01:33:40: -!- Doxin has quit (Remote host closed the connection).
07:28:27: -!- glogbot has joined #corewars.
10:02:49: -!- mialwe has joined #corewars.
11:47:40: -!- Doxin has joined #corewars.
11:47:41: -!- Doxin has quit (Changing host).
11:47:41: -!- Doxin has joined #corewars.
13:11:07: -!- inversed has joined #corewars.
13:53:50: <john_metcalf> Hi inversed :-)
13:57:25: <john_metcalf> Congratulations on 94x 1st place
14:09:50: <john_metcalf> Superstition on 94nop / tiny has a reverse clear... I think there are still a few improvements to be made.
14:11:16: <john_metcalf> I've been trying to write a new vamp for tiny, but the current hill seems a bit hostile to them.
14:18:35: <inversed> Hi John!
14:19:25: <inversed> Wow, you are really good at figuring out how my warriors work!
14:20:06: <inversed> Negative thinking does exactly the same
14:21:19: <john_metcalf> I didn't know. I suspected either reverse scan, reverse clear or both from the name...
14:22:08: <john_metcalf> But I tried making a '94 reverse clear because my '88 oneshot scores well on 94nop :-)
14:22:32: <inversed> Indeed, I was inspired by 88 oneshots
14:22:58: <inversed> Modern papers and pws are sooo vulnerable to it
14:23:14: <inversed> Because they were never optimized against it
14:27:14: <john_metcalf> I might see if I can come up with a decent scanner with fixed length attack next :-P
14:27:31: <john_metcalf> Or another '94 vamp/imp
14:28:46: <inversed> I was trying to improve my fixed length wipe scanner recently
14:29:11: <inversed> It is pretty good, but falls short of Recon2's performance
14:29:35: <inversed> I feel like I've created Recon 1
14:30:31: <inversed> You have to try to write one to understand just how good Recon 2 is
14:31:08: <john_metcalf> I have tried for a long time.
14:31:40: <john_metcalf> have you tried one on the tiny hill, or multiwarrior?
14:35:46: <inversed> Tiny scores are poor
14:36:34: <inversed> But another type with short scan phase can do better
14:37:00: <inversed> Like mod-8 scan, fixed length wipe -> DClear
14:37:44: <inversed> And I had the impression that on multiwarrior scanners get destroyed
14:37:51: <john_metcalf> Spl/jmp bombing scanner works on tiny. Also blur and mirage scans. (All with d-clear)
14:38:55: <inversed> Blur type with .25c carpet not so well, because the paper is so strong
14:40:32: <inversed> How do you think, should I add QScan to Amber on the 94x?
14:54:32: <inversed> Oh, and there are actually 3 different possible process orders for the mirrored imp spiral, not 2
14:56:11: <john_metcalf> It's worth a try. I'm not really sure of the advantage a qscan has on 94x. Maxlength is large enough to try some interesting things though...
14:56:50: <john_metcalf> e.g. Two parallel qscans
14:58:42: <impomatic> 1) 0, 4001, 2, 4003, .... / 4000, 1, 4002, ...
14:59:03: <impomatic> 2) 0, 4000, 1, 4001, 2, 4002, ...
15:00:24: <impomatic> 3) 0, 1, 4001, 4002, 2, 3, 4003, 4004, ... ?
15:00:39: <inversed> 3) 0, 4000, 4001, 1, 2, 4002, 4003, 3
15:00:53: <inversed> Yes, right
15:01:24: <inversed> Think of the whole spiral as two subspirals A and B
15:02:09: <inversed> Then the orders are 1. AAA ... BBB ..., 2. ABBAABBA..., 3. ABABABAB...
15:02:46: <impomatic> My '88 mirrored imp launcher is a bit clunky :-(
15:03:01: <inversed> I call them 1. Sequential 2. Interleaved-2, 3. Interleaved-1
15:03:22: <inversed> Yeah, since there is no > in 88
15:03:34: <inversed> But you can try the sequential order
15:03:59: <impomatic> I think most published mirrored imps are Interleaved-1
15:04:00: <inversed> It can be launched with unmodified impfinity launcher, you just have to place twio imps in advance
15:04:22: <impomatic> Mirrored imps were invented before normal imp spirals :-)
15:05:04: <inversed> And unfortunately have rarely been used since then
15:06:04: <impomatic> They'd already been around for 4 years before they got named!
15:06:19: <inversed> Interleaved-1 is in fact the weakest of 3 orders
15:07:26: <inversed> Other two orders have different tradeoffs
15:09:03: <inversed> I'm going to write a mini-newsletter, more like a series of notes, and the first issue will be focused on this subject
15:09:23: <impomatic> Oh, fantastic :-)
15:09:53: <impomatic> Have you chosen a name? I assume you'll be publishing to r.g.cw?
15:10:33: <inversed> Haven't settled on a name yet
15:10:43: <inversed> Maybe CoreNotes, but that sounds generic
15:11:14: <inversed> And I have problems posting to r.g.cw, Google groups just won't let me do it
15:11:50: <impomatic> Hmmm... I'd be happy to post for you. Google groups is a pain.
15:11:59: <inversed> Ok, I can publish on corewar.eu and you can repost to r.g.cw
15:12:16: <impomatic> Alternative, you could use the corewar-l gateway to r.g.cw
15:12:42: <inversed> What is corewar-l?
15:14:10: <impomatic> Email your post to corewar-l@koth.org and it will be posted to r.g.cw
15:14:36: <inversed> Oh, great
15:14:42: <impomatic> I'm working on a new FAQ, so I'll make sure that's included. Lots of people have trouble with Google groups.
15:15:10: <impomatic> You can also subscribe to corewar-l so r.g.cw posts get emailed to you.
15:15:59: <impomatic> There's normally a delay between sending the email and it appearing in r.g.cw (all posts have to be approved)
15:19:38: <impomatic> By the way, did the book arrive yet?
15:21:07: <inversed> No, you can rely on the Russian mail service to mess things up
15:21:41: <inversed> Usually they are very slooow
15:21:55: <impomatic> LOL. Normally anything I send to Russia takes 3-4 weeks :-)
15:22:55: <impomatic> Mail to Italy and South America is also really slow. Other countries I don't have a problem with.
15:24:34: <inversed> What do you send?
15:25:10: <inversed> I once ordered a bunch of CDs from St. Petersburg and it arrived 9 months later
15:26:37: <impomatic> Mostly light up balloons and paper farolito
15:27:23: <impomatic> I order a lot of books from the USA and they normally arrive within a week :-)
15:28:12: <inversed> A fine example of Russian mail service:
15:28:13: <inversed> http://www.youtube.com/watch?v=wOe3eloICT4
15:29:48: <impomatic> If it's not lost, it's smashed!
15:44:14: <impomatic> This looks a bit better http://corewar.co.uk/32.gif
15:44:29: <impomatic> I'll probably do the same for mirrored imps next.
15:46:09: <inversed> Nice, what software you used?
15:46:29: <impomatic> Just GIMP.
15:47:54: <impomatic> There's also one of pMARS http://corewar.co.uk/screenshots/pmars.gif
15:49:12: <impomatic> I want to do a couple of warriors being single stepped in pMARS, but first I need to write a CDB macro to improve the single step display.
15:49:36: <inversed> You should change the pMARS color scheme (at leats as a GIMP postprocessing)
15:49:57: <inversed> Red on Green is the worst color combination possible
15:50:25: <impomatic> How about blue and yellow?
15:50:40: <inversed> I'm slightly color blind an I can hardly tell the difference
15:50:52: <inversed> Blue / yellow is ok
15:51:13: <impomatic> Ah, okay. That's a good point.
15:52:02: <impomatic> I wonder if it's worth writing a pMARS for a color blind friendly scheme.
15:53:14: <impomatic> By the way, I like these :-) http://corewar.eu/forum/showthread.php?tid=135
15:53:37: <inversed> Or just use CoreWin, which has nice color scheme by default
15:54:12: <inversed> Yes, MDS maps are something that I wanted to do a long time ago
15:55:11: <inversed> I will do one for the 94nop after running a huge round-robin tournament
15:55:35: <impomatic> Is there any way to include more warriors? E.g. using a tiny font?
15:56:42: <inversed> Yes, I can use a smaller font, but the text will be less readable
15:56:55: <impomatic> Is it possible to use the Koenigstuhl score file? http://users.obs.carnegiescience.edu/birk/COREWAR/94/hill.scores
15:57:44: <inversed> ~80 warriors seemed like a good tradeoff
15:58:20: <inversed> Right, I'll use the Koenigstuhl file
15:59:11: <inversed> I am also working on my own infinite hill ranking system
15:59:36: <impomatic> Something different to flat scores?
16:00:03: <inversed> Actually, I came up with two different ranking methods
16:00:05: <impomatic> Did you see the diversity hill discussions in r.g.cw?
16:00:24: <inversed> One is similar to the Koenigstuhl recursive scores
16:00:45: <inversed> No, I'll have to look it up
16:01:48: <inversed> Flat scores are terrible, nearly useless in my opinion
16:06:14: <impomatic> I don't think the October tournament will have flat scores.
16:11:11: <inversed> Well, they can be used for small round-robin tournaments
16:11:39: <inversed> But they are not well suited for infinite hills
16:13:04: <inversed> Because there can be lots of poor, non-competitive warriors and they will have a huge impact on all scores
16:15:38: <impomatic> Lots of beginner warriors in tournaments too which affects the score.
16:16:03: <inversed> Exactly
16:16:26: <inversed> That's why I came up with an iteratively reweighted scheme
16:16:40: <impomatic> But flat scores work perfectly well on Koth / SAL :-)
16:17:22: <inversed> They do, because all the warriors on the hills are competitive
16:17:35: <inversed> Here's my scheme:
16:17:47: <inversed> 0. Initialize all weights to 1
16:17:59: <inversed> 1. Calculate weighted scores
16:18:21: <inversed> 2. Update weights: W = Max(Score - 100, 0)
16:18:35: <inversed> 3. goto 1 until convergence
16:22:55: <impomatic> You could run that over the Koenigstuhl score file and see what you end up with.
16:23:13: <impomatic> Warriors that do well on Koenigstuhl often don't do well on Koth.
16:23:45: <inversed> And sometimes it is the other way around
16:23:57: <inversed> I already run my method for the nano and tiny hills
16:24:24: <inversed> Will have to do some coding to read the Koenigstuhl format
16:28:27: <inversed> BTW, was it you who visited my soundcloud page yesterday?
16:36:29: <impomatic> Yes :-) I noticed one of your tracks has the same name as one of my warriors.
16:36:32: <impomatic> Creeping Horror
16:37:34: <inversed> Just a coincindence :)
16:55:28: * impomatic would like a hill where the coresize changes between battles.
17:04:38: <inversed> My second ranking method works like that
17:05:17: <inversed> Start with all warriors, calculate flat scores
17:05:47: <inversed> Push off lowest scoring warrior, repeat until no warriors left
17:24:20: <john_metcalf> There aren't many successful warriors containing sunset style papers
17:25:35: <inversed> No, I think there is a sufficient number of them
17:26:09: <inversed> Paper/Stones mostly
17:28:41: <inversed> There are at least 10 reasonably good PWS with Moore-style paper
17:29:22: <inversed> But different strategies remain unexplored
17:29:50: <inversed> Maybe adding an imp launcher
17:31:26: <inversed> Or combining it with a different type of paper (Venomous Philtre)
17:31:42: -!- catageek has joined #corewars.
17:32:03: <impomatic> I've tried adding several different types of imp launcher to the paper without much success. Also found an easy way to add an end-silk to make an extra copy.
17:36:10: <impomatic> Playing with this:
17:36:26: <impomatic> paper spl pstep1, {src
17:36:26: <impomatic> mov }src, }paper
17:36:26: <impomatic> src mov {paper+4, <1
17:36:26: <impomatic> djn.f @0, }pstep2
17:48:05: <inversed> Interesting, but you are using empty core as a pointer for the second silk, right?
17:48:57: <inversed> This does not looks like a reliable way
17:56:07: <inversed> In Neith, I just used a front silk for extra replication speed
17:57:39: <impomatic> Yes, empty core. I haven't figured out how to have the pointer in the warrior itself.
18:08:57: -!- catageek has quit (*.net *.split).
18:09:17: -!- impomatic has quit (*.net *.split).
18:09:40: -!- mialwe has quit (*.net *.split).
18:10:51: -!- jkw has quit (*.net *.split).
18:11:17: -!- Doxin has quit (*.net *.split).
18:11:30: -!- X-Scale has quit (*.net *.split).
18:12:32: -!- satoriuplink has quit (*.net *.split).
18:12:54: -!- inversed has quit (*.net *.split).
18:13:42: -!- ChanServ has quit (*.net *.split).
18:15:15: -!- glogbot has joined #corewars.
18:16:24: -!- 17SAAFGOG has joined #corewars.
18:16:24: -!- ggherdov has joined #corewars.
18:16:24: -!- Qu4Z has joined #corewars.
18:16:42: <inversed> Aside from sunset, you can also use Paperone-style paper that jumps back to its own front silk
18:17:27: <inversed> This strategy has fallen out of fashion long time ago, but it is viable
18:17:33: -!- 17SAAFGOG has quit (Write error: Broken pipe).
18:26:55: <impomatic> I tested all published warriors against a benchmark that closely matched the '94nop hill last month (although the hill balance has changed slightly since)
18:27:32: <inversed> And ...
18:28:04: <inversed> Any surprises?
18:28:07: <impomatic> I think there were 13 warriors which were the same as the hill and 7 which were close...
18:28:21: <impomatic> Yes, a few surprises :-) myBlur 2 did quite well.
18:29:47: <impomatic> Oliva was the best, with a significant lead. The pdqscan (which Kline resubmitted to confirm my claim that it would be koth)
18:31:27: <impomatic> Rust and Burning Metal were the highest of your published warriors.
18:32:38: <impomatic> There were a few 88 warriors which scored well, but I can't find the results at the moment.
18:33:00: <impomatic> I ought to update the benchmark and run again.
18:38:45: <inversed> I have a "timeline" idea
18:39:42: <inversed> Test all warriors against some benchmark, then sort by date they were created
18:40:27: <inversed> It would be interesting to plot the results to see how the strategies evolved
18:41:44: <inversed> But that would require collecting hill submission dates
18:42:36: -!- Doxin is now known as Doxin[buttstuff].
18:42:46: <impomatic> I have submission dates :-)
18:43:22: <impomatic> They're listed on the warrior pages for the Hall of Fame warriors http://corewar.co.uk/halloffame.htm
18:43:49: <impomatic> If not I have hill logs for at least the last 10 years http://corewar.co.uk/logs.htm
18:44:38: <inversed> Oh, great
18:45:10: <impomatic> The hall of fame will soon be extended to all warriors over the age of 400 (most of the warrior pages are there already).
18:45:20: <impomatic> Then maybe all warriors over 250 :-)
18:45:53: <inversed> Very nice new HoF page, haven't seen it before
18:47:49: <impomatic> I'll update the active warriors later. Monster_Human_Grunt has moved up to 11th place
18:55:25: <impomatic> Back in 30 minutes...
19:17:43: <impomatic> 22 to be more precise :-P
19:28:34: * impomatic has 80+ computers being delivered tomorrow!
19:29:00: <impomatic> I'm clearing a bit of extra space at the moment...
19:29:04: <inversed> For personal use? :)
19:30:23: <impomatic> They're retro computers. Ranging from the Camputers Lynx, Sinclair Spectrum, Commodore 64 up to a Cray.
19:32:14: <inversed> I thought that cryptocurrency miners have lots of hardware, but 80+ machines ...
19:32:21: <impomatic> I'd quite like a PDP to play with as well.
19:32:28: <impomatic> 80+ slow machines!
19:32:41: <inversed> Are you opening a personal museum?
19:33:14: <impomatic> I had that idea when I collected them (about 10 years ago)... But it's not going to happen.
19:33:19: <impomatic> We're in the wrong location.
19:33:40: <impomatic> They've been in storage for a while. Now I have space for them again :-)
19:34:43: <impomatic> Is it worth trying to mine cryptocurrency on normal hardware these days?
19:35:19: <inversed> That's like a programming geek's dream
19:36:07: <inversed> I wonder how people react if they ask how many computers you have
19:36:29: <impomatic> Depends on the person.
19:36:39: <impomatic> It makes my girlfriend angry.
19:37:06: <impomatic> It makes other people pester me to organise a retro gaming night!
19:38:48: <impomatic> The computer museum in Cambridge have offer to loan me other large machines to borrow / play with :-) They suggested the Symbolics Lisp Machine.
19:40:27: <inversed> 80+ ლ(ಠ益ಠლ)
19:41:25: <inversed> Room you store them in must look epic
19:43:55: <inversed> Speaking of cryptocurrencies, mining bitcoin on typical hardware is terribly unprofitable
19:44:41: <inversed> Mining litecoin on GPUs was profitable before the market crash few months ago
19:45:30: <impomatic> I was wondering about the dedicated hardware...
19:46:11: <inversed> Yes, the dedicated hardware is supposed to be good
19:46:51: <inversed> But the problem is, demand is high
19:47:09: <impomatic> I tried mining about 3 years ago, but it made my laptop too hot! I gave up after a day.
19:47:44: <inversed> If you place an order, there is a significant delay before it is shipped
19:48:18: <inversed> By that time, net hashrate and difficulty will go up and your profit won't be as good
19:48:25: <impomatic> I was watching this guy sell the cheap USB stick miners on eBay. He was selling for �15. The next week �30. The next week �60. And they still sold.
19:49:04: <inversed> I bought one AMD Radeon GPU for mining litecoin
19:50:13: <inversed> After a couple of months I've earned about its price
19:50:32: <inversed> Then prices crashed and I turned it off
19:51:15: <inversed> When the litecoin panic started, AMD GPUs were sooo hard to find
19:51:34: <inversed> Normally there is plenty of them lying on the shelves
19:52:58: <inversed> But one week after the sudden ltc/usd growth it was nigh impossible to find even a single one
19:57:29: <inversed> What scanline size do you use for pMARS fight visualization?
19:58:08: <impomatic> scanline?
19:58:37: <inversed> The core (1D) is translated into a picture (2D) using a scanline order
20:00:11: <inversed> I mean length of one line on display in core locations
20:00:37: <impomatic> I use pmars -v 033 which gives a width of 126.
20:00:51: <inversed> Can you specify other width?
20:01:42: <inversed> It has a major impact on the visualization clarity
20:02:32: <impomatic> For tiny I use pmars -v 093 which gives a width of 37 (very big blocks)
20:02:55: <inversed> Modulo properties are important
20:04:01: <impomatic> The width can't be set exactly, but the second digit of the -v number sets the cell size. I think it's also possible to get a width of about 315.
20:04:19: <inversed> According to my model, some good widths are: 120, 144, 64, 56, 36, 104, 160
20:04:22: <impomatic> On Linux it's different. I think there's a -geometry switch or something.
20:05:29: <inversed> With properly selected width, mod-x and mod-1 near mod-x patterns will look like lines
20:05:39: <impomatic> Ares width is 100. CoreWin is 128.
20:05:55: <inversed> And none of them are good ):
20:06:20: <impomatic> If the next version of CoreWin is released, it will have adjustable width.
20:06:41: <inversed> Will it ever be released?
20:07:39: <impomatic> Chip sent me an email earlier this year. There's some stuff he needs to do and it will most likely be released.
20:07:50: <impomatic> I don't want to pester him though...
20:08:07: <inversed> That's good news
20:11:04: <inversed> Nice chatting with you, but it's time to get some sleep
20:11:11: <impomatic> I'd like to see a MARS for Android. I don't want to learn Java just to write one though...
20:11:12: <inversed> Bye
20:11:16: <impomatic> good night :-)
20:15:37: -!- inversed has quit (Ping timeout: 246 seconds).
21:12:55: -!- catageek has quit (Quit: KVIrc 4.3.1 Aria http://www.kvirc.net/).