Date: Fri, 8 Mar 1996 03:16:26 -0500 From: jklewis@ren.us.itd.umich.edu (John K. Lewis) Subject: Tourney Results!!! Intelligent Tournament Results! After several hours of hard computer crunching the results of the first I.T. are in. Kurt Franke did an amazing job programming the winner System Trap. Talking with Kurt over e-mail, I got the impression he was very pleased with his handshaking system. He would have won even without it. Lets take a look at the challenge once again. The rules were fairly straight forward, they were exactly the same as the 94 hill on Pizza, with one small exeption: RedOS. The RedOS is a simulated operating system which resides somewhere in core with the two combatants. Should the RedOS die, for any reason, the computer will "crash". Scoring goes something like this; Points | Condition 3 Program kills opposing program and kept OS alive. 1 Program, opponent, and OS live. 0 Program dies, opponent and OS live. -3 OS dies. Kurt's program did a scan for the RedOS, then it would vamp the process and start a 1.5c coreclear (it's got another programs process helping it.) This worked amazingly well against stupido. Stupido basically went for the draw. It tried to force the RedOS into becoming an imp. While stupido forced most opponents into a draw, it almost always died against System Trap. The reason was the 1.5c coreclear. Once System Trap has the coreclear, it was pretty much over for stupido. Kurt wasn't the only one thinking about vamping the RedOS. Faster Finder by JS Pulido also attempted to vamp the RedOS. Once it got ahold of it, the OS turned into a paper and Fast Finder started bombing. Unfortunaly for Fast Finder, System Trap usually found RedOS first. Even when Fast Finder did get ahold of the OS first, System Trap's backup strategy stumped it. If System Trap couldn't find the OS, it turned into a paper. Specwar had a similar strategy to stupido, try to make the OS an imp and go for ties. The major difference was that Specwar did a scan for the OS while stupido just bombed with mov 0,1. Brian Haskin's viralOS used a copy of RedOS as a decoy. It was also tuned not to kill the real RedOS, but instead to bomb only other targets. I thougth that the decoy would have fooled System Trap, but upon closer inspection I noticed a very intensive test to see if the RedOS was the real one. Ape also used a copy of RedOS as a decoy. In a similar move to the others, it also tried for the tie. Hoping to hit the RedOS with an imp, and then set up a gate that a fast imp could get through. A few notes on the results. Several programs mentioned in thier strategies that they were not "intelligent". I think the scoring below shows the results fairly clearly. Kurt's program was the maximum length of 100 instructions. Every inch optimized for intelligence and strategy. I will be holding a second tounament next month and will post rules for that contest in rec.games.corewars as well. It will probably be exactly like this one, but I need input on what people want. I hope more people take part because this seem like a very interesting type of competition. Congratulations to Kurt and all the other participants, I hope you enjoyed it enough to enter the next one. :-) Table Data 1. 2. 3. 4. 5. 6. total 1. System Trap 600 265 230 530 80 286 1991 by: Kurt Franke 2. Fast Finder 13 291 119 476 -62 -77 760 by: JS Pulido 3. Specwar 140 83 165 182 0 132 700 by: J E Long 4. stupido 11 62 182 200 0 200 655 by: Nandor Sieben 5. viralOS -53 -44 34 261 188 88 474 by: Brian Haskin 6. Ape 37 -155 132 200 100 45 359 by: Calvin Loh *** If you are mystified by the scoring... you're not alone. Here is a quick guide: *** 1. Run pmars with the following command: pmars RedOS warrior1 warrior2 -r 200 2. Output will come up as follows: RedOS by jkl scores 548 Results: 0 102 70 28 warrior1 by player1 scores 348 Results: 20 12 70 98 warrior2 by player2 scores 564 Results: 5 96 70 29 To determine the score for warrior1 use the formula ((warrior2 loses - warrior1 single wins) * 3) + ( three warrior wins ) + (RedOS loses * -3) John K. Lewis ^^^^^