[07:34] Join: Jens joined #corewars [07:35] Part: Jens left #corewars [08:34] Join: brx_ joined #corewars [08:36] MSG: Ping timeout: 252 seconds [08:36] Nick Change: brx_ changed nick to brx [09:51] Join: fiveop joined #corewars [10:14] "But finally I realized that to me, Goedel and Escher and Bach were only shadows cast in different directions by some central solid essence. I tried to reconstruct the central object, and came up with this book." [14:15] Join: Jens joined #corewars [14:15] :) [15:03] Join: Roy joined #corewars [15:03] brx: So you are reading GEB? :) nice book! [15:03] Roy: you have read it? :) yes it is nice [15:05] Well, I think I did read it, a while ago [15:06] I know for sure I've lend it from the library [15:06] you think you read it? [15:06] Well, not sure anymore, just googled some stuff, and most sounds familiar, but I read a lot of books.. [15:06] the crab canon is nice :-) [15:07] does he get back to the crab canon later in the book? [15:07] I have just started and am currently at the formal system intro [15:08] No idea.. this discussion is fun: http://www.barryland.com/canon.html [15:09] ah no have not read it yet [15:09] so I won't spoil it by reading the website [15:09] :) [15:10] No you shouldn't :) [16:03] MSG: Read error: Connection reset by peer [17:03] Join: Nenad joined #corewars [17:03] Hi Nenad :-) [17:03] Hi Jens! [17:04] Is KryneLamiya now your oldest warrior ever? [17:04] yes. :) [17:05] Congrats! [17:05] thanks. it lived much longer than expected [17:05] btw, congrats on the Harmless Fluffy Bunny [17:06] You should ask JM, what is the oldest oneshot for 94nop. Maybe KL is already better :) [17:06] it isn't :) [17:06] Thanks [17:06] Toxic Spirit was very old [17:06] 600+ [17:06] Then KL has to live a little bit longer [17:06] CrazyShot2 was 211... [17:06] but not many oneshots lived long [17:07] but it's probably hardest to make a good stone today [17:07] even Beat This got beaten [17:08] Hmm, but that was an accident [17:08] papers are getting stronger and stronger thanks to optimax... [17:08] and other optimizing tools [17:08] Yes. That is why I soon have to finish my s/i [17:09] In case HFB drops too low [17:09] I'm currently optimizing a paper - although I'm not sure whether I should use it alone or with a stone... probably both [17:09] What kind of paper? [17:09] antiimp [17:09] current optimax score is 135,25 [17:10] which is nice for a paper [17:10] huh? [17:10] is the benchmark biased in such a way? [17:10] hi brx [17:10] hi Nenad :) [17:11] sorry to intrude [17:11] hello jens [17:11] no problem :) [17:11] biased? [17:11] brx: A fluff* hello :) [17:12] brx: what did you mean by biased? [17:12] well, is it easier for, say, stones to get good scores with the optimax benchmark in question? [17:12] s/p's get the most [17:12] papers tie too much [17:12] now is that a quality of s/p, or is it a quality of the benchmark [17:12] benchmark [17:13] you never see equal amount of warrior types on the 94nop [17:13] ah so the benchmark already takes that into account? [17:13] * brx scratches his head [17:14] brx: No, I don't think so. All kinds of warriors are equally part of the score [17:14] Jens: which of my questions does your statement answer? [17:15] brx: optimax doesn't correspond to 94nop [17:15] 19:13:41) brx: ah so the benchmark already takes that into account? [17:15] ah. well I knew that. [17:15] it's not possible :) [17:15] the hill is a moving target [17:16] Jens: Fluffy Bunny is antiAfield, right? [17:16] * Jens just went into stupid mode. He cannot remember anything now ... [17:16] it scored much better than my CoreRW ... it scores really well [17:16] it's just that since that is not the case, does it have merit for a benchmark to favor one type of warrior over the other? [17:17] or, if that is not the case, does it show that s/p is actually the most optimal currently known warrior configuration for a naive stab at an unknown hill? [17:17] well, I like s/p's a lot [17:18] and I think that it is the ultimate combo [17:18] but they have their Nemesis as well [17:19] s/most optimal/best [17:20] by the way, jens, do you remember my statements regarding score surfaces equating to noise for a sufficiently large benchmark? [17:20] yes [17:22] ... I mean, what are you aiming at? [17:22] I have to redraw some of the implications. [17:24] while that statement is probably true, it has no real merit in the actual world of corewars and thus doesn't substract of the worth of scoresurfaces. [17:25] but you probably knew that already. so, given that you are as slow as you like to state you are, I am slower :) [17:27] You can say whatever you want about the score surfaces. I can take it ;-) But you are probably right, when given a huge amount of (different) opponents, the average score surface might either become just noise or very smooth [17:27] Fortunately we only have to deal with 19 opponents, which makes the thing more interesting. [17:28] oh? do we? [17:28] I never even knew you were limiting yourself like that. [17:28] on the 94nop-hill [17:28] I know [17:28] I'm about to test my paper (best randry so far) against those 19(20) warriors and I hope it won't be just noise :) [17:28] but it is moving [17:29] yes, it is moving and so we have more than 19 opponents over time [17:29] Nenad: Good luck :) [17:29] Nenad: from me too [17:30] Jens: doesn't that contradict your previous statement? [17:31] A little bit, I only thought about one fixed moment and not about the development over time [17:31] But even SoV only had about 2600 different opponents to "care about" [17:31] mh, I assumed that the development was important to you [17:31] It is [17:31] only? :) [17:32] you know, what I mean ... compared to the total number of possible warriors [17:32] Jens: btw, did you code something for this rf round? [17:32] yes, thank you for reminding me [17:33] I didn't have the time... I'll code something quickly tomorrow [17:33] what I am really interested in is how a score surface against a benchmark of 2600 different warriors would look like. given that those 2600 warriors are "competitive", how does the score surface look like and what number would actually suffice to approximate that resulting surface [17:33] the number of what? [17:34] it would be interesting to observe the score surface against different groups of warriors sorted by type [17:34] different warriors. what is the maximum difference between two score surfaces against two warriors of the same kind [17:34] and even to make a surface of standard deviation for each group! [17:35] The current surface benchmarks has only 12 warriors. I have calculate the average group surfaces, but they don't show anything interesting. [17:35] that would help reveal the "magic steps" [17:35] define "interesting" [17:35] It is more or less the same as the overall average score surface [17:35] Nenad: There are no magic steps. Only, if you have a limited amount of opponents [17:36] brx: interesting != looks more or less the same, when scaled by a value [17:37] i never thought the diagonals were interesting for example.. [17:38] I think, a lot of people knew, that pStep2 = k * pStep1 is bad, but only on a subconscious level [17:38] well they weren't interesting because they didn't depend on the paper's opponents [17:38] The score surfaces only show, what they already knew. But it is nonetheless good to have a "prove" for that [17:39] brx: Because most of the pStep2 = k*pstep1-steps lead to suicide very soon! [17:39] I have calculates YAP vs a "jmp $ 0, $ 0" to see that. [17:39] Jens: exactly [17:39] ... but not yet published [17:40] I think, I now have the means to compute "Score Surfaces - The Movie" [17:41] I take a 3-silk-paper and calculate a 800x200x200-cube from the score-suce and let is runs as an animation [17:42] I hope it to be interesting [17:43] one 200x200-patch takes about 30 minutes on my computer (100 rounds are enough for that). So I will need less than 30*800 minutes [17:43] less, because 30 mins are normally the max. for such a patch [17:43] hmm... it seems that there is no food on the 94nop for antiBfield paper now... I'll have to wait for some more stones and s/i's... [17:45] Nenad: Try to reoptimize it against papers. Maybe it works better then [17:46] ... I've also sent SmallWaterBeast (which entered 94nop earlier) and it scores the same but the new paper (which is not Moore-style) is more balanced [17:47] so, I'm satisfied. I think it'll be ok in a s/p [17:47] quit [17:48] Jens: you've said that you were working on a s/i? something unordinary or the usual stuff? [17:49] Altough I haven't tried a lot of different warriors (I still haven't written a decent p/s), I'm looking now for unusual combinations [17:50] so IFB (Innocent Fluffy Bunny) has some "unsual stuff" in it [17:50] ... (if I every get it to work properly) [17:51] I've got two s/i's ready for submission, but I won't send them until KryneLamiya falls off (hopefulle never) ... they hurt it too much [17:51] One good opponent won't kill KL [17:51] Join: sf joined #corewars [17:51] Hi sf! [17:51] Hello [17:51] hi sf [17:52] Nenad: But I know, what you are thinking. I cannot send new papers to the hill, because they might harm HFB [17:52] Jens: but Arrow is already hurting it... if it weren't for Arrow.. [17:52] yes. the hill is too crowded with papers [17:52] KL ist koth at the moment. What else can you wish for apart from living longer than SoV [17:53] :) [17:53] My scanner didn't last long [17:54] which one? [17:54] sf: Try to improve it :) [17:54] I have written more and "optimized" more than 20 different versions of Yatima before it was good enough to enter the beginner hill [17:54] *written more -> written [17:55] it took me lots of time till I learned how to make good scanners [17:55] hehe ... and I still don't know that [17:55] I'm the highest new entry on Koenigstuhl recursive nano! [17:55] I can only write acceptable qscans [17:55] and there are several really good s/p's on 94nop now... [17:55] sf: :-) [17:55] :) [17:56] Jens: are you from Germany? [17:56] I wonder, what would happen, if JM sent type-1 to Koenigstuhl [17:56] sf: yes [17:57] Jens: do you have a guess? [17:57] 1st place [17:57] Your German page doesn't get updated as often as the English [17:57] Join: Neo joined #corewars [17:57] Hi Neo [17:57] hi [17:57] hi Neo [17:57] Jens: I think so too. [17:57] hi all [17:58] It wouldn't be first place [17:58] sf: Yes, my homepage is updated in waves ;-) [17:58] sf? [17:58] Neo: hi [17:58] Hi Neo! [17:58] sf = S.Fernandes = FatalC = http://corewar.atspace.com/ [17:59] brx = http://www.ociw.edu/~birk/COREWAR/NANO/hill_rec.html = 6th :/ [17:59] oh. I didn't know you have a web page [17:59] Most of the warriors on nano are good and stable core-clears and type-1 seems to have found a sweet spot against core-clears. I definitely think it will become koth on Koenigstuhl [18:00] * Neo congrats to brx [18:00] But we'll talk again once JM has sent it to Koenigstuhl [18:00] and where is his unbeatable warrior? [18:00] Neo: I was only teasing. :) [18:00] (the one for 94nop that he was talking about) [18:00] nice, now i have some nanos to optimize against }:) [18:01] Jens: actually... m [18:01] Nenad: Don't know, but you will probably notice it once it is ready :-) [18:01] probably :) [18:02] John is attempting to prove it is unbeatable. [18:02] brx: But you have the oldest active nano warrior! [18:02] Neo: it nearly dropped. [18:02] I have to go now... I'll be back tomorrow. [18:02] * Nenad waves [18:02] * sf waves [18:02] * Jens waves [18:02] * Neo waves [18:02] MSG: [18:03] * brx waves [18:03] hehe ... this time I was fast enough :) [18:03] darn [18:03] But along came a hero to the rescue! [18:03] I should think how it is possible for type-1 to get those scores off on-speed [18:03] Wow ... I really love quickscanners ... 13-30 points just from the qscan [18:04] resistance to decrements? could that be it? [18:04] * brx sighs [18:04] brx: Don't exaggerate, it is still 38th. [18:04] Neo: again, yes. it was 50th once. [18:04] look some challenge earlier [18:05] Perhaps a clear dodger [18:05] I have an idea now! [18:05] ack me too [18:05] :-O [18:05] OH YES [18:06] * Neo waves late.. [18:06] who left? [18:07] Nenad [18:07] 20:03 Neo waves [18:07] 20:03 <*> Quits ->Nenad [Nenad@82.208.192.96] [] [18:07] Nah, irc weird things, sorry [18:09] * Neo notes type-1 beats easily both on-speed,(clear) and my scanners [18:10] Yay .. my optimizer now works with fmars :) [18:14] Join: Fizmo joined #corewars [18:14] hi [18:14] :) [18:14] Hi Christian :-) [18:15] hi fiz [18:15] hi all :) [18:15] Hi Fizzer [18:16] any cw related news today? [18:16] Didn't see any. [18:16] no :( [18:17] has anybody tried an airbag oneshot with two scanloops so far? [18:17] Do you have one ready? [18:17] is that like twinshot? [18:17] still got the idea how to do that [18:18] heh, I got the idea after looking to twinshot [18:18] Fizmo: scanning in parallel? [18:19] yep, with two different loops somewhere in core [18:19] Email isn't working today. [18:19] if one found something start clear from both loops [18:20] but no idea so far how to adjust the clear if I just use one [18:21] I have! [18:21] :) [18:21] why don't you use the same pointer [18:22] grhhh, yes. That would be pretty good [18:22] the good point would be that you could two different scansteps [18:22] my idea too [18:22] :) [18:23] * Neo was faster [18:23] inc: spl #step,>step [18:23] loop: add.f {0,scan [18:23] scan: seq d1,d2 [18:23] jmz.a loop,loop [18:23] and.. [18:23] loop2: add.f }loop,scan2 [18:23] ... [18:23] more or less [18:24] I can't code that fast :( [18:24] heh, we could make a little challenge. The first one who get something like that onto the 94nop hill will earn Arrow's code as price ;-) [18:24] Yay! [18:25] Not me, my email is down [18:26] You think someone would be successful during this evening ;-) [18:26] Hey, that depends on the name :) [18:26] Neo: where is the airbag check in your scanloop? [18:26] jmz.a [18:27] and loop: add.f {0 then loop2: add.f }loop [18:27] Join: Roy joined #corewars [18:27] Hi all [18:27] hi Roy [18:27] Hi Roy [18:27] * Roy has a problem.. [18:27] hi Roy [18:27] :-( [18:28] I'm currently writing 2 pieces, maybe for the next CoreWarrior or for r.g.c or something [18:28] Nick Change: Neo changed nick to Oracle [18:28] Neo: Yes I see. That's nice [18:28] Yes, please for CoreWarrior [18:28] But one is called: "Improvemend on Reepicheeps paper" ... but it scores less then the original, but I can't explain why :) it should be a improvement [18:28] I've got all the answers [18:29] Change the benchmark to one which suits the results you want :-P [18:29] though, don't know were i put them.. [18:29] How can I publish something that is called "improvement" when it scores less :P [18:29] Oracle: What is the answer to life, the universe and everything? [18:30] Explain that it will be an improvement against an as yet undiscovered class of warriors which will be invading the hills in the near future [18:30] does it have any advantage against a certain strategy? [18:30] Roy: Everyone has a MAXCYCLE limit. [18:31] Uhm, I'll have to see, actualy Halycion on the hill could be better if I change it back to the original paper too :-( [18:31] Nick Change: Oracle changed nick to Neo [18:31] if yes, then you could name it "improvements against xxxx on Reepicheeps paper" [18:31] Oracle: Wrong, the answer is 42 :-P [18:31] :-P [18:32] Does the improvement make it combine with other strategies in a hybrid warrior in a better way? [18:33] Nope, it just scores worse :-P [18:34] I'll leave it this way, me writing about the huge nice improvement etc, and leaving me amazed with less scores... excersize for the reader to figure out why its worse :) [18:36] The other article is also very boring, most is copied from a old forgotten r.g.c post (but its nice to rediscover tricks) [18:36] Whatever, I'll just mail them to JM :) [18:40] * Neo sf I don't have that code here. It is just a clear, similar to on-speed. Wich kind of warrior are you looking for? [18:40] damn! [18:40] Got one [18:40] neo: Good snippet, now there is the problem to assign which of the two scan loops found something [18:41] Fizmo: wait, that loop is wrong [18:41] What loop? Snippet? [18:42] airbag oneshot with two independant scanloops [18:42] heh, we could make a little challenge. The first one who get something like that onto the 94nop hill will earn Arrow's code as price ;-) [18:43] Ooohhww, a challenge, I'm in! [18:43] Arrows code, would like that! [18:43] * Roy dives into the mars [18:44] hmm, if one of the loops is hit by a bomb it won't adjust the pointer for the clear [18:44] anymore [18:45] so I could let adjust both scanloops the pointer, and hopefully only the intact one will adjust it [18:46] but if one loop found something I don't want that the second loop adjust the pointer wrong, so there must be a second checksum [18:47] depends what kind of bomb [18:49] I didn't hope for a spl-bomb ;-) [18:50] would be unlikely for paper or stone/paper [18:50] Well, the biggest change with a paper is that you get partly overwritten [18:55] BTW, don't forget the next rf round ;-) [18:55] deadline is soon [18:55] * Neo has nothing.. [18:56] * Roy has plans...no code yet [18:58] hmm [19:01] Hmm, the snippet seems to work fine [19:04] Part: Jens left #corewars [19:06] Join: Mizcu joined #corewars [19:06] ugh, my brains needed some time off [19:06] Hi Mizcu [19:07] hi FatalC [19:07] 3.5 hours of maths, and 1.2 books out of 3 read so far [19:07] hi Mizcu :) [19:08] Want 2 hours of deep redcode to recover? [19:09] Think ill just test out this versioning of this paper-project, about 30 mins [19:09] MSG: Read error: Connection reset by peer [19:10] Join: sf joined #corewars [19:13] MSG: Read error: Connection reset by peer [19:13] hrmm.. [19:14] Join: sf joined #corewars [19:15] sf: sorry. Seems when i send a message to you, irc server pushes you out! [19:15] Try again [19:16] ok, good bye :p [19:16] MSG: Read error: Connection reset by peer [19:16] ! [19:16] Join: sf joined #corewars [19:16] Okay that was a coincidence [19:16] Try again [19:16] ... [19:17] Seems fine now [19:18] Maybe, but i don't trust in these irc-webs [19:18] Hmm, scores pretty much like.. between expectable and weird [19:22] MSG: Read error: Connection reset by peer [19:23] man, i don't known what's wrong with me this evening... [19:23] Give me your opinion about which is better; should paper-steps be constant, or changing? This is expecting that change doesnt need additional instruction(s). [19:24] Mizcu: You mean by decrements/increments? [19:24] y [19:24] decrements work well, but you'll need a long optimization i suppouse [19:25] Well, the other possibility would be to use dec/inc for attack, which brings one more constant to optimize [19:26] and a little more agressiveness [19:27] i've got an idea.. [19:27] But if you use djn, you get weakness against anti-djn bombs [19:28] bah, not so much, and 1,1 bombs are not very common [19:28] Unless you throw them yourself.. [19:28] ah, heh [19:29] dou you remember the bombing imp launcher in maelstrom? [19:29] dou equ do [19:29] No, but i know the idea [19:30] it could be adding instead of bombing.. and try to make it hit some of the spl's of the paper [19:30] add is still quite disrupting [19:33] Join: sf joined #corewars [19:33] It is difficult to make it work well, though [19:34] This is getting beyond my understanding [19:34] MSG: Read error: Connection reset by peer [19:34] Join: Core29 joined #corewars [19:35] Join: sf joined #corewars [19:35] 8-process paper throwing 30-instruction carpets without loop [19:36] yust try to hit the head of the silk using add instead of mov (dat), and you modify its pattern during the battle [19:36] ww [19:37] exit #corewars [19:37] i think i am on two chanels at the same time [19:39] changing the replication-distance has little effect (well, 10 ties -> losses) against SOV [19:41] gotta go now [19:41] * Fizmo waves [19:41] * Roy waves [19:41] MSG: [19:42] * sf waves [19:43] * Neo waves to Fizmo [19:43] thanks for the kick :) [19:43] :-) [19:44] * Neo blocks sf's pathetic attempt at a kick in the nuts and punches him in the face [19:44] :-( [19:44] this web-irc has nice commands :) [19:45] it is time for me to go [19:45] :-( [19:45] * sf waves [19:46] don't want to sleep in this cold place.. better at home [19:46] * Neo waves [19:46] Neo, on the hill! [19:47] * sf waves [19:47] hey good! [19:47] 38th [19:47] * Neo goes now [19:47] MSG: Quit: Physical Wreck IRC webscript: http://www.pswclan.com/cgi-bin/irc/irc.cgi [19:49] We have a type-1 Killer! [19:51] MSG: Quit: Quick! Kill your client! Bersirc 2.2 is here! [ http://www.bersirc.org/ - Open Source IRC ] [19:53] ugh, gotta ask Met if he appears tomorrow, have no idea what to do with this [20:11] Join: Jens joined #corewars [20:14] hmm ... Fluffy Paper IIa was supposed to be a test warrior for my next text. It shouldn't enter any hill ... damn, I hate that ;-) [20:15] I dont get it, Weirdpaper with Moore-paper gets 165 points against SoV, and 160 against Uninvited, Without MP 131&174 [20:21] Part: Core29 left #corewars [20:30] bye [20:30] Part: Jens left #corewars [20:30] MSG: [22:18] MSG: [23:03] MSG: Quit: humhum