03:35:31 Experimental (bcrawl) branch on underhound.eu updated to: 0.23-a0-5261-gd9800d219b 03:36:55 <02Monkooky> @Odds some experience actually trying the bias now- not enough, will need to play more, but enough for first impressions. I don't think there's a good case for a bias from 1-10. You will gladly hare off down a corridor as far as it goes, but also double back at very odd times. (With bias 10 I believe as soon as you don't have an unexplored hex within vision range you'll double back and try another path) 03:40:02 <02Monkooky> Bias 30 feels much better, but is in some respects unsatisfying. 03:41:14 <02Monkooky> On this level layout with bias 30, autoexplore visited the immediate surroundings of the stairs, then the entire northwest segment before exploring any other corridors 03:41:19 <02Monkooky> https://cdn.discordapp.com/attachments/747522859361894521/1496460867120795778/image.png?ex=69e9f749&is=69e8a5c9&hm=4ec876bd102ab800f67e30f2a0a8c0ff2dba36fc8ce70a3249a28ffa8f570d86& 03:44:06 <02Monkooky> This allowed some wights to show up as I was returning from the north by northwest corridor, as 03:49:07 <02Monkooky> It was also a strong enough bias that I had some pointless and silly doublebacks later on when the level was almost fully explored 04:02:10 <11Odds> Thanks! 04:02:39 <11Odds> Yeah definitely at 30 or even 20 you get some silly doublebacks 04:03:31 <11Odds> Interesting on the inconsistency of 10 which makes a lot of sense. I actually felt it did make quite a big difference but yeah inconsistently so as it will hare off to infinity 04:04:09 <11Odds> The wights in that screenshot came from roughly where you are now? 04:27:51 <02Monkooky> came from west by northwest corridor 04:29:04 <11Odds> Oh right I was mixing up east and west 10:58:55 New branch created: pull/5221 (1 commit) 13https://github.com/crawl/crawl/pull/5221 10:58:56 03CrawlOdds02 07https://github.com/crawl/crawl/pull/5221 * 0.35-a0-292-gadc6a812b9: Apply coglin's rev penalty to melee attacks. 10(10 minutes ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/adc6a812b911 13:30:53 <02Darby> so, the thing about revmeleebug is that as the dust has settled across the past 2 years, it's come to feel as if an uncompensated nerf to melee Co (and only melee, as rev has always affected the overwhelmingly most popular, though partly meme, CoHu^Oka) is uncalled for, which raises the question of what to do about it, and there's many proposed options: - officially remove rev's effect on melee, changing the text but not the current 13:30:54 gameplay, retaining it mainly for gizmos (each of which it affects normally), potentially making rev apply to more gizmos - compensate with melee apt/stat buffs (but of course, melee Co, especially axes or long blades, already have roughly human apts and stats besides a mere -1 to armour/dodging, and straightforwardly stronger melee damage than a comparable 2h, on the basis that jewelry is that big of a deal) - change the numbers on rev itself (it's 13:30:54 telling that almost nobody noticed this for 2 years and all learned it entirely from studying the code) - simply nerf/fix melee Co on the assumption that ranged popularity among winning players is not relevant to intended balance and that their melee in particular can take a nerf 13:32:08 <11Odds> Ah interesting! 13:32:53 <11Odds> How big is rev when working as intended? 13:36:09 <11Odds> I'm not sure it's actually super telling that people didn't notice this from gameplay, damage is super high variance and you can have pretty significant effects on it without them noticing (see also: double damage bug) 13:40:15 <02Darby> it true 13:40:45 <02Darby> and of course, the numbers on rev are clear, how consequential those numbers versus the average length of a fight, etc, is harder to say 13:41:05 <11Odds> Scales from 2/3 to 1 damage over 40 aut, from a quick read? 13:41:38 <02Darby> yeah. I don't know how I feel myself. 13:42:15 <11Odds> Seems... like it probably matters a bit. And not loads. More early, when fights are shorter. 13:42:27 <11Odds> Maybe it matters loads on D:1 13:42:50 <02Darby> melee coglins do currently do extremely well at the earliest XLs, iirc 13:45:08 <11Odds> It would feel a bit weird if rev affected range and not melee, after all this time believing it did both 13:46:41 <02Darby> yeah, it's definitely been nontrivial to settle on a plan of action so far 13:47:20 <11Odds> (At any rate I'll close that PR as clearly there's loads of ongoing thought on this I wasn't aware of) 13:53:45 Placebo of seeing "Rev+" in your status is huge apparently 15:20:24 <02Monkooky> There were definitely people who assumed rev was a damage buff 15:27:53 <02Monkooky> explore bias 80, I autoexplore to this point 15:27:54 <02Monkooky> https://cdn.discordapp.com/attachments/747522859361894521/1496638703500529754/image.png?ex=69ea9ce9&is=69e94b69&hm=89c95e95e70316392af12a7db2ee070e5ec6bd0f872775a6d865b1dfa398d697& 15:29:14 <12ge0ff> maybe there should be some max distance from the closest stairs, when the autoexplore just forgets about stairs altogether? 15:29:40 <02Monkooky> this is something weirder, because this screenshot contains the entire unexplored area 15:30:57 <02Monkooky> oh wait it was probably trying to look through some of the glass to the left of that point 15:31:09 <12ge0ff> since fast monsters and slow species exist, it doesn't always make sense to worry about the distance if it gets too big 15:32:06 <02Monkooky> Well, part of the value of always caring is that's necessary for the algorithm to maintain a clearish path to stairs 15:33:58 <02Monkooky> See here- when you have this kind of fork in the path, since the bias is relative it will currently function similarly to if the stairs were at the fork point 15:36:08 <02Monkooky> that said, in practice I think a 'explore within this radius of stairs first' might just function better- the current bias doesn't do that very well without getting an extreme level of silly doubling back later on 15:39:35 <12ge0ff> yeah, it feels like there should be some cap for silly doubling back/oscillating 15:43:22 Unstable branch on underhound.eu updated to: 0.35-a0-293-ga3442ab723 (34) 15:48:07 <02Monkooky> Well, moreover it's hard to actually get a particular radius explored reliably- the exponential nature of the explore distance means a corridor starting at the wrong point can lead you very far from safety with relatively little explored 17:45:12 -!- Discord|1 is now known as Discord| 18:31:14 <08nicolae> FUN FACT: if all the lines of the MAP in your subvault have a blank space in the front, the level builder will assume the subvault is wider than it is 18:31:55 ...... ah 18:32:19 I assume that was what was causing the rotation problem? 18:35:08 <08nicolae> no, i abandoned the vault that had the rotation problem, i don't think subvaulting handles non-symmetric subvaults real well and i opted to not fuck with it more. this was a separate vault i made with better-shaped subvaults. 18:35:40 <08nicolae> me, before: subvaults are a fun and easy way to delegate complicated decisions to other vaults 🙂 me, now: every single subvault is a tile or two away from tearing open the fabric of reality 18:38:12 maybe if you make your subvaults just reference their own subsubvaults to handle the complexity then that'll make things easier 18:38:41 reality: it's all a hack and you're lucky any of it works at all 18:54:59 <08nicolae> ah, this time i made sure that the subvaults would place properly BEFORE making a ton of them 😌 18:55:50 03WizardIke02 07* 0.35-a0-294-g2ccebcafee: Fix tile variation duplicate detection 10(11 minutes ago, 3 files, 36+ 57-) 13https://github.com/crawl/crawl/commit/2ccebcafee38 19:13:41 <02Darby> reality reality: it's both. always has been. 19:14:03 <02Darby> well, okay, not "easy" 20:26:28 <06pleasingfungus> re rev: ;_; 20:27:44 <09hellmonk> you have to come back and fix it 20:27:49 <09hellmonk> it's the only way 20:48:17 <02Darby> I still think it's a bit funny that guns akimbo apparently speaks to people such that the majority of players will suffer having exactly two 1h weapons in their class and, unknowingly, a mechanic that nerfs only them and nobody else, to get it 20:56:38 03WizardIke02 07* 0.35-a0-295-g2878072334: Fix RTILE not respecting tile variant weights 10(14 minutes ago, 1 file, 5+ 3-) 13https://github.com/crawl/crawl/commit/287807233436 20:57:13 <06pleasingfungus> the genius of my design..?! 22:36:03 Unstable branch on crawl.develz.org updated to: 0.35-a0-295-g2878072334 (34) 22:59:56 Windows builds of master branch on crawl.develz.org updated to: 0.35-a0-295-g2878072334 23:13:42 Unstable branch on cbro.berotato.org updated to: 0.35-a0-295-g2878072334 (34) 23:56:05 Monster database of master branch on crawl.develz.org updated to: 0.35-a0-295-g2878072334