00:00:02 -!- supercode has quit (Quit: Client closed).
00:19:58 -!- salpynx has joined.
00:22:03 <salpynx> Lymia: hi! I didn't realize you were on here still :) Thanks for the comments and explanation of how nyuroki works (I'll happily add it to the wiki as part of organising my notes)
00:25:45 <salpynx> it is impressive, i chose it to work with because the high-level description was so short compared to some of the other high performers -- i'm mainly fixating on the nyuroki-esoteric.jx version which seems extra elegant, even though it's not the best performer
00:27:17 <salpynx> I can see what it's doing, but it's a lot harder to understand _why_ it does so well, your explanation helps with that
00:32:24 <salpynx> i'm happy to see my unstable_atom dropped about 3 places with the addition only a few other warriors lower in the rankings. Instability from over-tuning was a 'design goal'
00:34:53 <salpynx> the optimised nyuroki i ran the other does does slightly worse with the new additions, but not too badly: points: 22.21 -> 21.79 , Markov score 59.82 -> 58.50
00:38:04 <salpynx> Lymia: seeing your evolver would be interesting. i feel like i've got a feel for optimising details, but i need to work on understanding higher level strategies. bfjoust is a bit hard to 'get into' without having participated in the early evolution.
00:39:44 <salpynx> i misread logs and thought there was a proposal for a bot to monitor the ranking ladder for new entries, then immediately evolve and autosubmit a re-tuned version of itself (nasty!)
00:40:22 <salpynx> ... there was an unrelated 'ladder' esolang idea interleaved in the chat
00:40:39 <korvo> Oh, machine search is allowed for BF Joust? Interesting.
00:41:59 <salpynx> ... well, is it? It seems controversial. It's hard to tell how some warriors were made. I thought Sookie was hypothesised (by ais523?) to be machine generated.
00:44:06 <salpynx> i'd happily consider my recent 'optimisations' to be in a similar vein ... except i wasn't smart enough to code any kind of GA, and just evolved them by hand to develop an intuition on how to improve warriors
00:44:54 <korvo> I don't know. I imagine it'd be a per-hill policy.
00:45:39 <korvo> Relevant: Putting "Functional Genetic Programming with Combinators" https://www.cs.hmc.edu/~oneill/papers/Combinators-ASPGP.pdf on my rereading list.
00:49:43 <ais523> I think Sookie is machine-generated, although I don't have any proof
00:51:25 <salpynx> hand optimising was quite fun, i felt like i was exploring valleys and peaks of multi-dimensional terrain, in a particular local minimum, with only 2 floats as feedback
00:52:16 <salpynx> I just played a bit of bee-game, and there was a similar exploration vibe, but I know less people have explored the bfjoust terrain
00:54:06 <salpynx> b_jonas: which reminds me, Baba is you is not on your game list. I discovered via this channel (and enjoyed it!)
00:56:58 <esolangs> [[Special:Log/newusers]] create * Cobaltbluestars * New user account
01:03:24 -!- ais523 has quit (Ping timeout: 260 seconds).
01:07:36 -!- Sgeo has joined.
01:10:19 -!- amby has quit (Quit: so long suckers! i rev up my motorcylce and create a huge cloud of smoke. when the cloud dissipates im lying completely dead on the pavement).
01:16:31 <korvo> int-e: Okay, everything's pushed now; bbgauge.info will update in a minute. I ended up making the code change in my BF optimizer and pushing that as well.
01:17:10 <korvo> Along the way, I noticed that my BF implementation takes about 2.5s to run the typical mandel.b, and I think maybe that's due to writing a single character at a time. Do I feel like taking on output buffering? Maybe dinner first.
01:25:43 -!- salpynx has quit (Quit: Leaving).
01:32:23 <Lymia> I did realize a small optimization I could put out for a nyuroki3 that'd be pretty quick to write and maybe would 'stabilize' the ladder a little bit.
01:33:01 <Lymia> Since the way the same code is generated creates a lot of times, I could make each version slightly different programatically in a way that throws off timing/etc based tuning.
01:43:19 <esolangs> [[User:B jonas]] https://esolangs.org/w/index.php?diff=143959&oldid=143890 * B jonas * (+14) /* Games that the esolangs community plays */
02:23:37 -!- salpynx has joined.
04:16:35 <Lymia> !ztest nyuroki3 https://dl.aura.moe/paste/lymia/nyuroki3_0rg9zqhv38wwh1n9zi3ybyy985sp2cxllbjpai2qyya8fv9klp6j.bf
04:16:35 <zemhill> Lymia: URL fetch problems: No such file or directory @ rb_sysopen - https://dl.aura.moe/paste/lymia/nyuroki3_0rg9zqhv38wwh1n9zi3ybyy985sp2cxllbjpai2qyya8fv9klp6j.bf
04:19:58 <Lymia> !zjoust test_prog <
04:19:58 <zemhill> Lymia.test_prog: points -46.00, score 0.00, rank 47/47
04:41:00 <Lymia> !zjoust nyuroki2 <
04:41:00 <zemhill> Lymia.nyuroki2: points -45.00, score 0.00, rank 47/47 (-45)
04:47:52 <Lymia> !zjoust pushiguess <
04:47:52 <zemhill> Lymia.pushiguess: points -45.00, score 0.00, rank 47/47
04:48:05 <zemhill> Lymia.test: points -45.00, score 0.00, rank 47/47
04:50:01 <salpynx> I haven't used the url feature of those commands, but the URL works for me, and your command looks fine too, idk why it's not happy
05:18:01 <salpynx> Lymia: fwiw, here are the changes I made to nyuroki, in one git diff https://github.com/hornc/JoustExt/commit/162f51d6a0d12688e241059d6f6a2c4ce5ecc2d3 when viewed like this the changes to the strategy look minimal, and my rush changes are just obscure..
05:19:59 <salpynx> the repeated 255s stood out, after reviewing existing code repetition counts i noticed 256 present in multiple of ais523s warriors, so i think i possibly black box reverse engineered those particular values
05:20:13 <Lymia> I'm working on a version with randomized constants, and generally retuned.
05:20:20 <Lymia> Should make timing-based attacks extremely difficult.
05:22:32 <salpynx> I think i need to go back to the drawing board and try to understand the higher level strategies better, i can't really explain the differences between the rush. I don't fully understand the originals, nor do i understand my changes. Was hoping that studying the differences would reveal something interesting
05:27:48 <salpynx> I was struck by the fact that sensing anything is limited _and_ expensive, and the notation supporting the hardcoded magic numbers. Tuning the numbers seemed the only (or at least most obvious) deterministic strategy. Perhaps i took it too far though
05:30:03 -!- Sgeo has quit (Read error: Connection reset by peer).
05:31:54 <salpynx> feel free to use anything good in those changes back. I liked the [-.-+] wobble, i did change some to [-] where it made no difference to the score originally (to reduce size), but stopped because [-.-+] seemed much stronger when it mattered
05:39:55 <Lymia> I used a 7-cycle wobble to decrease the chance of it being locked by something designed to lock 5-clears.
05:41:43 <zemhill> web.Lymia_nyuroki3: points 16.19, score 42.83, rank 1/47
05:42:22 <Lymia> !zjoust test_prog <
05:42:23 <zemhill> Lymia.test_prog: points -46.00, score 0.00, rank 47/47 (-43)
05:51:11 <salpynx> oh no, nyuroki3 lost first place just because of the `<`?
05:54:41 <salpynx> and nyuroki2 has gone -- how do you swap out one of your own warriors like that to avoid spamming minor variations (I am hesitant to submit anything atm)?
05:56:27 <salpynx> oh, ok i see what you did earlier, clever (but sad to see it go)
06:26:07 -!- tromp has joined.
06:26:45 -!- tromp has quit (Client Quit).
06:30:19 -!- tromp has joined.
06:42:10 <Lymia> !ztest blind_bat (>)*9([(+.)*256>{}])*-1
06:42:11 <zemhill> Lymia.blind_bat: points -30.57, score 3.82, rank 47/47
06:42:40 <Lymia> !ztest blind_bat (>)*9([(+.)*256>{}]>)*-1
06:42:41 <zemhill> Lymia.blind_bat: points -30.57, score 3.82, rank 47/47
07:14:42 -!- molson_ has quit (Ping timeout: 252 seconds).
07:38:03 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
07:39:23 -!- Lord_of_Life_ has joined.
07:39:53 -!- Lord_of_Life has quit (Ping timeout: 252 seconds).
07:40:05 -!- tromp has joined.
07:42:19 -!- Lord_of_Life_ has changed nick to Lord_of_Life.
07:57:08 <esolangs> [[BigFish]] https://esolangs.org/w/index.php?diff=143960&oldid=102702 * Ractangle * (+72) /* Examples */
07:59:07 <esolangs> [[User:B jonas]] https://esolangs.org/w/index.php?diff=143961&oldid=143959 * B jonas * (+16)
08:21:38 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
08:26:33 <zemhill> web.Lymia_nyuroki3: points 16.29, score 42.75, rank 2/47 (--)
08:30:28 <Lymia> ais523: ash is pretty fun. Consistently beats nyuroki even after retuning for hill.
08:31:30 <Lymia> ash, nyuroki and impatience definitely seem to be a bit of a "new generation" going on here.
08:32:18 -!- tromp has joined.
09:31:06 <fizzie> The URL thing is probably a Ruby change I didn't account for.
09:32:44 <fizzie> Apparently Ruby 3 removed the facility of the standard `open` method to open URLs, so it's trying to open it as a file. Should be easy enough to fix.
09:39:17 -!- zemhill has quit (Remote host closed the connection).
09:39:27 -!- __monty__ has joined.
09:39:33 -!- zemhill has joined.
09:40:22 <fizzie> !ztest just_a_test https://zem.fi/git/?p=hill;a=blob_plain;f=web.Lymia_nyuroki3.bfjoust;h=e67c7e291837ba2827649966557628aab7b319de;hb=af8810ea990f34634637ddc4677c00f401b44c90
09:40:22 <zemhill> fizzie.just_a_test: points 15.29, score 40.83, rank 2/47
09:40:38 <fizzie> Yeah, a four-character fix, adding `URI.` in front of `open`.
09:51:19 <esolangs> [[BF Joust champions]] https://esolangs.org/w/index.php?diff=143962&oldid=143321 * Lymia * (+1208) Update description for Lymia.nyuroki2.
09:53:19 <esolangs> [[BF Joust champions]] M https://esolangs.org/w/index.php?diff=143963&oldid=143962 * Lymia * (-4) Small wording changes to nyuroki2 description.
10:01:08 <Lymia> Going to try and start work on the framework for an evolver again. Going to have to dig up the old "every bfjoust program ever" repos, lol.
10:01:25 <Lymia> I think I put them into cold storage somewhere already, since I can't find them.
10:35:23 <esolangs> [[STRTRAN]] https://esolangs.org/w/index.php?diff=143964&oldid=143711 * Froginstarch * (+2)
10:48:45 <esolangs> [[Chinese]] https://esolangs.org/w/index.php?diff=143965&oldid=122897 * None1 * (+17)
10:54:12 -!- X-Scale has joined.
11:15:11 <Lymia> My old code didn't bitrot too much, at least. Running a hill of 11000 bfjoust programs again.
11:22:29 -!- X-Scale has quit (Ping timeout: 256 seconds).
11:26:07 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
11:27:23 <esolangs> [[User:ATProtogen/Sandbox]] M https://esolangs.org/w/index.php?diff=143966&oldid=143945 * ATProtogen * (-104)
11:34:00 -!- tromp has joined.
11:44:18 <esolangs> [[Translated !]] https://esolangs.org/w/index.php?diff=143967&oldid=143891 * None1 * (+46)
11:44:36 <esolangs> [[Translated !]] https://esolangs.org/w/index.php?diff=143968&oldid=143967 * None1 * (+2) /* See also */
11:47:29 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
12:15:03 -!- tromp has joined.
12:52:10 -!- wWwwW has joined.
13:23:41 <esolangs> [[Translated ORK/PSTF Again17]] N https://esolangs.org/w/index.php?oldid=143969 * PrySigneToFry * (+2179) Created page with "Translated ORK/Mihai Again13|<span style="font-family:Unifont;">Sorry, your Android system has been hacked, please deal with it as soon as p#~=/?+{$M??}?}WD_??}?}u?O4bD?k??#?????eJ_o+#}e?/2??{IS|? D'!0#0Nb\?V ..."
13:24:15 <esolangs> [[Translated ORK/Mihai Again13]] https://esolangs.org/w/index.php?diff=143970&oldid=143950 * PrySigneToFry * (+78)
13:24:56 <esolangs> [[Joke language list]] https://esolangs.org/w/index.php?diff=143971&oldid=143694 * PrySigneToFry * (+74)
13:59:39 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
14:00:13 -!- craigo has joined.
14:10:28 <esolangs> [[]] https://esolangs.org/w/index.php?diff=143972&oldid=143841 * PrySigneToFry * (+93)
14:11:00 -!- lynndotpy has quit (Quit: bye bye).
14:11:50 -!- lynndotpy has joined.
14:13:02 -!- tromp has joined.
14:16:18 -!- craigo has quit (Quit: Leaving).
14:32:48 <esolangs> [[]] M https://esolangs.org/w/index.php?diff=143973&oldid=143972 * PrySigneToFry * (+1080)
14:39:18 -!- chiselfuse has quit (Read error: Connection reset by peer).
14:39:39 -!- chiselfuse has joined.
14:40:15 -!- wWwwW has quit (Ping timeout: 256 seconds).
14:46:09 -!- FreeFull has quit.
14:50:00 -!- FreeFull has joined.
15:01:05 -!- salpynx has quit (Quit: Leaving).
15:26:36 -!- ais523 has joined.
15:38:14 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
15:39:28 -!- amby has joined.
16:00:34 -!- tromp has joined.
17:03:23 <esolangs> [[User talk:PkmnQ/Alternate Universe Underload]] N https://esolangs.org/w/index.php?oldid=143974 * Ais523 * (+432) I have an unreleased project that was going down similar lines is it worth making the terminology the same?
17:25:36 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
17:39:12 <ais523> <salpynx> the repeated 255s stood out, after reviewing existing code repetition counts i noticed 256 present in multiple of ais523s warriors, so i think i possibly black box reverse engineered those particular values ← when I use exactly 256 it is generally not an arbitrary choice, but rather an attempt to ensure that a lock is "stable" i.e. the same values are encountered each time around the opponent's loop
17:43:22 -!- tromp has joined.
18:30:34 -!- ais523 has quit (Quit: quit).
18:42:10 <esolangs> [[4ME]] https://esolangs.org/w/index.php?diff=143975&oldid=143947 * Ractangle * (+77) /* Commands */
18:55:47 <esolangs> [[4ME]] https://esolangs.org/w/index.php?diff=143976&oldid=143975 * Ractangle * (+22) /* Commands */
18:57:52 <esolangs> [[Language list]] https://esolangs.org/w/index.php?diff=143977&oldid=143928 * Ractangle * (+32) /* N */
19:01:41 -!- salpynx has joined.
19:52:41 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
19:53:42 -!- Everything has joined.
20:12:06 -!- tromp has joined.
20:14:26 -!- ais523 has joined.
20:17:41 <ais523> Lymia: nyuroki's antishudder is the sort of thing that ash struggles with – it wants to lock programs in their antishudder but isn't consistent enough in the timing to manage it against one that doesn't do two checks
20:18:13 <ais523> and if the program doesn't have one it wants to shudder them to death, and isn't likely to hit the timing that survives
20:18:31 <ais523> ("ash" is an acronym for "anti-shudder", incidentally – it's the part of the program it attacks)
20:23:37 <Lymia> That... makes sense. It explains well why it's one of those programs that blows out markov scoring without high points. :D
20:24:26 <ais523> I've been wondering whether it's possible to adapt ash to get wins in some of the scenarios that are currently draws – probably not given that there aren't many draws that correspond to a win at the other polarity
20:24:40 <ais523> only monolith and polexchange are draws on one polarity and wins on the other
20:26:31 <ais523> I actually think that the current state of BF Joust defence programs is in an equilibrium between shudders and shudder-triplocks; unlike most defensive styles it doesn't seem possible to make a program that reliably beats both, whereas most defensive mechanisms do have a reliable way to beat them
20:27:16 <ais523> preparation and growth have both started mainly losing to programs which can out-decoy them, which is a bit of a surprise given the lengths they go to to set up a lot of decoys before starting their rushes
20:29:23 <ais523> in general I feel like the "era of pokes" is over – pokes seem to be reliably beaten by tape length estimation methods, which is a bit surprising and not what I expected them to lose to
20:29:46 <ais523> perhaps this is a sign that the range of possible tape lengths needs to be larger
20:30:23 <ais523> I have been wondering about 8 to 32 so that there are 50 games in a match rather than 42 – with each game being win/lose/draw that would let you record a program's performance against another as an integer percentage
20:33:24 <Lymia> There's a trick I found in nyuroki that makes me unsure about increasing the tape length spread.
20:33:52 <Lymia> Namely, skipping a large range of cells (I've found that even 8 works on the current ladder) after the rule of 8.
20:34:34 <Lymia> Increasing the tape length spread will make those tactics way stronger.
20:34:44 <ais523> people have been trying that for ages and mostly couldn't make it work
20:35:11 <ais523> I think the way to make it work has been cracked, though – you check for clashing decoy setups before doing it
20:35:44 <ais523> there are definitely matchups where it hurts, e.g. it's the reason why nyuroki draws with omnipotence rather than beating it
20:36:47 <ais523> omnipotence can only probabilistically beat nyuroki in normal circumstances, and the probability is less than 50%, but nyuroki shoots off the tape on the short tape lengths
20:36:53 <ais523> which brings it back to even
20:36:56 <Lymia> Yeah, it's a tradeoff as always.
20:38:57 <ais523> many opponents can be discourage from trying that sort of thing by making a fake decoy in front of the opponent's front decoy, because most programs check in front of their decoys to look for clashing decoy setups and/or opponents' clears
20:39:11 <ais523> but you can't reliably do that against nyuroki because it only checks inside its own decoys
20:39:35 <ais523> and except on short tapes, it's hard to get inside quickly enough
20:39:50 <ais523> (unless you unconditionally skip the first decoy)
20:40:19 <Lymia> That's the exact thing the short probablistic lock targets.
20:41:37 <ais523> that only works if the opponent is clearing, though
20:41:51 <ais523> if you get inside, set a fake decoy, then go back to setting up your own decoys and then finally rush
20:41:58 <Lymia> Yeah, a poke has better chances there. Huh.
20:42:03 <Lymia> I did have an idea to try.
20:43:01 <ais523> stealth4 isn't a champion but it did teach me a lot about the current hill
20:49:04 <ais523> I think a program that sets a lot of decoys, then alternates between rushing, and repairing its rightmost uncleared decoy, would do quite well – but it'd be a lot of work to write
20:50:02 <ais523> and you'd need to do an offset-clear-like thing of your own decoys in order to distinguish them from opposing trails (but you wouldn't clear all the way, just enough ro repair)
20:51:00 <ais523> btw, I got inspired by Lua Joust to wonder if it would be possible to create an esolang that allowed complex-ish BF Joust programs to be expressed succinctly, but which wasn't powerful enough to lead to, e.g., reliable triplock escapes
20:51:34 <ais523> my current plan involves a state machine + 1 counter + one pointer to a tape cell (which can only be set to your current cell), control flow can check the distance you are from the pointer or your own flag, or the value of the counter against a constant
20:52:41 <ais523> one nice thing about this is that it has decidable halting, so you can detect whether the programs have got stuck in an infinite loop or whether it will resolve, and check to see if a program can enter an infinite loop with no wait/move/tape-change/tape-check commands
20:54:09 <ais523> meanwhile, another nice balance property of "tape length estimation" programs is that they lose more short lengths against low-decoy defence programs than they do against typical rush programs, which helps to keep defensive programs viable
20:54:54 <Lymia> (geez, a basic poke right now is bottom of hill. Difficult hill...)
20:55:07 <ais523> incidentally, is there a reason why web.Lymia_nyuroki3 has no "n" characters in it? there are plenty of comments that look like they should have one, but have spaces instead
20:56:13 <Lymia> It's fine on the file on my disk.
20:57:24 <ais523> I have noticed that if I use the web form, my newlines get replaced with spaces
20:57:30 <ais523> maybe there's something that is replacing both \n and regular n?
20:58:49 <ais523> while thinking about "basic hills for beginners to practice against", I divided BF Joust programs into six main categories (with the idea that each hill should have a fixed number from each category)
20:59:24 <ais523> two fast rush categories (one for "stable fast rushes" that don't shoot off the tape against defenders / flag-adjusters / no-decoy programs, and one for "reckless fast rushes" that do, at least sometimes)
20:59:41 <ais523> two defence categories (one for programs that defend on the flag, and one for programs that defend in front of the flag)
21:00:21 <ais523> and two slow rush categories (one for programs that try to out-decoy the opponent, with pokes, big offsets, and the like; and one for programs that instead try to win through a superior number of decoys, often by using small decoys and by skipping forward if the tape seems to be large)
21:01:01 <ais523> the last of those categories has been doing very well recently, nyuroki belongs to it obviously, but also most of my best rushers
21:04:29 <ais523> actually I feel like what the hill is really missing is a good fast rusher – they tend to fall off the hill faster than the other techniques and not to get near the top
21:04:57 <Lymia> I'm trying to build a more modern poker right now.
21:06:38 <ais523> so the most survivable pokers on the hill are quintopia's space pokes, ais523.slowpoke, and two defensive poke programs (ais523.omnipotence and Oj742.quicklock)
21:08:54 <esolangs> [[4ME]] https://esolangs.org/w/index.php?diff=143978&oldid=143976 * Ractangle * (+57)
21:09:11 <ais523> defensive pokes arguably make more sense than aggressive ones – they are more survivable against being counterpoked, and timer clears mean that locks often break nowadays so getting a head start on the clear increases the chance you win before the lock breaks
21:11:21 <ais523> one of the reasons my state-machine+counter+pointer plan has the pointer is so that programs can remember the location they poked to
21:19:03 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
21:20:06 <Lymia> !ztest lugh https://dl.aura.moe/paste/lymia/lugh_0yak4ycn8wq0809bm3331lzw36bk7q5qvpsma2461q1i3adz44jl.bf
21:20:09 <zemhill> Lymia.lugh: points -10.86, score 12.29, rank 45/47
21:20:22 <Lymia> Very difficult. :D
21:20:35 -!- X-Scale has joined.
21:20:55 <Lymia> !zjoust lugh https://dl.aura.moe/paste/lymia/lugh_0yak4ycn8wq0809bm3331lzw36bk7q5qvpsma2461q1i3adz44jl.bf
21:20:58 <zemhill> Lymia.lugh: points -10.86, score 12.42, rank 45/47
21:21:00 <Lymia> Going to put it up and see how it gets screwed, lol.
21:21:33 <Lymia> Beats impatience, huuuh.
21:22:06 <esolangs> [[Translated ORK/Mihai Again14]] N https://esolangs.org/w/index.php?oldid=143979 * MihaiEso * (+2058) Created page with "[[Translated ORK/PSTF Again17|<span style="font-family:Unifont;">Sorry, your brand new iPhone 16 Pro Max is brokenX|[;2ddAgb=>_q@9pJD#9hj{jO:)e-cI0B4ZXCAny,boErNsK:x^f*k>+!cS</span>]] 1. Drag out that Y[@$RLOI..."
21:22:41 <esolangs> [[Translated ORK/PSTF Again17]] https://esolangs.org/w/index.php?diff=143980&oldid=143969 * MihaiEso * (+45)
21:24:25 <zemhill> Lymia.lugh: points -6.93, score 15.40, rank 41/47 (+4)
21:25:46 <esolangs> [[Joke language list]] https://esolangs.org/w/index.php?diff=143981&oldid=143971 * MihaiEso * (+393) /* Horribly translated variants */
21:28:03 <esolangs> [[Joke language list]] https://esolangs.org/w/index.php?diff=143982&oldid=143981 * MihaiEso * (-130) /* Horribly translated variants */
21:37:07 <zemhill> Lymia.lugh: points -6.60, score 15.69, rank 41/47 (--)
21:38:18 -!- Everything has quit (Read error: Connection reset by peer).
21:39:20 <zemhill> Lymia.lugh: points -6.50, score 15.96, rank 41/47 (--)
21:42:30 -!- Everything has joined.
21:49:58 <ais523> Lymia: so there's a bug in impatience where if there's a blank cell between two decoys, the same timer is used across both decoys
21:50:05 <ais523> rather than resetting it
21:50:33 <ais523> but it's hard to fix while fitting into BF Joust's control flow limitations, unless you remove the check to see if a cell is zero before trying to clear it
21:50:55 -!- __monty__ has quit (Quit: leaving).
21:54:51 <zemhill> Lymia.lugh: points -5.86, score 16.36, rank 40/47 (+1)
21:55:24 <Lymia> That explains something I'm running into.
22:04:30 <fizzie> Ugh. I noticed that 'n' thing (because I used web.Lymia_nyuroki3 as the test program for the URL), but thought "well they're probably been using some tool where 'n' is a special debug instruction and therefore it's had all 'n' letters removed", not that that makes so much sense.
22:06:02 <fizzie> The reason why the web thing strips '\n' is that it talks to the Ruby program (that runs the bot and generally manages the hill) over a Unix domain socket using a line-oriented protocol, so the program must be on one line.
22:09:20 <fizzie> The CGI script does `code = code.tr("\n", ' ')` in Ruby, but I think that should just be a newline in the first string.
22:10:38 -!- Sgeo has joined.
22:13:51 <zemhill> Lymia.lugh: points -5.71, score 16.75, rank 38/47 (+2)
22:23:19 <fizzie> `` ruby -e 'print("nanana\nfoo".tr("\n", " "))'
22:23:31 <fizzie> Looks like that shouldn't do anything to 'n's.
22:23:37 <zemhill> web.ais523_impatience: points -46.00, score 0.00, rank 47/47 (-45)
22:23:50 <ais523> !bfjoust impatience http://nethack4.org/pastebin/impatience.bfjoust
22:23:59 <ais523> !zjoust impatience http://nethack4.org/pastebin/impatience.bfjoust
22:24:00 <zemhill> ais523.impatience: points 16.86, score 44.28, rank 2/47
22:24:49 -!- X-Scale has quit (Quit: Client closed).
22:25:50 <esolangs> [[DoEverything();]] https://esolangs.org/w/index.php?diff=143983&oldid=84808 * Timothytomato * (-15) im pretty sure turing machines cant cure cancer
22:26:42 <Lymia> Oh, that explains why I'm getting weird scores, you updated impatience.
22:30:04 <ais523> I fixed a bug that made the timers expire too quickly, and then I could make the timers shorter because I didn't need to compensate for the bug any more
22:33:16 <zemhill> Lymia.lugh: points -6.86, score 15.29, rank 42/47 (-4)
22:33:34 <Lymia> Negative score, but... don't have the version that got the 38 anymore.
22:33:40 <Lymia> (And I think it's impatience's fault anyways)
22:34:35 <Lymia> Lugh's... polarizing right now.
22:34:45 <Lymia> Lots of great immediate wins, and lots of bad loses.
22:37:15 <esolangs> [[BF Joust champions]] https://esolangs.org/w/index.php?diff=143984&oldid=143963 * Ais523 * (+106) /* 2024 */ updates to impatience (I fixed a bug in it, meaning I didn't need to compensate)
22:37:41 <ais523> right, I noticed the bug in impatience by watching it against lugh
22:37:50 <ais523> the bug was triggered via two decoys with an empty cell between them
22:38:04 <ais523> (which lugh does, or at least did at the time)
22:41:22 <Lymia> Yeah, it was the impatience matchup fucking it up.
22:41:31 <Lymia> Against all logic, more compact decoy setups were losing more often instead of winning more often.
22:41:35 <Lymia> Which doesn't make sense otherwise.
22:54:39 <zemhill> Lymia.lugh: points 0.21, score 21.01, rank 20/47 (+22)
22:54:52 <Lymia> That's an aggressive poker that'll stick on the hill for a while.
22:58:26 <Lymia> One last feature I intended to implement which should win it a few more matchups.
22:58:30 <Lymia> But don't know where it's going from there.
22:58:45 <Lymia> Decent mid-ladder bot with an outdated strategy should be good enough. :D
22:58:55 <esolangs> [[BIT]] https://esolangs.org/w/index.php?diff=143985&oldid=101514 * Fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff * (+10)
23:02:03 <zemhill> Lymia.lugh: points 0.93, score 22.21, rank 16/47 (+4)
23:09:17 -!- Everything has quit (Quit: leaving).
23:10:31 <zemhill> web.salpynx_kalb: points -3.74, score 17.10, rank 34/47
23:11:15 <ais523> hmm, that lugh update made leviathan raise 21 places
23:11:27 <ais523> and then it dropped 13 when kalb was posted
23:11:41 <Lymia> Malkov scoring is super-unstable.
23:12:19 <ais523> although, oddly, it's normally fairly stable with respects to posting a champion or very high-scoring program
23:12:23 <ais523> it's the mid-scoring programs that mess things up
23:13:10 <zemhill> Lymia.lugh: points -1.45, score 20.26, rank 20/47 (+11)
23:13:33 <Lymia> Should be a feature-complete Lugh.
23:17:57 <salpynx> !ztest experiment_optinyuroki https://raw.githubusercontent.com/hornc/JoustExt/897620ad597db89b3378aa7881023c354fe887af/examples/nyuroki-opti.bf
23:17:57 <zemhill> salpynx.experiment_optinyuroki: points 19.74, score 51.57, rank 1/47
23:20:15 <zemhill> Lymia.lugh: points 0.14, score 20.91, rank 20/47 (--)
23:20:49 <salpynx> the kalb bot is a collaborative drawing program which detects the playfied size and draws a scaled image, then i tried optimizing to score ok. I was trying to see how bots could collaborate
23:21:33 <Lymia> Lugh is the perfect midtier, lol.
23:23:57 <ais523> oh, a licensing note: "MIT license" is ambiguous, there is more than one version
23:25:48 <zemhill> Lymia.lugh: points 1.17, score 22.39, rank 14/47 (+6)
23:28:48 <esolangs> [[Special:Log/newusers]] create * Mwwv * New user account
23:29:07 <Lymia> Oh, that explains why polarity messes with it so much.
23:29:11 <Lymia> Both my tripwires are the same polarity.
23:30:08 <zzo38> ais523: I knew that there is more than one BSD license, I did not know about MIT but now we can know, good
23:31:01 <esolangs> [[Esolang:Introduce yourself]] https://esolangs.org/w/index.php?diff=143986&oldid=143774 * Mwwv * (+172) hiya
23:32:40 <zemhill> Lymia.lugh: points 1.45, score 22.52, rank 16/47 (-2)
23:32:57 <Lymia> It's a moderate loss in markov score, but... better overall points, and more stable against a common bot type.
23:34:47 <esolangs> [[User:Mwwv]] N https://esolangs.org/w/index.php?oldid=143987 * Mwwv * (+407) create
23:36:16 <zemhill> Lymia.lugh: points 2.07, score 22.81, rank 16/47 (--)
23:37:37 <Lymia> Going to leave it at that. Have stuff to attend to.
23:38:08 <Lymia> ais523: messy as hell, but. Rank #16 fast poke with minimal defenses: https://github.com/Lymia/JoustExt/blob/master/examples/lugh.jx
23:38:17 <Lymia> Probably a lot of little room for improvement I've missed.
23:39:54 <ais523> at some point, a minimally defended fast poke turns into a weird sort of fast rush program
23:42:05 * ais523 vaguely wonders whether fast rush has ever been good
23:42:43 <ais523> the poke/fast-rush hybrids can be / have been
23:43:01 <Lymia> It repairs its own flags only once, after the poke.
23:43:09 <Lymia> Doesn't even repair flag, but sets decoys.
23:43:18 <Lymia> And they're in the 10-20 range, so... definitely a kind of fast rush.
23:44:07 <ais523> flag repair is a strategy that wasn't used much in the past but seems to be a core part of modern BF Joust, both in and against fast rush programs
23:44:40 <Lymia> (Its main "gimmick" is a routine that bets that it's standing on the flag with the first big decoy it finds)
23:44:54 <Lymia> (And it executes a massive reverse offset clear)
23:45:02 <Lymia> (Well, big cell I should say*)
23:45:10 <ais523> that's standard in fast pokes and on short tapes, I think
23:45:42 <ais523> e.g. ash does that if it detects clashing decoy setups
23:46:56 <ais523> well, if it detects that the second tape element is disturbed, that's the only cell it checks for fast rushes and clashing decoys
23:47:13 <salpynx> i was struggling with the MIT license, indicating what is source and what is compiled output is tricky, and trying to highlight changes to both source and output
23:47:17 <ais523> the third tape element is checked for activity, much later, as an anti-passive-defender mechanism
23:47:50 <Lymia> salpynx: Go ahead and treat it as CC0 if you'd like. XD
23:49:01 <salpynx> thx :) not sure MIT had generated bf code in mind when they created their software license...
23:49:58 <salpynx> ... I've just seen how my drawing program can distinguish polarity flip and hopefully draw the same image together... i need a break tho
23:50:10 <Lymia> ais523: Next thing I want to try is another try at an evolver. I have a basic idea of how to approach it a second time with a few more years general programming experience, heh. :p
23:51:41 <esolangs> [[User talk:Fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff]] https://esolangs.org/w/index.php?diff=143988&oldid=142287 * Fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff * (+43)
23:56:59 <salpynx> just noticed my 1337 program is doing ok now, it had almost dropped. It literally does not attack, it sets a decoys in the adjacent cell and defends the flag