[00:01] whats the highest score ever made against wilkies? [00:01] 200 or 210 [00:02] made by Fire & Ice (p-spacer) [00:03] oh, I had random data on. actually 53. [00:11] do you have ideas for improvement, or should I scrap this one? [00:11] throw the source of better version [00:12] well, i do have a semi-ready article for improving stones, but it might be a little bit too think read for now [00:12] http://pastebin.com/m1d9c77bb [00:13] instead of add.a #-1, bomb | jmp loop you need just one instruction [00:13] Does it make sense for an imp to tie a dwarf? [00:14] boaf: yes [00:14] Really? How so? [00:14] boaf: imp cannot defeat at opponent (though some warriors can be tricked into suicide) [00:14] But a dwarf should be able to, no? [00:14] because opponent has to run a dat to die, and imp's only copy itself [00:15] So because a dwarf has so many more opcodes to run through it takes longer and therefore at the end imp has more copies of itself? [00:16] imp does not copy itself [00:16] well, it does, but the meaning is different [00:16] Yeah, that's what I meant. [00:16] :P [00:16] @Mannerisk: http://pastebin.com/m3d81aa77 <- quick hack to reduce it to the very basics [00:17] elkauka: now now, you did not need to optimize it ready for him [00:17] boaf: when an imp overwrites the dwarf, the dwarf becomes another imp [00:18] Right. [00:18] boaf: dwarf can defeat an imp by reaching it by its bombing-run [00:20] however, since the bombing-run bombs only every fourth instruction, it is caught only about 1/5th of time [00:20] Ah. [00:20] (fifth, because there are cases where it cannot reach the imp) [00:21] then there is the special case [00:21] what happens when imp overwrites the dwarf's first instruction, and dwarf runs the add-instruction next? [00:27] MSG: Quit: http://www.mibbit.com ajax IRC Client [00:28] djn.a loop,$bomb [00:29] sorry elkauka, I don't understand yours yet :P [00:36] I get it now. what I don't understand is how you got that in 3 minutes [00:37] playing corewars for 6 years now ;) [00:37] year from now you can do the same [00:47] how many regular players are there? [00:47] only about a handful, dozen lurkers and dozen retired-but-still-following [00:48] i kinda miss fluffy and fizmo :( [00:50] players grow up faster than new players arrive [00:51] no one likes assembly when they can use Java [00:51] i blame doom for creating a generation that cannot play without audiovisual experience [00:51] my assembly class a couple years ago, I was one of a couple that refused to use macros if I could help it [00:51] so we need a mars that beeps every time an instruction is executed :P [01:03] 3am, time for coffee [01:04] my next goal is to have the program copy itself somewhere else after 1000 bombs and start again [01:04] I shall christen it the pole-vaulting dwarf. [01:04] i wouldnt suggest that [01:05] why not? [01:05] sure, you can do it, but itll score badly [01:05] would I understand the reason? [01:05] the code to copy the dwarf will be about as big as the dwarf itself [01:05] so scanners can find it? [01:06] well, since dwarf is already a bomber, it will not worsen the scores that much [01:06] but you spend time relocating while you could be continuing the bombing [01:07] I figured it would be cheaper than ensuring it didn't bomb itself once it gets down to a distance of 5 [01:09] the fact is that if you are reducing the MOD of your bombing-pattern, you will hit yourself later [01:10] I ran into that at 6 of my code and I'm sure elkauka's version does the same [01:11] no, not later, meant to write sooner or later [01:12] ...so change the offset [01:12] which requires more lines [01:13] and different bombing-steps throw different amount of bombs until returning back, so the counting will be problem again [01:15] I'm reasonably sure someone has done a variable-offset dwarf before [01:16] sure, but the only version that isnt worthless doesnt leave the main loop at any point [01:17] it seems to me that all versions of everything have been tried already [01:18] which is why the development of warriors has also slowed down, most players dont want to redo or reoptimize old things [01:22] this game needs a memory wipe. all the previous programs are deleted and everyone starts over [01:23] I think I might attempt to start some local beginner tournaments [01:25] Mannerisk: that's why i ended up creating templates of code snippets and have optimax figure out how to glue them together ... not much handwritting involved [01:25] + you end up with even more cryptic code: http://www.nomorepasting.com/getpaste.php?pasteid=17604 [01:26] and which is why I, Roy and Inversed despise optimax [01:26] what is optimax? [01:27] program that fight a warrior against a benchmark, but changes some numbers in between rounds, in order to find the best-scoring combination(s) [01:27] http://www.corewar.info/optimax/ [01:36] optimizing couple numbers, ok, putting every number into churn, like bringing calculator into 8th grade math-test [01:37] of course, if you still believe that Metcalf used optimizer to get the step-sizes on OOS, i wont wonder [01:38] seems to me like it could be expanded to also do instructions [01:38] pit it against every available warrior, and let it crunch numbers on a supercomputer for a few minutes [01:39] there isnt supercomputer strong enough on earth [01:39] heck, they've almost solved checkers [01:42] some quick figuring... 10 operations with about 4 modifiers, give it 8000 numbers to use and 20 lines of code [01:43] 6.4 million variations. you'd want to test against 100 programs 100 times each [01:43] 64 billion [01:46] if you can run a test in a tenth of a second it will take 30 years [01:46] but only 3 if you just pick the 10 best [01:47] but my desktop runs tests faster than that, so a supercomputer should be able to do it in a few hours [01:48] heh [01:48] bvowk has koeningstuhl #63 with purely evolved code [01:48] ;strategy This warrior is from a series of runs that has averaged [01:48] ;strategy approximately 80Gflops for nearly 4 months. [01:48] but do you see bvowk currently at top of KOTH? [01:49] I do not [01:49] why didn't it work? [01:50] the evolver didnt actually start creating any "smart" structures even during that time [01:50] and exhaustive search takes even more time/power than that evolver [01:51] it's probably a good thing. therein lies death by robots [01:52] also, he didnt want to "loan" the school's processing power anymore, so now he only uses a 14-processor system [01:56] no supercomputer here :( and the only thing that doesnt make any noise (so i can sleep) is an atmel atmega32l board. guess evolving is out of question for me :D [01:59] wilmiz-draft updated [02:03] link? [02:03] http://users.evtek.fi/~mikaos/bench.txt [02:15] doing deep analysis, or fell asleep? [02:19] moi? [02:19] doi [02:21] 70's show is on, got a lil distracted :P [02:32] bedtime anway, c ya [02:32] MSG: [02:38] time for me too [04:13] MSG: Ping timeout: 245 seconds [05:02] herm. [05:23] Join: Mannerisk joined #corewars [06:40] yuck, i forgot to kill the one [10:45] Join: cywhale joined #corewars [12:23] Join: asiekierk joined #corewars [12:23] hi [12:23] Hello [12:23] how may i be of help? [12:23] Nothing... yet. [12:24] first i must check the newbies start here URL [12:25] Checked [12:25] our main tutorial was nabbed my a domain-squatter, but here is a working link: http://web.archive.org/web/20071219024634/http://vyznev.net/corewar/guide.html [12:29] Heh. I made my first robot [12:29] before reading the tutorial [12:29] It's just an extended Dwarf, i must say. [12:30] A double-direction Dwarf. [12:31] With a single-direction dwarf, it's really random, but most of the time there's a tie [12:40] Well [12:40] if the core size % 7 == 0, THEN it works. [12:40] but i'll try it with a 7-space dwarf now [12:42] duhh, my seems to lose now [12:43] "When all other things are equal (which they never are), stone with smaller MOD wins" [12:43] MOD? [12:43] oh [12:43] MODulo. [12:44] MSG: Read error: Operation timed out [12:44] Now, i fixed it to use 8 (mine) and 4 (normal dwarf). [12:45] Why does mine win 80% of the time when there's no tie? [12:45] Weird. [12:45] Maybe because i use double-direction [12:45] and one direction carries DAT, the other carries SPL 0 (just for fun and testing) [12:45] you win by hitting the dwarf earlier than it hits you [12:45] :) [12:46] i could tell you how my code looks like [12:46] when stepsizes are this small, its still easy to deduct who hits whom first [12:46] i can probably guess already, but if you feel like showing, use a pastebin [12:47] well, not now [12:48] Well, i'm thinking to make a different warrior [12:48] It's a simple double-direction dwarf, easy to find out how it's made [12:48] It's just a dwarf with double ADD and MOV [12:49] accordingly modified [12:51] why do i need to send corewars robots thru email :( [12:52] because the one service where you didnt need to got shut down [12:52] mostly because its author didnt want to rewrite hes scripts every time sourceforge changed their rules about script-usage [12:55] 2565 4021 3414 [12:57] thats a lot of fights [13:00] 10000. [13:01] my bot vs. dwarf [13:01] thats a little bit excessive, since there are only 2620 possible starting-positions in standard settings [13:02] when benchmarking, i use 100 fights per match [13:02] 201 0 2419 - using 2620 rounds, me vs. Imp [13:15] MSG: Read error: Connection reset by peer [13:18] Join: cywhale joined #corewars [13:19] Join: asiekierk joined #corewars [13:19] list [13:19] oh [13:19] sorry [13:19] What, there's only one channel on here? [13:21] MSG: Read error: Connection reset by peer [14:38] Part: cywhale left #corewars [15:41] Join: elkauka joined #corewars [16:12] Join: Dwaddle joined #corewars [16:29] Join: Metcalf joined #corewars [16:29] Hi :-) [16:30] hi [16:30] Hi Mizcu [16:30] subtle change at Wilmiz, also each warrior has now scores against the benchmark [16:31] I've uploaded Ilmari's guide - http://corewar.co.uk/karonen/guide.htm [16:31] good [16:36] It uses a Creative Commons license, I just need to notify Ilmari [16:36] I'm not sure of the best way to fix the impring [16:38] depends on of Ilmari has a archived copy of its source, and the codes to the company that hosts the ring's [16:46] Join: OoS joined #corewars [16:49] MSG: Ping timeout: 245 seconds [16:58] the irc server has hit a new max user count (16) :D [16:59] Hi Elkauka [17:00] 2 people on IRC but not in #corewars [17:14] MSG: [17:15] you forgot the ;kill, AGAIN [17:15] christ what a mess, i KNOW i killed those [17:15] that was the main reason i even entered anything [17:15] screw it i'll just figure out this password crap [17:15] it didn't work when i tried it before [17:19] i did add ;kill and it's not responding to @kill wtf :( [17:19] killing doesnt take effect until next warrior is submitted [17:19] ah, need hillkey [17:20] sigh :\ [17:20] is hillkey "94b" or "redcode-94b"? [17:20] latter [17:24] Join: asiekierk joined #corewars [17:25] Hey; i'm back [17:25] Is there a tutorial on making better corewars robots than imps and dwarfs? [17:25] that's what i used, ok [17:25] also, can you only ;kill one warrior? [17:25] i found out what happened [17:25] you can ;kill multiple, just give line for each [17:25] the first time was a typo, the second time was ??? [17:25] hmm [17:25] ok [17:26] Hi asiekierk [17:26] asiekierk: not really, but feel free to start reading the Corewarrior -newsletter, which has a lot of interesting content [17:26] no space allowed? [17:26] space is allowed [17:26] that's the only other thing i can think of, there's a newline before the ;kills [17:27] well i guess i'll just wait and see if the @kills work [17:27] maybe make something else [17:29] MSG: Read error: Connection reset by peer [17:30] Join: asiekierk joined #corewars [17:31] MSG: Ping timeout: 245 seconds [17:33] Join: OoS joined #corewars [17:36] MSG: Read error: Connection reset by peer [17:37] Join: asiekierk joined #corewars [17:37] Hi [17:37] i found out my bot wasn't using double-direction later in the progress [17:37] Now, while sacrificing itself, it works fairly better against a mix of 5 different code files (one of them was made by mixing too) with RedRace [17:38] The mix had an imp, a dwarf, a robot i dont remember the source of, my bot's old ver, and later the output of the 4 i mentioned before too [17:38] Now mine is fairly better [17:38] since i'm using a different SPL, performing a full clear, and fixing up doubledirection, also, for kicks, i replaced DAT #42, #42 with a DIV 3, 0 [17:39] It was meant to be DIV #3, #0 [17:40] Out of 300 rounds, my Rofler scores 43, top.red gets 24, 233 are ties. [17:40] Anyone has a medium-difficulty robot with which i can test my bot? [17:41] "medium" is a bit of sliding term [17:41] www.koth.org/wilkies/ has a simple benchmark [17:45] with these, i have no change [17:45] chance* [17:46] that just means that you have to practise more [17:47] I'm using a "dwarf-spl mix" thing [17:47] i am not really sure if its smart to mix both spl's and dats in bombing [17:56] MSG: Ping timeout: 245 seconds [17:58] Join: elkauka joined #corewars [18:01] Join: OoS joined #corewars [18:02] there was a wiki once which collected strategies like the corewar lexicon (http://www.corewar.info/lexicon/). anyone got the url or in case its gone a backup somewhere? [18:05] back [18:06] elkauka: it was on pvk.ca and it's disappeared. [18:06] I haven't got a backup and neither has pvk [18:06] http://corewar.co.uk/biblio.htm might be useful [18:08] There's no backup in the Internet archive because it's blocked [18:09] time to start a new wiki then? [18:11] MSG: Read error: Connection reset by peer [18:12] Join: asiekierk joined #corewars [18:13] MSG: Ping timeout: 245 seconds [18:18] Join: OoS joined #corewars [18:23] MSG: Ping timeout: 245 seconds [18:26] Join: OoS joined #corewars [18:30] I don't know if anyone would be interested in editing a new wiki? [18:38] MSG: Read error: Connection reset by peer [18:41] MSG: Ping timeout: 245 seconds [18:44] Join: asiekierk joined #corewars [18:59] MSG: Read error: Connection reset by peer [19:07] Join: asiekierk joined #corewars [19:07] :/ [19:07] I wonder if there's space for #corewars innovation [19:08] i mean [19:08] Core Wars innovation [19:12] MSG: Ping timeout: 245 seconds [19:14] MSG: Read error: Connection reset by peer [19:15] probably [19:16] but youd have to try hard! [19:16] * evitable talks to empty space [19:16] i have a thing i want to try hehe [19:18] Join: elkauka joined #corewars [19:20] diffrent hills can force that a bit, like nano of fortress did. i would like to see a hill with max cycles way closer to coresize [19:20] id prefer higher.. [19:20] like 8192/100000 [19:21] 100/200 :P [19:33] i had an idea for a possibly interesting MARS modification actually [19:33] i didn't think it through much though yet [19:34] i was thinking you could "link" two mars machines together by having a common stretch of core [19:34] like, 100 instructions long or something [19:34] then you could make a whole "network" to take over :D [19:34] also it'd involve strategy changes since you'd have to find and access the other machines before you could even attack whatever may be loaded in them [19:40] quantum corewars allows it already [19:58] ?? [19:58] never heard of that [20:00] http://bio.freelogy.org/wiki/Coreworld_TAC [20:11] finally, took care of it [20:15] thats an interesting scoring [20:15] it was an idea i wanted to try, but i haven't spent as much time as i should on it [20:16] the hill mess was annoying me :P [20:16] it throws flaming imp bombs! ;) [20:22] Join: OoS joined #corewars [20:30] MSG: [20:39] i've been playing around with optimax the last couple days too [20:39] is there a way to seed it with already chosen values? [20:39] for comparative purposes [20:41] i mean like, instead of starting off random, start with one that already works; or throw it in the list at least [21:14] MSG: Ping timeout: 245 seconds [21:35] hm it's about time i tried a scanner [21:37] either i don't understand the hill well or i somehow ruined the warrior i was trying to improve [21:37] originally i just wanted to add a decoy to see how it'd affect its score vs crackhammer [21:37] but it didn't even make it back to #2 [21:37] randomness maybe? [21:38] unlikely [21:38] how far you are from the decoy? and before or after? [21:43] you can try optimize the distance against the plain crackhammer itself, since the source is available at my page [21:50] boss had my attention [21:50] heh [21:51] ok. [21:51] it's not crackhammer [21:51] i mean, the first 'decoy' was just a simple for loop for 50 instructions [21:51] and that version did the best [21:51] then i tried adding a boot away so i could run it through optimax to find a good distance [21:51] (maybe that's part of it -- longer delay before it starts attacking) [21:52] but even the first versino with a decoy scored #3 [21:52] anyway they all seem to do decent vs that one, but in general they are worse [21:52] crackhammer is really the only warrior that cares about decoys on the hill [21:52] yeah, i'm not surprised [21:54] anyway, i learned stuff, and that's what counts [21:54] i haven't tried to make a #1 winner yet, but i though maybe if i could tweak the one which already had #2 a little it might put it over the top [21:54] right now i'm running bolo through optimax and seeing interesting results [21:55] i picked the original steps just for fun [21:55] so it actually sorta looked like a bolo hehe [21:55] but those aren't very effective values [21:55] interestingly enough the very first randry is still at the top of the list [21:56] i think i pasted that one in here once, it had what i thought was an interesting idea [21:56] i haven't heard of any paper-ish warriors like it at least [22:02] i dont recall the source right now [22:17] i'm insulted! :P [22:17] i expect everyone to remember all the code i paste [22:18] (1sec) [22:20] http://www.nomorepasting.com/getpaste.php?pasteid=18834 [22:22] thats more of a hopper really since it doesnt make new copies [22:23] ok then [22:23] hadn't heard that word [22:24] i just liked the synergy :) [22:26] there is nothing bad in it really, just add a spl in it, and change the jmp into jmp top, {ptr (or modify the add so that it modifies the ptr into 0, newnumber) [22:27] that's the core idea, i was playing with tacking on something to the end as a fall through [22:27] which is the version on the hill, but its very broken [22:27] how long you are still in the work? [22:27] ? [22:28] i don't understand the question [22:28] i'm not sure i want it to copy actually [22:28] also it depends on process timing [22:28] < evitable> boss had my attention [22:28] oh [22:28] ha, hours yet [22:29] that's why i'm replying intermittently [22:29] while the machine is running [22:29] ill see in couple hours if i can make a couple not-too-releaving tips for the bolo [22:29] revealing [22:30] ha [22:30] sure [22:31] i know it is pretty fragile [22:31] i keep trying to decide to try and make a "winner" but i get sidetracked by random ideas [22:31] which is, to me, more fun [22:32] (so i haven't spent much or any time optimizing or improving anything yet, it'll be good to get some practice at that too) [22:42] I like the quantum coreworld idea [22:42] has it been implemented anywhere? [22:43] i think there is a version of MARS that can run it, but ive never tested myself [22:45] i think it's not quite like i meant [22:45] but i haven't gotten to read it in depth yet [22:49] it is different [22:49] instead of a program finding a link between nodes, more programs are created automatically I think [22:50] but into other worlds with different parameters [22:50] then the article starts getting into quantum states and flies over my head [22:50] lol [22:50] well 'finding' the link isn't so much the goal -- you could easily brute force it [22:50] the goal would be to take over ALL the MARS machines [22:50] instead of just one [22:50] or end up with the most [22:51] or something [22:51] it does seem like the next step in the game [22:53] the interesting part comes with network architecture and taking over machines [22:53] each machine would run in parallel, independently [22:53] so if you have two machines and the enemy has one, you run twice as many instructions per turn [22:53] though it doesn't really help if there's only one link i guess [22:53] but you could have things connected like a mesh, or through a central hub [22:54] with the central hub you'd have multiple points of attack and a single point that is a choke point for accessing the others [22:54] etc [22:54] and you wouldn't really have to change redcode [22:54] except perhaps some command to launch a process on the "alternate" server [22:54] spl could perhaps be hijacked for that purpose but that doesn't sound like a clean solution [22:57] Part: Dwaddle left #corewars [23:02] it's been 14 years since the last standard. about time for another one [23:02] http://xkcd.com/350/ [23:03] perhaps [23:03] a lot of things get worse when people try to make them better [23:03] tetrinet, for example [23:03] (tetrinet 2.0 yuck) [23:03] i think strategy games tend to have simple rules and complicated combinations [23:03] making complicated rules doesn't help [23:03] Go, anyone? [23:04] (i'd play but i don't have the sequential time to play, also i'm a noob -- barely know the rules even) [23:04] there still isn't a half-decent AI for that game [23:04] I've picked it up through a facebook app of all things. The game was invented thousands of years ago, and I'm playing it on web 2.0 [23:04] not surprising [23:05] the ai bit, i mean [23:05] i want to go to a go parlor [23:05] the rules are evens simpler than chess, yet mediocre players beat the best AIs [23:05] or maybe join som sort of club [23:05] and learn [23:06] seen that go app that lets you play on arbitrary node meshes? :) [23:06] you can play on mobius strips and spheres and stuff [23:06] that could be fun :D [23:10] I'm totally going to do my senior thesis on Netwars [23:11] haha [23:11] fuck, if you do you better mod pmars for it [23:11] so we can play [23:27] unfortunatly my grad thesis is 6 years away ;) [23:28] I guess it will be an early start