From news-rocq.inria.fr!jussieu.fr!oleane!plug.news.pipex.net!pipex!tank.news.pipex.net!pipex!usenet.eel.ufl.edu!usenet.cis.ufl.edu!caen!stimpy.us.itd.umich.edu!not-for-mail Mon Mar 25 11:33:23 1996 Article: 4526 of rec.games.corewar Path: news-rocq.inria.fr!jussieu.fr!oleane!plug.news.pipex.net!pipex!tank.news.pipex.net!pipex!usenet.eel.ufl.edu!usenet.cis.ufl.edu!caen!stimpy.us.itd.umich.edu!not-for-mail From: afabbro@stimpy.us.itd.umich.edu (Andrew Philip Fabbro) Newsgroups: rec.games.corewar Subject: Help My Poor Scanner Date: 25 Mar 1996 05:55:40 GMT Organization: University of Michigan Engineering, Ann Arbor Lines: 68 Message-ID: <4j5cgs$peh@srvr1.engin.umich.edu> NNTP-Posting-Host: stimpy.us.itd.umich.edu X-Newsreader: TIN [UNIX 1.3 950824BETA PL0] This is the first warrior I've submitted for group diagnosis, so please be gentle...I'm merely a Young Grasshopper. Antibodies is basically a JMZ.F scanner using incediary munitions, with a small twist. Rather than avoid itself by the usual SLT/JMP back method, it uses a mod-8 scanning pattern and hides in the spaces. I'd also hoped that it would be faster than the SLT/JMP avoidance method because (1) it wouldn't have to spend time executing the SLT/JMP when it found its own code, and (2) it wouldn't have to spend time executing the SLT when it found the enemies'. Its basic plan is to stun and then kill with a core clear. I miss stones (as expected), win about 33% vs. other scanners, and fail miserably against modern papers, silks, and recent replicators. Against older papers (e.g., the classic Mice), I clean up, but not against their descendents. Frequently, the enemy is stunned and my cookbook coreclear (which wraps through core several times) reduces the replicator to only a few processes...but it rebuilds itself, resulting in a tie or a loss for me. I've watched my warrior reduce Marcia Trionfale to only 3 active processes and then lose as Marcia rebuilds and swamps me. Any suggestions on nailing the coffin tight enough that the occupants don't clamber back out? Timing the jump to core-clear is also proving tricky. At the moment, it transitions after making 200 bomb runs...using 400 instead works better against well-bred papers, but makes it prone to fail to core clear (and therefore tie). All in all, it scores about 95 on the beginner hill, so I think it has some potential. I rewrote it as a mod-4 and it performed worse against papers with no change against other types. Suggestions? ;redcode-94b ;name Antibodies ;author Andrew Fabbro ;version Mark I ;assert CORESIZE==8000 radar ADD #2936, #3999 ; mod-8 scanning JMZ.F radar, @radar bombrun MOV bomb1, >radar MOV bomb2, @radar adjust SUB #1, radar ; restores pointer to mod-8 DJN radar, #200 JMP cclear marker DAT #0, #0 ; **************** bomb1 SPL 0, 8 bomb2 MOV -1, <-1 cclear SPL 0, 0 ; standard recipe core-clear MOV marker, <-1 JMP -1, -1 END radar Andrew Fabbro | "Millihelen, n.: Amount of ITD Marketing Research | beauty necessary to launch afabbro@umich.edu | a single ship." afabbro@katana.ucore.itd.umich.edu |