57% 3102 35% 16 Cheng 4.43 HCE 3138 13 13 2000 55% 3102 29% 17 Rybka 3 1-cpu 3138 13 13 2000 55% 3102 28% 18 Monolith 2 popcnt 3121 13 13 2000 53% 3102 32% 19 Mr Bob 1.1.0 popcnt 3119 12 12 2000 53% 3102 33% 20 Protej dev_20240805a 3102 3 3 50000 41% 3165 30% 21 Polaris 1.8.1 popcnt 3089 12 12 2000 48% 3102 34% 22 Naum 4.6 3063 13 13 2000 45% 3102 30% 23 FabChess 1.16 POPCNT 3054 12 12 2000 43% 3102 36% 24 Cheese 3.1.1 pop 3033 13 13 2000 41% 3102 30% 25 PeSTO 2.210 3013 13 13 2000 38% 3102 25% 26 Crafty 25.3 JA Wb2Uci 3000 13 13 2000 36% 3102 27% Pool of 25 engines, average UBC rating 3165, TC=4"+0.1, Little Blitzer testing framework STC Total: W: L: D: Elo: +21 vs Protej 0.6.6
alexRetest of version 0.6.6. Rank Name Elo + - games score oppo. draws 1 rofChade 2.3 3324 10 10 4560 82% 3081 24% 2 Laser 1.7 3306 9 9 4560 80% 3081 24% 3 Schooner 2.2-sse 3265 9 9 4560 75% 3081 27% 4 Roc 1.9.1 3261 9 9 4560 74% 3081 25% 5 Booot 6.5 popcnt 3248 9 9 4560 73% 3081 29% 6 Wasp 6.50 HCE 3243 9 9 4560 72% 3081 29% 7 Drofa 3.3.22 popcnt 3213 9 9 4560 68% 3081 27% 8 IPPOLIT 0.080a intel JA 3210 9 9 4560 67% 3081 20% 9 Senpai 2.0 3208 9 9 4560 68% 3081 30% 10 iCE 4.0.853 popcnt 3207 9 9 4560 68% 3081 30% 11 SlowChess Blitz Classic 1.8 3202 9 9 4560 67% 3081 29% 12 Pedantic 1.1.0 3180 8 8 4560 64% 3081 31% 13 Protector 1.9.0 3168 9 9 4560 62% 3081 28% 14 Hannibal 1.7 3167 9 9 4560 62% 3081 28% 15 Cheng 4.43 HCE 3142 9 9 4560 58% 3081 26% 16 Rybka 3 1-cpu 3137 9 9 4560 58% 3081 25% 17 Berserk 4.1.0 popcnt 3135 8 8 4560 58% 3081 36% 18 Mr Bob 1.1.0 popcnt 3115 8 8 4560 55% 3081 31% 19 Monolith 2 popcnt 3106 8 8 4560 54% 3081 32% 20 Polaris 1.8.1 popcnt 3085 8 8 4560 51% 3081 31% 21 Protej 0.6.6 3081 2 2 114000 39% 3163 28% 22 FabChess 1.16 POPCNT 3050 8 8 4560 46% 3081 32% 23 Naum 4.6 3048 9 9 4560 45% 3081 27% 24 PeSTO 2.210 3022 9 9 4560 42% 3081 25% 25 Cheese 3.1.1 pop 3021 8 8 4560 42% 3081 29% 26 Crafty 25.3 JA Wb2Uci 3000 9 9 4560 39% 3081 26% Pool of 25 engines, average UBC rating 3165, TC=4"+0.1, Little Blitzer testing framework STC Total: 114000 W: 28305 L: 53720 D: 31975 Elo: 3081
alexNew official release.
It includes all updates released with the dev versions after Protej 0.6.5:
Non-functional changes:
It includes also:
Non-functional changes:
Rank Name Elo + - games score oppo. draws 1 Senpai 2.0 105 12 12 2500 68% -17 31% 2 Deep iCE 4.0.853 x64/popcnt 100 12 12 2500 67% -17 30% 3 Texel 1.07 96 12 12 2500 66% -17 27% 4 SlowChess Blitz Classic 1.8 89 12 12 2500 65% -17 30% 5 Wasp 3.75 86 12 12 2500 65% -17 30% 6 Hannibal 1.7 x64 68 12 12 2500 62% -17 28% 7 Protector 1.9.0 66 12 12 2500 62% -17 27% 8 Vajolet2 2.8.0 56 11 11 2500 61% -17 31% 9 Weiss 1.3 54 12 12 2500 60% -17 29% 10 Counter 4.0 42 11 11 2500 58% -17 31% 11 Cheng 4.41 35 12 12 2500 57% -17 27% 12 Mr Bob v1.1.0 35 11 11 2500 58% -17 31% 13 Berserk 4.1.0 30 11 11 2500 57% -17 37% 14 Combusken 1.2.0 8 11 11 2500 54% -17 30% 15 Invictus r382 8 11 11 2500 54% -17 35% 16 Monolith 2 6 11 11 2500 53% -17 32% 17 Stash v28.0 5 11 11 2500 53% -17 29% 18 Tucano 9.00 -11 11 11 2500 51% -17 32% 19 Protej 0.6.6 -17 2 2 70000 47% 1 29% 20 Koivisto 64 4.0 -18 11 11 2500 50% -17 30% 21 Polaris 1.8.1 -18 11 11 2500 50% -17 31% 22 Minic 2.33 -40 11 11 2500 47% -17 29% 23 Naum 4.6 -51 12 12 2500 45% -17 27% 24 FabChess v1.16 -52 11 11 2500 45% -17 32% 25 Winter 0.8 SSE4.2 -53 12 12 2500 45% -17 26% 26 Cheese 3.1.1 -63 11 11 2500 44% -17 29% 27 PeSTO 2.210 -80 12 12 2500 41% -17 25% 28 GOOB 1.8.9 -123 12 12 2500 36% -17 25% 29 bright-0.4a -365 17 17 2500 13% -17 8% Pool of 28 engines, average CCRL rating 3095, TC=4"+0.1, Little Blitzer testing framework STC Total: 70000 W: 22795 L: 27040 D: 20165 Elo: +84 vs Protej 0.6.5
Rank Name Elo + - games score oppo. draws 1 Protej 0.6.6 31 4 4 21000 59% -31 26% 2 Protej 0.6.5 -31 4 4 21000 41% 31 26% Selfplay vs. Protej 0.6.5, TC=4"+0.1, Little Blitzer testing framework STC Total: 21000 W: 9639 L: 5963 D: 5398 Elo: +62 vs Protej 0.6.5
Rank Name Elo + - games score oppo. draws 1 Protector 1.9.0 98 50 50 107 55% 64 42% 2 Winter 0.8 SSE4.2 97 50 50 107 55% 64 40% 3 Weiss 1.3 83 49 49 107 53% 64 44% 4 Vajolet2 2.8.0 72 49 49 107 51% 64 43% 5 Wasp 3.75 72 48 48 107 51% 64 50% 6 Deep iCE 4.0.853 x64/popcnt 70 49 49 107 51% 64 48% 7 Protej 0.6.6 64 10 10 3000 57% -3 39% 8 Senpai 2.0 59 49 49 107 49% 64 44% 9 SlowChess Blitz Classic 1.8 58 52 52 107 49% 64 33% 10 Texel 1.07 57 50 50 107 49% 64 40% 11 Hannibal 1.7 x64 57 52 52 107 49% 64 31% 12 Koivisto 64 4.0 50 49 49 107 48% 64 47% 13 Invictus r382 46 49 49 107 48% 64 45% 14 Combusken 1.2.0 44 47 47 107 47% 64 56% 15 Berserk 4.1.0 41 49 49 108 47% 64 44% 16 Mr Bob v1.1.0 36 49 49 107 46% 64 45% 17 Counter 4.0 31 48 48 107 45% 64 49% 18 Cheng 4.41 20 50 50 108 44% 64 41% 19 Stash v28.0 3 48 48 107 41% 64 51% 20 Minic 2.33 3 49 49 107 41% 64 46% 21 Tucano 9.00 -4 52 52 107 41% 64 30% 22 FabChess v1.16 -8 49 49 107 39% 64 47% 23 GOOB 1.8.9 -24 51 51 107 38% 64 33% 24 Naum 4.6 -27 50 50 107 37% 64 40% 25 Monolith 2 -29 52 52 107 37% 64 30% 26 Polaris 1.8.1 -46 52 52 107 36% 64 28% 27 PeSTO 2.210 -77 53 53 107 31% 64 27% 28 Cheese 3.1.1 -119 53 53 108 25% 64 29% 29 bright-0.4a -726 273 273 108 0% 64 0% Pool of 20 engines, average CCRL rating 3095, TC=2'+1" (CCRL Blitz), Little Blitzer testing framework LTC Total: 3000 W: 1124 L: 697 D: 1179 Elo: 3159 (CCRL Blitz scale)
alexMassive optimization of types, variables, and data structures; introduction of new assembler functions. Non-functional change: improved printing related to compilation date. Debug: in console, the function displaying the position now indicates values in centipawns instead of pawns. Rank Name Elo + - games score oppo. draws 1 Senpai 2.0 114 12 12 2250 69% -24 29% 2 Deep iCE 4.0.853 x64/popcnt 100 12 12 2250 67% -24 28% 3 SlowChess Blitz Classic 1.8 91 12 12 2250 66% -24 30% 4 Texel 1.07 86 12 12 2250 65% -24 27% 5 Wasp 3.75 86 12 12 2250 66% -24 30% 6 Protector 1.9.0 74 12 12 2250 64% -24 27% 7 Hannibal 1.7 x64 65 12 12 2250 63% -24 28% 8 Vajolet2 2.8.0 52 12 12 2250 61% -24 31% 9 Berserk 4.1.0 50 12 12 2250 61% -24 35% 10 Weiss 1.3 50 12 12 2250 60% -24 28% 11 Counter 4.0 49 12 12 2250 61% -24 31% 12 Cheng 4.41 28 12 12 2250 57% -24 27% 13 Mr Bob v1.1.0 16 12 12 2250 56% -24 30% 14 Stash v28.0 9 12 12 2250 55% -24 31% 15 Combusken 1.2.0 1 12 12 2250 54% -24 29% 16 Monolith 2 0 12 12 2250 53% -24 32% 17 Tucano 9.00 -1 12 12 2250 53% -24 30% 18 Invictus r382 -1 12 12 2250 53% -24 36% 19 Polaris 1.8.1 -16 12 12 2250 51% -24 31% 20 Koivisto 64 4.0 -21 12 12 2250 50% -24 30% 21 Protej dev_20231226b -24 2 2 63000 46% 1 29% 22 Minic 2.33 -29 12 12 2250 49% -24 28% 23 Naum 4.6 -46 12 12 2250 47% -24 28% 24 Winter 0.8 SSE4.2 -47 12 12 2250 47% -24 26% 25 FabChess v1.16 -49 12 12 2250 46% -24 31% 26 Cheese 3.1.1 -65 12 12 2250 44% -24 31% 27 PeSTO 2.210 -92 12 12 2250 41% -24 23% 28 GOOB 1.8.9 -110 12 12 2250 38% -24 28% 29 bright-0.4a -369 17 17 2250 13% -24 8% Pool of 28 engines, average CCRL rating 3096, TC=4"+0.1, Little Blitzer testing framework STC Total: 63000 W: 19942 L: 24987 D: 18071 Elo: +8 vs Protej dev_202311212a
alexParameter tuning for reduction handling in singular nodes. Rank Name Elo + - games score oppo. draws 1 Deep iCE 4.0.853 x64/popcnt 87 14 14 1750 67% -32 30% 2 Senpai 2.0 86 14 14 1750 67% -32 31% 3 SlowChess Blitz Classic 1.8 85 14 14 1750 66% -32 30% 4 Wasp 3.75 69 14 14 1750 64% -32 31% 5 Texel 1.07 68 14 14 1750 64% -32 30% 6 Hannibal 1.7 x64 55 14 14 1750 62% -32 30% 7 Protector 1.9.0 47 14 14 1750 61% -32 28% 8 Counter 4.0 47 14 14 1750 61% -32 31% 9 Weiss 1.3 44 14 14 1750 61% -32 30% 10 Vajolet2 2.8.0 43 14 14 1750 61% -32 33% 11 Berserk 4.1.0 30 13 13 1750 59% -32 39% 12 Mr Bob v1.1.0 22 14 14 1750 58% -32 32% 13 Cheng 4.41 22 14 14 1750 57% -32 29% 14 Combusken 1.2.0 10 14 14 1750 56% -32 31% 15 Stash v28.0 2 14 14 1750 55% -32 30% 16 Monolith 2 -8 14 14 1750 53% -32 32% 17 Invictus r382 -13 13 13 1750 53% -32 39% 18 Koivisto 64 4.0 -20 13 13 1750 52% -32 32% 19 Tucano 9.00 -29 14 14 1750 50% -32 31% 20 Minic 2.33 -30 14 14 1750 50% -32 29% 21 Polaris 1.8.1 -31 14 14 1750 50% -32 31% 22 Protej dev_20231212a -32 3 3 49000 45% 1 30% 23 FabChess v1.16 -52 13 13 1750 47% -32 32% 24 Winter 0.8 SSE4.2 -57 14 14 1750 46% -32 25% 25 Naum 4.6 -70 14 14 1750 45% -32 28% 26 PeSTO 2.210 -76 14 14 1750 44% -32 23% 27 Cheese 3.1.1 -81 14 14 1750 43% -32 28% 28 bright-0.4a -96 14 14 1750 41% -32 26% 29 GOOB 1.8.9 -123 14 14 1750 38% -32 26% Pool of 28 engines, average CCRL rating 3096, TC=4"+0.1, Little Blitzer testing framework STC Total: 49000 W: 14786 L: 19410 D: 14804 Elo: +8 vs Protej dev_20231129d
alexRemoval of certain data structures, replaced by calculation functions. Refinement of the outpost evaluation function. Introduction of the singular extension. Rank Name Elo + - games score oppo. draws 1 Deep iCE 4.0.853 x64/popcnt 97 14 14 1800 69% -40 30% 2 Senpai 2.0 94 14 14 1800 69% -40 30% 3 Texel 1.07 86 14 14 1800 67% -40 27% 4 SlowChess Blitz Classic 1.8 83 14 14 1800 67% -40 31% 5 Wasp 3.75 71 14 14 1800 66% -40 30% 6 Protector 1.9.0 53 14 14 1800 63% -40 29% 7 Hannibal 1.7 x64 50 14 14 1800 63% -40 28% 8 Counter 4.0 50 13 13 1800 63% -40 31% 9 Weiss 1.3 40 14 14 1800 61% -40 29% 10 Vajolet2 2.8.0 37 13 13 1800 61% -40 34% 11 Berserk 4.1.0 26 13 13 1800 60% -40 37% 12 Cheng 4.41 25 14 14 1800 59% -40 27% 13 Mr Bob v1.1.0 16 13 13 1800 58% -40 32% 14 Combusken 1.2.0 14 14 14 1800 58% -40 28% 15 Monolith 2 -5 13 13 1800 55% -40 33% 16 Invictus r382 -8 13 13 1800 55% -40 38% 17 Stash v28.0 -10 13 13 1800 54% -40 30% 18 Koivisto 64 4.0 -17 13 13 1800 53% -40 32% 19 Polaris 1.8.1 -17 13 13 1800 53% -40 33% 20 Tucano 9.00 -30 13 13 1800 52% -40 32% 21 Protej dev_20231129d -40 3 3 50400 44% 1 30% 22 Minic 2.33 -41 13 13 1800 50% -40 29% 23 Naum 4.6 -56 13 13 1800 48% -40 28% 24 FabChess v1.16 -59 13 13 1800 47% -40 31% 25 Winter 0.8 SSE4.2 -64 14 14 1800 47% -40 26% 26 PeSTO 2.210 -86 14 14 1800 44% -40 24% 27 Cheese 3.1.1 -93 13 13 1800 42% -40 32% 28 bright-0.4a -99 14 14 1800 42% -40 26% 29 GOOB 1.8.9 -117 14 14 1800 39% -40 27% Pool of 28 engines, average CCRL rating 3096, TC=4"+0.1, Little Blitzer testing framework STC Total: 50400 W: 14632 L: 20601 D: 15167 Elo: +28 vs Protej dev_20231118d
alexSearch functions: parameter tuning, code optimization, and improved execution speed. Rank Name Elo + - games score oppo. draws 1 Deep iCE 4.0.853 x64/popcnt 89 19 19 1000 72% -68 29% 2 SlowChess Blitz Classic 1.8 82 19 19 1000 71% -68 27% 3 Senpai 2.0 82 19 19 1000 71% -68 29% 4 Texel 1.07 81 19 19 1000 71% -68 27% 5 Protector 1.9.0 58 19 19 1000 67% -68 27% 6 Hannibal 1.7 x64 57 19 19 1000 67% -68 27% 7 Counter 4.0 57 19 19 1000 67% -68 28% 8 Vajolet2 2.8.0 46 18 18 1000 66% -68 31% 9 Wasp 3.75 43 18 18 1000 66% -68 30% 10 Berserk 4.1.0 41 18 18 1000 66% -68 35% 11 Weiss 1.3 32 18 18 1000 64% -68 29% 12 Cheng 4.41 23 18 18 1000 62% -68 26% 13 Monolith 2 15 18 18 1000 62% -68 31% 14 Combusken 1.2.0 4 18 18 1000 60% -68 29% 15 Invictus r382 3 18 18 1000 60% -68 34% 16 Mr Bob v1.1.0 3 18 18 1000 60% -68 28% 17 Stash v28.0 1 18 18 1000 60% -68 30% 18 Koivisto 64 4.0 -1 18 18 1000 59% -68 30% 19 Polaris 1.8.1 -12 18 18 1000 58% -68 33% 20 Tucano 9.00 -19 18 18 1000 57% -68 29% 21 Minic 2.33 -29 18 18 1000 55% -68 29% 22 FabChess v1.16 -60 18 18 1000 51% -68 33% 23 Protej dev_20231118d -68 3 3 28000 40% 2 29% 24 Naum 4.6 -68 18 18 1000 50% -68 30% 25 Winter 0.8 SSE4.2 -75 18 18 1000 49% -68 27% 26 bright-0.4a -81 18 18 1000 48% -68 29% 27 PeSTO 2.210 -87 18 18 1000 47% -68 25% 28 Cheese 3.1.1 -94 18 18 1000 46% -68 29% 29 GOOB 1.8.9 -125 18 18 1000 42% -68 24% Pool of 28 engines, average CCRL rating 3096, TC=4"+0.1, Little Blitzer testing framework STC Total: 28000 W: 7164 L: 12684 D: 8152 Elo: +11 vs Protej dev_20231108c
alexCode restructuring with grouping of types, constants, and global functions. Optimization of tapered scores handling routines. Rank Name Elo + - games score oppo. draws 1 Senpai 2.0 98 16 16 1450 74% -79 27% 2 Texel 1.07 93 16 16 1450 73% -79 23% 3 Deep iCE 4.0.853 x64/popcnt 81 16 16 1450 72% -79 27% 4 Wasp 3.75 76 16 16 1450 71% -79 27% 5 SlowChess Blitz Classic 1.8 62 16 16 1450 70% -79 29% 6 Protector 1.9.0 59 16 16 1450 69% -79 26% 7 Hannibal 1.7 x64 53 16 16 1450 68% -79 26% 8 Counter 4.0 50 15 15 1450 68% -79 31% 9 Vajolet2 2.8.0 48 15 15 1450 68% -79 32% 10 Weiss 1.3 46 16 16 1450 67% -79 27% 11 Berserk 4.1.0 38 15 15 1450 67% -79 33% 12 Mr Bob v1.1.0 21 15 15 1450 64% -79 27% 13 Cheng 4.41 10 15 15 1450 62% -79 26% 14 Monolith 2 2 15 15 1450 62% -79 31% 15 Combusken 1.2.0 2 15 15 1450 61% -79 28% 16 Invictus r382 -4 15 15 1450 61% -79 34% 17 Koivisto 64 4.0 -8 15 15 1450 60% -79 30% 18 Stash v28.0 -9 15 15 1450 60% -79 30% 19 Minic 2.33 -27 15 15 1450 57% -79 28% 20 Tucano 9.00 -27 15 15 1450 57% -79 30% 21 Polaris 1.8.1 -28 15 15 1450 57% -79 31% 22 FabChess v1.16 -52 15 15 1450 54% -79 33% 23 Naum 4.6 -56 15 15 1450 53% -79 28% 24 Winter 0.8 SSE4.2 -71 15 15 1450 51% -79 26% 25 Cheese 3.1.1 -73 15 15 1450 51% -79 28% 26 Protej dev_20231108c -79 3 3 40600 39% 3 28% 27 bright-0.4a -94 15 15 1450 48% -79 26% 28 PeSTO 2.210 -99 15 15 1450 47% -79 24% 29 GOOB 1.8.9 -115 15 15 1450 45% -79 24% Pool of 28 engines, average CCRL rating 3096, TC=4"+0.1, Little Blitzer testing framework STC Total: 40600 W: 9960 L: 19200 D: 11440 Elo: +11 vs Protej dev_20230919a
alex
alexRetest of version dev_20230919a. Rank Name Elo + - games score oppo. draws 1 Deep iCE 4.0.853 x64/popcnt 91 21 21 929 74% -90 19% 2 Senpai 2.0 86 21 21 928 73% -90 21% 3 Protector 1.9.0 80 21 21 928 73% -90 23% 4 Wasp 3.75 78 21 21 927 73% -90 23% 5 Counter 4.0 72 21 21 929 72% -90 22% 6 Texel 1.07 68 21 21 928 71% -90 21% 7 SlowChess Blitz Classic 1.8 66 20 20 928 72% -90 28% 8 Vajolet2 2.8.0 61 20 20 927 71% -90 26% 9 Weiss 1.3 60 20 20 927 71% -90 24% 10 Hannibal 1.7 x64 54 20 20 929 70% -90 23% 11 Berserk 4.1.0 31 19 19 930 67% -90 30% 12 Mr Bob v1.1.0 25 20 20 929 65% -90 20% 13 Invictus r382 21 19 19 929 66% -90 29% 14 Cheng 4.41 14 20 20 930 65% -90 25% 15 Combusken 1.2.0 -7 20 20 929 61% -90 22% 16 Monolith 2 -7 19 19 929 62% -90 27% 17 Stash v28.0 -11 19 19 928 61% -90 27% 18 Koivisto 64 4.0 -21 20 20 929 59% -90 23% 19 Minic 2.33 -28 19 19 929 59% -90 25% 20 Polaris 1.8.1 -34 19 19 928 58% -90 24% 21 Tucano 9.00 -39 19 19 928 57% -90 26% 22 FabChess v1.16 -68 19 19 929 53% -90 29% 23 Winter 0.8 SSE4.2 -70 19 19 927 53% -90 23% 24 PeSTO 2.210 -77 19 19 928 52% -90 23% 25 Naum 4.6 -77 19 19 929 52% -90 24% 26 Cheese 3.1.1 -81 19 19 930 51% -90 27% 27 Protej dev_20230919a -90 4 4 26000 37% 3 24% 28 bright-0.4a -92 19 19 930 50% -90 23% 29 GOOB 1.8.9 -106 19 19 929 48% -90 24% Pool of 28 engines, average CCRL rating 3096, TC=4"+0.1, Little Blitzer testing framework STC Total: 26000 W: 6513 L: 13157 D: 6330 Elo: +11 vs Protej 0.6.5 (recalculated)
alexSearch parameters tuned. Rank Name Elo + - games score oppo. draws 1 Mr Bob v1.1.0 214 16 16 1400 64% 112 21% 2 Tucano 9.00 175 16 16 1400 59% 112 26% 3 Cheese 3.1.1 122 16 16 1400 51% 112 25% 4 Protej dev_20230919a 112 4 4 28000 65% -6 21% 5 PeSTO 2.210 102 16 16 1400 49% 112 23% 6 bright-0.4a 96 16 16 1400 48% 112 22% 7 Berserk 3.3.0 83 16 16 1400 46% 112 29% 8 Quazar 0.4 x64 55 16 16 1400 42% 112 23% 9 Cheese 3.0.0 54 16 16 1400 42% 112 23% 10 Daydreamer 2.0.0-pre2 (2.0.0-pre2-dirty-97e5c32) 41 16 16 1400 40% 112 24% 11 Spike 1.4 7 16 16 1400 35% 112 24% 12 Leorik 2.3 -25 17 17 1400 31% 112 23% 13 Arminius 2018-12-23 -32 17 17 1400 30% 112 23% 14 Pawny 1.2.x64.SSE4.2 -34 17 17 1400 30% 112 21% 15 GreKo 2021.06 -39 17 17 1400 30% 112 20% 16 K2 v.0.99 -47 17 17 1400 29% 112 21% 17 odonata v0.6.2 -75 17 17 1400 25% 112 20% 18 Coiled 1.1 x64 -159 20 20 1400 17% 112 14% 19 Devel 3.8.4 -187 20 20 1400 14% 112 15% 20 The Baron 3.44.1 64-bits version -204 21 21 1400 13% 112 13% 21 SmarThink 1.98 -258 24 24 1400 11% 112 7% Pool of 20 engines, average CCRL rating 2910, TC=4"+0.1, Little Blitzer testing framework STC Total: 28000 W: 15196 L: 6973 D: 5831 Elo: +9 vs Protej dev_20230918a
alexEvaluation of attacked, defended and undefended pieces optimized. Rank Name Elo + - games score oppo. draws 1 Mr Bob v1.1.0 219 16 16 1400 66% 103 21% 2 Tucano 9.00 181 16 16 1400 61% 103 24% 3 Cheese 3.1.1 110 16 16 1400 51% 103 26% 4 PeSTO 2.210 104 16 16 1400 50% 103 21% 5 Protej dev_20230918a 103 4 4 28000 64% -5 21% 6 Berserk 3.3.0 100 15 15 1400 50% 103 30% 7 bright-0.4a 92 16 16 1400 48% 103 23% 8 Cheese 3.0.0 55 16 16 1400 43% 103 25% 9 Quazar 0.4 x64 53 16 16 1400 43% 103 23% 10 Daydreamer 2.0.0-pre2 (2.0.0-pre2-dirty-97e5c32) 44 16 16 1400 42% 103 24% 11 Spike 1.4 -6 16 16 1400 35% 103 22% 12 Pawny 1.2.x64.SSE4.2 -23 16 16 1400 32% 103 25% 13 Leorik 2.3 -37 17 17 1400 31% 103 23% 14 Arminius 2018-12-23 -41 17 17 1400 30% 103 24% 15 GreKo 2021.06 -42 17 17 1400 31% 103 20% 16 K2 v.0.99 -52 17 17 1400 29% 103 22% 17 odonata v0.6.2 -76 17 17 1400 26% 103 20% 18 Coiled 1.1 x64 -152 19 19 1400 19% 103 15% 19 Devel 3.8.4 -179 20 20 1400 15% 103 17% 20 The Baron 3.44.1 64-bits version -214 21 21 1400 13% 103 12% 21 SmarThink 1.98 -236 22 22 1400 13% 103 9% Pool of 20 engines, average CCRL rating 2910, TC=4"+0.1, Little Blitzer testing framework STC Total: 28000 W: 14829 L: 7184 D: 5987 Elo: +8 vs Protej 0.6.5
alexNew official release.
Includes all updates from dev versions from a to k:
Non-functional changes:
Rank Name Elo + - games score oppo. draws 1 Mr Bob v1.1.0 215 14 14 2000 67% 95 22% 2 Tucano 9.00 173 13 13 2000 61% 95 25% 3 PeSTO 2.210 128 13 13 2000 55% 95 24% 4 Cheese 3.1.1 109 13 13 2000 52% 95 27% 5 bright-0.4a 107 13 13 2000 52% 95 23% 6 Protej 0.6.5 95 3 3 40000 63% -5 22% 7 Berserk 3.3.0 88 13 13 2000 49% 95 27% 8 Cheese 3.0.0 54 13 13 2000 44% 95 27% 9 Quazar 0.4 x64 54 13 13 2000 44% 95 23% 10 Daydreamer 2.0.0-pre2 (2.0.0-pre2-dirty-97e5c32) 39 13 13 2000 42% 95 25% 11 Spike 1.4 3 13 13 2000 37% 95 26% 12 Leorik 2.3 -31 14 14 2000 32% 95 24% 13 GreKo 2021.06 -36 14 14 2000 32% 95 21% 14 Pawny 1.2.x64.SSE4.2 -37 14 14 2000 32% 95 23% 15 Arminius 2018-12-23 -50 14 14 2000 30% 95 22% 16 K2 v.0.99 -55 14 14 2000 30% 95 22% 17 odonata v0.6.2 -72 14 14 2000 28% 95 20% 18 Coiled 1.1 x64 -160 16 16 2000 18% 95 16% 19 Devel 3.8.4 -173 16 16 2000 17% 95 16% 20 The Baron 3.44.1 64-bits version -203 17 17 2000 15% 95 14% 21 SmarThink 1.98 -249 19 19 2000 12% 95 9% Pool of 20 engines, average CCRL rating 2910, TC=4"+0.1, Little Blitzer testing framework STC Total: 40000 W: 20680 L: 10610 D: 8710 Elo: +125 vs Protej 0.6.4
Rank Name Elo + - games score oppo. draws 1 Protej 0.6.5 74 6 6 10000 70% -74 30% 2 Protej 0.6.4 -74 6 6 10000 30% 74 30% Selfplay vs. Protej 0.6.4, TC=4"+0.1, Little Blitzer testing framework STC Total: 10000 W: 5550 L: 1491 D: 2959 Elo: +148 vs Protej 0.6.4
Rank Name Elo + - games score oppo. draws 1 Mr Bob v1.1.0 244 48 48 150 61% 158 20% 2 Protej 0.6.5 158 11 11 3000 71% -8 24% 3 Tucano 9.00 152 45 45 150 49% 158 37% 4 Berserk 3.3.0 144 45 45 150 48% 158 33% 5 PeSTO 2.210 114 45 45 150 43% 158 35% 6 bright-0.4a 88 46 46 150 40% 158 34% 7 Quazar 0.4 x64 68 47 47 150 37% 158 28% 8 Spike 1.4 27 48 48 150 32% 158 29% 9 Daydreamer 2.0.0-pre2 (2.0.0-pre2-dirty-97e5c32) 24 48 48 150 32% 158 25% 10 Cheese 3.1.1 23 49 49 150 32% 158 21% 11 Devel 3.8.4 5 49 49 150 30% 158 22% 12 The Baron 3.44.1 64-bits version 1 49 49 150 28% 158 26% 13 Cheese 3.0.0 -40 51 51 150 24% 158 21% 14 Arminius 2018-12-23 -55 51 51 150 22% 158 23% 15 GreKo 2021.06 -77 53 53 150 20% 158 20% 16 Pawny 1.2.x64.SSE4.2 -84 53 53 150 19% 158 21% 17 K2 v.0.99 -95 54 54 150 18% 158 21% 18 Coiled 1.1 x64 -105 55 55 150 17% 158 19% 19 Leorik 2.3 -118 56 56 150 15% 158 20% 20 odonata v0.6.2 -147 61 61 150 15% 158 12% 21 SmarThink 1.98 -328 92 92 150 5% 158 5% Pool of 20 engines, average CCRL rating 2910, TC=2'+1" (CCRL Blitz), Little Blitzer testing framework LTC Total: 3000 W: 1767 L: 527 D: 706 Elo: 3068 (CCRL Blitz scale)
alexProtej 0.1.1 vs 1000 CPP1 v0.1038 +30 =11 -19 35.5 - 24.5 Protej 0.1.1 vs 1074 LaMoSca 0.1 +6 =48 -6 30.0 - 30.0 Protej 0.1.1 vs 1306 T.rex 1.8.5 +21 =0 -39 21.0 - 39.0 Protej 0.1.1 vs 1379 S-Chess 1.42 +18 =0 -42 18.0 - 42.0 Protej 0.1.1 vs 1281 Cassandre 0.24 +0 =9 -51 4.5 - 55.5 Protej 0.1.1 vs 1337 Raffaela 0.11.0 +2 =2 -56 3.0 - 57.0 Protej 0.1.1 vs 1017 PreChess 0.7.8 +2 =1 -57 2.5 - 57.5 Protej 0.1.1 vs 1385 SharpChess 0.06 +0 =4 -56 2.0 - 58.0 Protej 0.1.1 vs 1168 Sinapse 1.0 +2 =0 -58 2.0 - 58.0 Protej 0.1.1 vs 1156 Pierre 1.7 +1 =0 -59 1.0 - 59.0 Protej 0.1.1 vs 1238 TheLightning 2.04 +0 =1 -59 0.5 - 59.5 Protej 0.1.1 vs 1277 Gaia 1.0 +0 =0 -60 0.0 - 60.0 Protej 0.1.1 vs 1206 Chad's Chess 015 +0 =0 -60 0.0 - 60.0 Protej 0.1.1 vs 1332 Pentagon 1.2 +0 =0 -60 0.0 - 60.0 Protej 0.1.1 vs 1132 FUSCH# 1.07 +0 =0 -60 0.0 - 60.0 Rank Name Elo + - games score oppo. draws 1 Gaia 1.0 309 301 301 60 100% -430 0% 2 Chad's Chess 015 309 301 301 60 100% -430 0% 3 Pentagon 1.2 309 301 301 60 100% -430 0% 4 FUSCH# 1.07 309 301 301 60 100% -430 0% 5 TheLightning 2.04 203 218 218 60 99% -430 2% 6 Pierre 1.7 199 218 218 60 98% -430 0% 7 Sinapse 1.0 133 181 181 60 97% -430 0% 8 PreChess 0.7.8 86 158 158 60 96% -430 2% 9 SharpChess 0.06 56 143 143 60 97% -430 7% 10 Raffaela 0.11.0 49 143 143 60 95% -430 3% 11 Cassandre 0.24 -62 106 106 60 93% -430 15% 12 S-Chess 1.42 -249 88 88 60 70% -430 0% 13 T.rex 1.8.5 -297 85 85 60 65% -430 0% 14 LaMoSca 0.1 -428 62 62 60 50% -430 80% 15 Protej 0.1.1 -430 31 31 900 13% 29 8% 16 CPP1 v0.1038 -497 77 77 60 41% -430 18% Pool of 15 engines, average ChessWar rating 1219, TC=1'+1", STC Total: 900 W: 82 L: 742 D: 76 Elo: 789
alex