[09:01] Join: Nenad joined #corewars [09:30] MSG: [10:26] Join: fiveop joined #corewars [12:53] Join: Nenad joined #corewars [12:53] ... ... [13:31] MSG: [13:50] Join: Nenad joined #corewars [14:09] * Nenad draws a summoning circle and starts chanting [14:10] thou who seeks the first and final warriors, come [14:10] * Nenad thinks it doesn't have any effect whatsoever [14:12] this channel has been really dead for a while. [14:12] we ought to change that soon [14:13] to the log readers: come to a sunday irc minitournament at 20h GMT this sunday, if you can. :) [14:21] * Nenad waves [14:22] MSG: [18:36] MSG: Read error: Operation timed out [19:43] Join: Core_old joined #corewars [20:53] Join: Nenad joined #corewars [21:02] hi Nenad [21:04] hi! [21:04] it seems that there IS someone alive here after all [21:06] always ready to serve [21:06] :) [21:06] well, with a bit smaller reaction-time now that i've done a little switching and my highlight-voice isnt on anymore [21:07] smaller -> slower [21:07] it has been a long time since there were 10+ people here and a nice chat&discussion [21:07] y [21:07] I hope that there will be a little more activity now that vacations are over [21:08] I've already said in the rec.games.corewar that I'd like to hold a tournament in November [21:08] but i hope that there will actually be poeple to participate [21:09] in the meantime... I think that I'll do something cw-related for my graduation paper :) [21:09] still a fw months to go, but it's good to think in advance [21:09] that post of yours looks horrible [21:09] ? [21:10] define: horrible [21:10] text for whole height of my browser with no separation [21:10] oh, that :) [21:11] well, wait to see my java code and you'll think that the text posted looks quite nice :)) [21:12] I don't like breaking lines into parts, so sometimes I even get a 500 character line... heh [21:12] (when there are some serious calculations involved, of course) [21:14] i hope my little probe on sal-b rises a little attention, and encourages people to ask from the community (i included mail-addy in strategy-line) [21:14] anyway, I plan to make a really pretty corewar tool soon (in a few months), but it'll probably be useless since there's noone to use it [21:14] it'll certainly have an advanced optimizer (hopefully much stronger than optimax) incorporated [21:15] and maybe also some options for evolving [21:16] but, it seems that there is some opposition to optimization at the moment [21:17] however, I have always felt that good warrior ideas deserve to have their potential fulfilled through the optimization process [21:17] otherwise, their true strength couldn't display on the hills [21:22] Oneshot at #1.. 'nop [21:23] heh :) [21:24] I am a little puzzled in it being only an S(s?)D-clear [21:25] maybe he meant to say SSD [21:25] but who knows :) [21:26] i would've expected, say, SMD [21:26] there aren't that many papers on the hill [21:26] well, yes, many scanners [21:26] why? i've tried it, probably the first [21:26] but it has a flaw [21:26] you can only do a good smd with antiimp mov, if I recall [21:27] and then you'll score bad against Bfield imps [21:27] papers are more and more often having imps [21:27] but, it is a good antiAimp weapon [21:28] yes, well papers need something to save them from clr [21:28] and antiImp clears are lareger, therefore score worse against everything except for imppapers and a portion of sbi and sabi [21:29] so they are not so often utilized (at least I haven't) [21:30] I have memory about there being a Mul-bomb really good against 3-point spirals, but i cant remember exactly [21:30] and last time i tried to search rgc, founded nothing [21:31] heh. seems like my oneshot does decently, as well... will try improving it a little bit in the next few days and resubmit, though... [21:32] Mul-bomb? hmm sounds interesting, but haven't seen it before [21:32] I think it was Mod, not mul [21:32] Memory hazy [21:33] well, what? MOD.F istep, istep ? [21:33] It was something like Mod >2667, >5334 [21:33] Part: Core_old left #corewars [21:33] yes :) that's even better [21:34] but looks hard for incorporating it into a warrior [21:34] paper would work, maybe in a stone [21:34] but oneshot-clear, nah [21:34] it isn't deadly for other chunks of code, only to imps [21:35] mov bombs are also not deadly per se, but can force a loop to overwrite itself with dats, maybe... [21:35] no chances for that in case of MOD [21:36] so my guess is that MOD paper would have too few wins against non-imp warriors to survive on the hill [21:36] hmm.. mov.i 1, >1 | spl -1, 1 ... [21:36] yes :) a nice one :) [21:36] used it a few times [21:37] i've had boring lessions is school, so i tend to either scribble on notebook or think about CW (or sleep) [21:37] there is such a thing as a not boring lesson? :) just joking. or not? most of are lectures tend to be boring [21:38] but not because of the subjects, more because the people are lousy lecturers... [21:38] There is a difference between boring lecture that i need to learn, and boring i already know [21:40] been there, done that [21:40] unfortunatelly... [21:40] but they are obligatory, so... [21:41] i have to say that i kinda like baconfist.. (the probe at sal-b) [21:43] is there an age limit there now? [21:43] yes [21:43] I don't remember killing Svarog there... oh... now it's clear :) [21:43] I've been worried that there is no acitivity at the hill [21:44] but I didn't get reports, since it was ;killed by the system [21:44] I should probably send some of my old beginner warriors there for the same purpose [21:44] or something else not too strong [21:44] qscan -> oneshot is also a nice idea [21:45] but Jens claimed that standard qscans do better [21:45] Ah, but that one uses tablescan so it is different [21:45] tablescan? [21:45] you mean, it doesn't have to recalculate? [21:46] or what? [21:46] something in form of sne *table, @table djn.f -1, -1 [21:47] that is 1c [21:47] when something is found, 10 dats and the clear is booted to b-found, and the clear's djn starts near a-location [21:47] qscans do 2c [21:47] but it looks nice, as well [21:47] probably much smaller [21:47] which can be a benefit [21:48] when something is found, it launche an SSD-clear in middle of A and B -locations [21:48] but you could've also made a 0,8c scan with a longer scanning run, maybe even a better option [21:48] why in the middle? [21:49] I had the table-scan ready, i didnt want to make something different [21:49] Because when the clear starts, the Spl's will clear on the B-field, and the djn will wipe A-field [21:50] Actually it is booted in b-field, but it boots 10 dat's with the clear [21:50] Hopefully instakilling something in the B [21:50] well, why didn't you use my two-way clear instead :) [21:50] Well, one could [21:51] That one was relatively fastly made, and the table-attack could be improved, yes [21:53] yes, but you don't want to be koth there, now, don't you? :) [21:53] did you try it at 94nop? [21:53] Didnt try at nop, but it wouldnt do well [21:53] It scores quite interestingly in quick benchmarking; [21:53] It scores under 100 points against wilfiz :) [21:54] but it wipes the floor with wilkies, over 150 points [21:54] why do you test againt those, anyway? [21:55] use nop_2 or nop_3 fsh from optimax [21:55] I use them as a quick-bench to see if my change has made a difference [21:55] :) every change makes a difference [21:55] A notable difference [21:56] Because once i get to the point of tweaking small changes, it is probably time to use optimax anyway [21:57] :) [21:57] yes, optimax is nice. [21:57] but has a huge problem - it's random [21:57] and the search space is usually huge [21:57] you want baconfist's source? [21:57] in papers and warriors with many values to optimize [21:58] it would be nice, if you'd like to share it :) [21:58] i see no reason not to spread it around since it is not special [21:58] sounds like an interesting warrior [21:58] adaptive strategies are always nice [21:58] I also experimented with scans that invoke different routines depending on what they think have found [21:59] paper dodgers, stone dodgers, regular oneshots... [21:59] http://koti.mbnet.fi/mizcu/temp.red [21:59] or maybe starting a paper if a bomb is located, or smt like that [22:01] 94b is really weak :( [22:01] weaker than I thought it would be [22:01] not surprizing [22:01] if a simple nonagressive replicator without imps can have 2.9% losses, it means that there is basically no scanners on the hill [22:02] Scanners are tough to make.. [22:02] oneshots - tough? well, good ones require some imagination, but you can make a decent one with the most basic scan loop and the most basic coreclear [22:03] we're not talking about HSAish scanners [22:03] beginners dont know oneshots [22:03] Join: codgeriff joined #corewars [22:03] (sigh) well, i certainly need to start making some more CE issues soon [22:04] hi Paul [22:04] (or am I wrong?) [22:04] hi, it's me. [22:04] :) another living cw_coder specimen. nice to see you here [22:05] thanks, I'm also reviewing the log for today to catch up. [22:05] how have you been doing lately? everything going fine, I hope? [22:06] well, still working on Shamu. changed the name to eliminate the apostrophe [22:07] at first it killed a bunch of stuff, but was clobbered by scanners [22:07] so now it's modified to kill them too, but not doing so well against others. [22:07] good to see some good oneshots. my favorite strategy, but haven't been coding anything lately [22:07] test is also a oneshot, but just something put together in a few mins. I'll try improving it later [22:09] oneshot is sort of a compromise between a qscan and a real scanner [22:10] you can do a fast coarse scan but will you get there in time to do any good? [22:11] oneshots aim is only to get close enough, jump to the coreclear (which is smaller and a therefore a smaller target) and hope to wipe it before it gets hit [22:11] ah, well. then I need a new name :-) [22:12] :) [22:12] than there is G2 which is a nasty twist to oneshotness [22:12] and Geist [22:12] yes, Geist in 5 lines. very nice endgame. [22:12] well, even Arrow, in a way [22:12] because of its small scan step [22:14] i'd like to see a good oneshot-> dclear + imps now. there hasn't been one in a while [22:14] since there aren't that many papers atm, it may even be the right moment for something like that [22:15] and with inversed's imps impervious to papers that could be good [22:16] there are also quite a few scanners on 94nop, so there's an opportunity [22:16] well, not that unusual, I think... I mean, I've also made some incendiary + imps combos before, but never as successfull as his Grunts... [22:16] at least I guess that that it what he is doing [22:16] (or did he publish it somewhere already?) [22:17] actually i think he doesnt use incendiary [22:17] not that I know of, but there is a clue in the irc log a while back about [22:17] the imps [22:17] SotBS and Grunts score so close [22:17] (scored) [22:18] so i know pretty well what the code is, but there might a naughty spin somewhere in it [22:19] P.K. : so, if I organize the november oturney, will you be interested in taking part? [22:20] btw, what is he referring to by mpttw in mag? [22:20] ? [22:20] always interested in a tourney, just time-limited [22:21] tell me about it... I've been time-limited for months, and probably will be until I graduate [22:21] I wonder if i should do some remaking on baconfist, it probably wont get on 'nop anyway, but just to experiment.. [22:21] why not? :) [22:22] Because it is vulnerable to qscans [22:22] what is baconfist's basic strategy? [22:22] tablescan/oneshot into bombing paper [22:23] i gave the code little before you came, just see the log [22:23] just boot your tablescan executable and have two short papers so both wont be killed [22:24] hmm [22:25] so booting two or three components, running one component and then jumping to another [22:25] tablescan is slower than qscan but about half the size too. [22:25] or leave the papers and hope they don't both get dat-bombed [22:25] qscan attacks are usually pretty grainy [22:28] the only use for tscanning is to kill qscans, it is just too coarse for anything [22:28] else [22:28] so, with your scanning engine safe elsewhere you should get a few qscan kills [22:29] but will he? qscans are usually a little bit separated from the code. until he copies away to attack, the enemy will probably already boot away [22:29] he needs a quicker attack to get qscans [22:30] ah, but what enemy attacks he's own qscan? [22:30] oh, a scanner, a stone, etc. [22:30] a paper. [22:31] by that time a paper is already staggering from the clear [22:34] maybe. [22:35] large decoy + 1c boot is usually good enough against qscans, imho [22:36] it keeps the pressure on them to simplify their attacks [22:36] otherwise they would be much more dangerous [22:36] mad mad idea appeared [22:36] yes, as long as they cannot be sure what their target is... [22:37] they can't afford to lose too much time attacking [22:40] MSG: Quit: Lost terminal [22:40] have to go now... but will be back tomorrow, probably :) [22:40] * Nenad waves [22:40] bye [22:40] MSG: [22:42] 100.09 points again fshnop03 [22:42] the current, unchanged/barely optimized version [22:43] does it clobber any particular classes? [22:43] I like to find something that kills one or two classes very well [22:43] then expand it to include the others [22:43] it does surprisingly bad against everything [22:43] it does least-bad against papers [22:44] probably a simple coding error :) [22:44] loses to almost all scanners, and against many bombers too - probably because of the relatively simple paper [22:45] launching two copies could probably give 5 points straight without tinkering [22:46] is it a bombing paper? [22:46] yes [22:46] i'm not sure a bombing paper works fast enough for today's scanners [22:47] quite true [22:47] the early cycles are spent replicating instead of bombing [22:48] pretty quick you get spl-wiped and have to undo that somehow [22:51] the whole thing would have to be reconstructed.. [22:51] ..to get on 'nop, that is [22:52] like i say, "it's a hobby, it's supposed to take time" :) [23:08] gotta go! [23:08] MSG: