[08:05] nice irc network [08:09] ... [08:18] Join: yoR joined #corewars [08:18] java [08:18] err... [08:18] why not mocha? [08:18] wrong window :( stupid popup-irc-window [08:19] Was typing /join ##java in another window, but for some odd reason TinyIRC decides to switch instantly to windows that just connect [10:18] Join: mooskagh joined #corewars [10:18] evening [10:21] hi [10:23] Mizcu, where are you from? [10:23] europe [10:24] I can guess this from your timezone. :) [10:24] Which country? [10:25] well, why dont you try to find out? (takes ~0.5 secs..) [10:25] ok :) [10:25] but anyway, what brings you here? [10:25] ah, even less than 0.5 :) [10:27] I wonder if there are any corewar championships are up now, and if there are anything new in corewar since 1990s [10:27] proper 'championship' -contests are few and rare, theres none going right now [10:28] though hills are 24/7 active [10:28] and 'new', not really, just silent improvement of code and few hills with new settings [10:29] unless by 1990s you mean before 94 [10:29] It has slowed down a bit lately, was much more active about 3-4 years ago [10:30] nope, I mean middle 1990s :) And I cannot find any good replacement to corewars. I tired robocom, but it seems more boring and stops developing as well. [10:30] gridwars died out real quick too.. [10:33] That not necessary should be something assembler-like.. But if it uses higher-level languages, i'd prefere use existant language rather than something designed specially for the game (usually pour subset of C). [10:34] somehow i have the idea that using C for a programming game always ends up stupid [10:34] Roy, correct me? [10:35] yep, i agree with it. [10:35] at least I didn't see any nice one. [10:35] Well Mizcu, I really liked the Roshambo competition, it was in C [10:36] But the year I wanted to join...they stopped holding it :-( would have probably be in the top 5 [10:38] though world is saved from VisualbasicWars and Javawars. I think. Or rather, i wouldnt even want to know. [10:45] Ah, roshambo is nice. I wanted to participate as well. But as all ideas from past years are open, I think all ideas would be exhausted pretty soon. [11:32] Thats a bit naive :) [11:33] Its all about reactive pattern recognition, a very import field of computer science today (people are trying to predict everything, from the weather to earthquakes, stockmarkets to hurricanes) [11:34] dont forget data compression [11:35] Ah yes, and what Mizcu says :) complex prediction algorithms do well in data compression (especially the hybrid ones) [11:35] pattern recognition in computer sight too, but it not as big [11:36] The bigger problem in computer sight is object recognition [11:36] But thats kind of related [11:38] :) in any case roshambo has too few variants to select from, so the time when this problem will be completely studied would come soon. [11:42] No... there are very limited choices indeed, but thats a good thing, forces you to learn to understand the opponent [11:42] If you grasp the algorithms the opponent uses, you always win [11:44] What if you grasp that he uses completely the same algorithm that you use? :) [11:46] then you die [11:46] (because then you fight yourself, and 'he' gets all the scores for wins, instead of ties) [11:46] and in that game there's no better or worse algorithms. All you can do is to write counteralgorithm against all known algorithms. [11:48] Tic-Tac-Toe problem [11:50] tic-tac-toe is completely another kind of problem as it has non-losing strategy [11:51] and isnt counteralgorithm to all algorithms such? [11:52] in tic-tac-toe? there's no such courteralgorithm against the best algorithm. [11:53] Well no, rock-paper-scissors problem, not tic-tac-toe [11:53] Tic Tac Toe is the 3-in-a-row game Mizcu [11:54] Roy: i actually wikipediad the name because is wasnt 100% sure [11:54] rock-paper-scissors is the same as roshambo we were talking about, isn't it? [11:54] Yes, but the intresting thing is, all beats all (1>2, 2>3, 3>1) [11:55] in ttt, once you know the game well enough (eg. all scenarios of a game), you can play always the best move (unless youve done bad moves beforehand), once youve analyzed all strategies in rps, cant you not play a game with only good choices? [11:56] so every algorithm will win 0.5 of all other algorithms. (if not count draws) [11:56] No, rps always depends on the opponent, if the opponent switches strategy you have to adept [11:56] (in ttt, the center tile has higher value than others, so it remains debatable, and rps has only equivalent choices) [11:57] (it remains debatable it the strategies between the two are comparable) [11:57] rps (when played in sequences) is infinit, there is no real end, so you can't calculate a perfect solution [11:57] in talking myself again into a bag [11:58] if there are infinite problems, there are infinite solutions [11:58] in rps all solutions are equal. Any unequality appears when you know anything about your opponents. [11:59] Mizcu: Indeed, thats the whole problem, there are infinite solutions, so the problem is making one that understands most solutions other people made and beating those [12:00] The only problem is this: Once you loose you can go into random play, random in rps means you tie... (unless the opponent can calculate the random function) [12:00] so thats inf * inf -problem, which is still equal in size to making infinitely long strategy [12:00] (ok, in lacking in mathematical theory, so im a little on thin ice) [12:01] in that contest there were 1000 rounds for each fight between two programs. [12:01] And thus... There are (3^2)^1000 possible solutions. [12:01] Yeah, but that doesn't really matter, the only thing that matters is reading/understanding the opponent moves, if it tries to do something logical you could (in theory) counter it [12:03] this is why we need more newbies, finally discussion that has challenge [12:03] Hehe, we should make our own RPS competition [12:03] Neo's challenge again? [12:04] Well yeah, it has been done [12:04] thats not a fault [12:04] But would be fun to incorparate it with RedOS [12:04] ..what? [12:04] The IO-mars, writing your decision to the IO [12:05] i started thinking the RedOs from RedFrenzy [12:05] The fun part is: In CW you don't have much memory to store all the past moves, neither is there a good random function, you have to make a rational decision, making yourself vulnarable to counter-logic [12:07] how many 1-2-3 choices can you save in 8000? [12:08] What's CW? [12:08] Core Wars [12:08] ah [12:09] I think pseudorandom function is easily implementable. [12:10] 8000 bytes? you can store log[3](256^8000) = 40379,504228573275974369735317937 values of 1..3 [12:11] no, in a value of 0-8000 [12:12] ah, last time I played core wars was around 1998, so don't remember anything. :) [12:12] standard coresize is 8000, so there are no higher values [12:12] so you are limited to values 0-8000 per datafield [12:13] memory has 8000 cells each of them can store value from 0 to 7999, it that right? [12:13] yes [12:13] 8000 choices, not 8001, you are correct [12:14] So each cell can store log[3](8000)=8,1804990835825261241882839090925 values of (1..3) [12:14] so that 128000 previous fights minus warriorsize [12:15] (plus minus p-space, if used) [12:15] But it's hard to use that ".1804990835825261241882839090925" with so pour language. :) [12:15] So let it be 8 per cell [12:15] The trick would be: how :P [12:16] the same way as you pack numbers in binary, except now you use trinary [12:16] *bonk* [12:16] yep. [12:16] Yeah, but in code, thats a bit harder ;-) [12:17] my heads a bit woozy, looks like i havent drinked that ED next to me, time to take a sip [12:18] well, Lukasz implemented a three-dimentional array in redcode when the last RPS contest was done, so it just takes a bit of thinking [12:20] three-dementional array with each value is stored in one cell? [12:21] looks like Neo hasnt reuploaded the warriors on the new website, so i cant check the source [12:22] (i never analyzed the switchers, other than seeing Neo use add #2667, tactic to notice brainwashing) [12:23] http://labarga.atspace.com/mc6.html [12:23] http://groups.google.com/group/rec.games.corewar/browse_thread/thread/d5c407bd4571544d/b247d3018acc2c2c?lnk=gst&q=PointA+Rewritten#b247d3018acc2c2c [12:23] theres the code an explanation [12:32] Calculating trenary isn't very hard in CW [12:34] decode equ 8 [12:34] loop mov.ab ptr, @ptr [12:34] mod #3, >ptr [12:34] div.a #3, ptr [12:34] jmn.a loop, ptr [12:34] ptr dat #decode,2 [12:34] end loop [12:34] Decodes the value 8 into : dat 0,2/dat 0,2 (22 ternary) [12:35] 653 (for example) becomes: 220012 [12:35] Thats how you can store multible rps numbers into one variable [12:46] have you heard about icfp programming contest? [12:47] Uhh no? What is it? [12:49] it's annual programming contest which takes part in summer. The most succesive one was in 2006. The teams needed to write the interpreter for very simple virtual machine language. And using this language they ran the code which was given by organisators. [12:49] And that code was very-very impressive. They written unix-like operating system using that language and some compilers as well as other applications. [12:49] Hehe thats always fun [12:50] I sometimes join the recmath competition, also a lot of fun [12:50] And participants needed to reseach that code and solve some tasks and look for some secrets. [12:51] http://www.boundvariable.org/ here the page about icfp 2006 contest. [12:51] You can check "technical report" and "slides" to have an idea what it was like [12:51] but all participants were very-very impressed. [12:52] the contest lasts 3 days, and all participants usually don't eat and sleep all three days. There's no limit in how many people can be in the team. [12:53] recmath?.. With one task given for a long time, and participants are to write some heuristical solutions? [12:54] Yeah, fun mathematical problems [12:55] I like Topcoder Marathon matches, which are very similar, but more people are active there, and tasks are usually more interesing. One match usually takes 1 week, but sometimes there are 2 week matches. [12:55] http://www.topcoder.com/ [12:56] I have no time for Topcoder... pfft, the concept is nice, I should have known ten years ago [12:58] http://www.boundvariable.org/press/icfp2006web.swf - that can give a very good idea what icfpc 2006 was like. 2007 was nice as well. "the DNA of the alien" was the code, and we needed to research it as well. [12:59] But it was too hard for most teams to do it within 3 days. [13:00] I'm encouraging my manager to form a program-competition group within our company, most big companies have these teams that compete in national/international competitions [13:04] afk brb [14:22] well, yay [15:27] yay! [15:27] (why?) [15:29] i say afk, and everyone goes quiet [15:29] Sure, we listen if you command us [15:29] well, anyone, friend asked how much a 4/4 SDSL would be from the largest finnish ISP. 370e/m and 900 bucks for install [15:29] anyone -> anyway [15:30] we did a little scouting around and found one for 199e/m w/ 90 buck install [15:30] That sounds kind of cheap in a way.. [15:31] considering you can get a 24/2 adsl2+ for hundred buckaroos, 370 is kinda ripoff [15:32] but damn, that 199e is not bad at all for 5/5 (Swedes and people at Oulu, STFU) [15:33] MSG: Quit: dat [18:56] Join: Metcalf joined #corewars [18:56] Hi :-) [19:01] Mizcu: mc6 warriors aren't available on the old site either [19:02] Some of them are available on Koenigstuhl [19:06] MSG: Ping timeout: 245 seconds [19:48] Join: fiveop joined #corewars [20:50] MSG: Ping timeout: 245 seconds [23:00] Join: onerc joined #corewars [23:16] MSG: Quit: onerc