IRC logs for #openttd on OFTC at 2019-07-12
⏴ go to previous day
02:54:02 *** Eddi|zuHause2 has joined #openttd
03:12:02 *** Eddi|zuHause2 is now known as Eddi|zuHause
04:39:13 *** Wormnest has joined #openttd
05:57:29 *** Thedarkb1-T60 has joined #openttd
06:01:09 *** rocky11384497 has joined #openttd
08:16:29 *** ChanServ sets mode: +v tokai
08:20:17 *** Wormnest has joined #openttd
09:43:20 *** Laedek_ has joined #openttd
11:06:30 *** Wormnest has joined #openttd
11:30:54 *** Thedarkb-T60 has joined #openttd
11:42:03 <Dakkus> Hi! Trying to get anything to work that has daylength patch on it. Anything I manage to get to actually run has the [Download] button in the Check Online Content greyed out. Why's that?
11:43:03 <Eddi|zuHause> how are you compiling things?
11:43:27 <Dakkus> Been mostly downloading binaries.
11:43:44 <Dakkus> The latest one being reddit's patch, because traditionally that has been the most uncomplicated one.
11:43:58 <Eddi|zuHause> then that is not the issue :)
11:44:35 <Dakkus> It shows all downloadable things, but each one's size is 0 bytes.
11:44:53 <Eddi|zuHause> sounds like a communication error?
11:45:39 <Eddi|zuHause> did you actually select anything to download?
11:46:28 <Dakkus> Haven't been playing OTTD for a few years, and my memory seems to have failed me.
11:46:56 <Dakkus> Forgot that it's not enough to choose one thing that I want to download and press the Download button, but I should check the box(es) as well.
11:47:05 <Dakkus> Feeling a bit stoopid here now ...
11:47:22 <Eddi|zuHause> you're probably not alone :p
11:48:09 <Dakkus> Okay, so now the next problem is to figure out what xUSSR patch is called these days.
11:48:48 <Dakkus> Or whether it exists at all. I always much liked the long wagons with low capacity, and having to juggle with the DC and AC electrifications.
14:32:29 <Eddi|zuHause> why wouldn't xUSSR set exist anymore?
14:46:52 <peter1138> Eddi|zuHause, maybe if you thought it was a "patch" it could be obsolete.
14:47:58 <ntsbmvnk> Eddi|zuHause: blame Gorbachev
14:48:29 <Eddi|zuHause> ntsbmvnk: it wouldn't be "x" otherwise
14:49:34 <peter1138> I wonder how to pronounce ntsbmvnk
15:06:51 *** supermop_work has joined #openttd
15:09:51 <ntsbmvnk> peter1138: very carefully
15:27:42 *** frosch123 has joined #openttd
15:43:23 <Samu> 1007 savegames generated
15:44:11 <Samu> now I need to figure a way how to score
15:45:27 <Samu> i see that 1.9.2 is released, i made all test on 1.9.1
15:45:55 <nielsm> nothing that affects gameplay should have changed
15:46:56 <Samu> I was thinking to score this in multiple ways
15:47:26 <Samu> i still didn't log this info, would have to open up all 1007 savegames
15:47:35 <nielsm> - Change: Allow building road stops over self-owned one-way/blocked road (#7547)
15:47:36 <nielsm> - Fix #5685: Check for free wagons in ScriptVehicleList (#7617)
15:47:49 <nielsm> those could theoretically affect something
15:48:54 <Samu> 48 AIs, 1st place gets 48 points, 2nd gets 47, etc..
15:49:07 <Samu> if it didn't bankrupt, gets double the points
15:49:39 <Samu> 1st place no bankrupt 96 points, 2nd place bankrupt 47 points, 3rd place no bankrupt 92
15:55:34 <Samu> 12 points for not finishing still seems too much
15:58:31 <Samu> well there's only a few AIs that build one-way roads
15:59:18 <Samu> WmDOT if I recall, but never uses road vehicles
16:01:27 <Samu> maybe I need to score victorious after bankrupt in a different way
16:02:36 <Samu> group all non-bankrupters first, so they always get highest score, then bankrupters after, then ties after, then losses
16:03:45 <Samu> victorious after bankrupt still pose a problem for the one that loses
16:06:05 <Samu> if the loser is still not bankrupted, should it be scored different when the winner had bankupted already?
16:08:54 <Samu> loser score when winner didn't bankrupt, when loser didn't bankrupt: X points
16:09:03 <Samu> loser score when winner didn't bankrupt, when loser did bankrupt: Y points
16:09:12 <planetmaker> there's different metrics you can measure success in a game...
16:09:16 <Samu> loser score when winner did bankrupt, when loser didn't bankrupt: Z points
16:09:35 <Samu> loser score when winner did bankrupt, when loser did bankrupt: W points
16:09:58 <planetmaker> economic success is only one. And you could do that on a linear scale. Or relative to the overall assets of all companies combined
16:10:07 <planetmaker> or something else
16:12:11 *** sla_ro|master has joined #openttd
16:21:19 *** Wormnest has joined #openttd
17:09:44 <DorpsGek_II> [OpenTTD/OpenTTD] Ansbaradigeidfran opened issue #7648: Dead keys aren't interpreted correctly when typing accented letters (Linux) https://git.io/fjPh9
17:27:45 <Dakkus> planetmaker: Thanks :)
17:30:06 <LordAro> ^ issue could be solved by sdl2
17:34:01 <nielsm> I'd accept the sdl2 patch right away if it was made as a new driver and not a replacement for the old
17:34:16 <nielsm> (even if the two can't be compiled in at the same time)
17:37:32 <peter1138> Has you noted that in the PR?
17:38:03 <peter1138> It would certainly allow a more direct comparison.
17:39:31 <DorpsGek_II> [OpenTTD/OpenTTD] nielsmh commented on pull request #7086: Change #6173: Update SDL driver to use SDL 2.0 https://git.io/fjPjC
17:45:28 <DorpsGek_II> [OpenTTD/OpenTTD] nikolas commented on pull request #7086: Change #6173: Update SDL driver to use SDL 2.0 https://git.io/fjPjP
17:58:02 *** HerzogDeXtEr has joined #openttd
18:01:34 <Samu> I'm nearly done, figuring the score rules
18:03:37 <Dakkus> Can anybody say anything about the processor intensivity of different AIs? I'm running on a computer from 2010 and would like to play a large map, so finding ways to preserve CPU time is really valuable!
18:03:41 <Samu> should I group all non-bankrupters first, so they always get highest score, then bankrupters after, or just double score non-bankrupters
18:04:26 <Samu> meaning that bankrupters may at times score higher than nonbankrupters
18:05:32 <Eddi|zuHause> you can limit the number of opcodes AIs can run per game tick. this will dumb them down, but might save CPU time if you're limited
18:06:05 <Dakkus> Eddi|zuHause: And, I can probably do this towards the endgame, when AIs are rather irrelevant anyway? :D
18:06:24 <Dakkus> Is there a GUI thing for that?
18:06:33 <Eddi|zuHause> no, you'd have to use the command line for that
18:06:41 <Eddi|zuHause> the ingame console
18:07:10 <Eddi|zuHause> but in the ?-menu you can find a performance viewer, to check whether AIs even are a problem
18:09:19 <Samu> must think: losses get 0 to 3 points, ties get 6 points, worst victory may get a minimum of 12 points, best victory gets 35 points
18:09:45 <Samu> non bankrupting victories get double the score
18:10:50 <Samu> imagine best time is of a bankrupted company: gets 35 points
18:11:12 <Samu> the 24th time is of a bankruptless company: gets 12 points, and doubles to 24
18:14:13 <Samu> my other alternative way to score this would be: group bankruptless times first, bankrupted times second
18:14:15 <Dakkus> Eddi|zuHause: Aghkackekacke, I cannot find the command to alter opcode amount. Sorry, but could you be so nice and just write it here for me? :)
18:14:52 <Dakkus> Ah, console commands might be it?
18:15:19 <nielsm> I think you can alter it in settings
18:15:21 <Eddi|zuHause> ai_max_opcode_till_suspend or script_max_opcode_till_suspend
18:16:07 <Eddi|zuHause> use them with the "set" command, or "list_settings"
18:17:51 <Samu> if in 24 rounds there were 10 bankruptless times, and 14 bankrupted times, the best time of a bankrupted company would get 35-10=25 points, and the worst time of a bankruptless company would get 26 points.
18:19:33 <Samu> bankruptless score could still be doubled, but not necessary
18:19:53 <Dakkus> Eddi|zuHause: Okay, seems to default to 10 000.
18:20:21 <Eddi|zuHause> Dakkus: note that those are separate settings, for AIs and GameScripts, respectively
18:20:25 <Dakkus> Then one last questiönchen: Does it spare any CPU time if all opponents use the same AI or the same AI script?
18:21:12 <Dakkus> Good, that's kinda the answer I was hoping to hear, even though it's a design choice that hampers performance :))
18:21:19 <Eddi|zuHause> in general it's a bad idea to use multiple of the same AI, as they think too much alike and constantly get in each others way
18:21:22 <Dakkus> Means that I can have a bit of extra variation.
18:21:35 <Dakkus> Ok, interesting thing to hear!
18:21:58 <Dakkus> I just hope that some day in the future I can have different trainsets for different players. That would be so much fun!
18:22:38 <Dakkus> Obviously it's possible with human players already now: Use NARS, DBSet and xUSSR set at the same time and decide that one only uses trains from one.
18:23:49 <Dakkus> Plus, require each one to start in some specific area and require in a way or another that the network must be connected or anything new must be at least somewhat in the vicinity of what already exists.
18:23:56 <Eddi|zuHause> theoretically the game could support that, each enginge has a bitmask which players have access to it (this is used for exclusive preview), but nobody seriously considered adding an interface to actually influence this bitmask beyond the current "nobody" "exclusive preview" and "all" states
18:24:20 <Samu> (if in 24 rounds) should be read as (if in a round consisting of 24 matches)
18:25:39 *** gelignite has joined #openttd
18:26:03 <Dakkus> And sorry for all these questions. It's just that the way I play this game (starting in 1830's, with daylength set to 9), I'm going to be playing it for at least half a real time year, possibly several. And I would prefer not crippling my game for September 2020 by configuring it badly in July 2019 :)
18:26:33 <Eddi|zuHause> you could hope you get a new computer by september 2020 :p
18:27:06 <Samu> daylength patch is in openttd now?
18:27:19 <Eddi|zuHause> but in general, map size is a much bigger impact on performance than anything else
18:27:24 <Dakkus> Samu: Nope, I think it should be, but it isn't. And probably will never be.
18:27:41 <Dakkus> Eddi|zuHause: That's what I'm pondering. 2k*2k or 4k*4k.
18:27:55 <Dakkus> I've never actually used a whole quarter of a 4k*4k...
18:27:58 <Samu> i dont think ais can manage daylenght properly
18:28:19 <Dakkus> I mean, I've been playing with my setup often enough and am very content with it.
18:29:29 <Dakkus> Basically it's xUSSR+daylength=9+FIRS_or_ECS+increased construction costs
18:29:54 <Dakkus> I mean decreased... But anyway.
18:30:37 <Dakkus> One idea is to have a very large map with very few cities, effectively increasing the space I have between cities so that the junctions wouldn't fill 90% of space between two cities.
18:31:15 <Dakkus> I've been playing TTD and later OTTD since 1995 and this is very clearly how I prefer it :)
18:31:51 <Samu> back to my business, which scoring method is fair
18:32:41 <Dakkus> Appreciate you taking care of the scoring method, BTW! Thanks! I'll try to be off now -->
18:35:27 <Dakkus> Eddi|zuHause: Can you beat something that has had content developed for it for 20 years? I mean, look at Sawyer's attempts at making a new TTD. They've been fine, but they've goto zero chance of success because the community is here.
18:37:12 <Samu> pick a method: 1) bankruptless get double score, 2) bankruptless first, bankrupted second
18:39:36 <nielsm> Samu: you've got a problem in that you ran your experiment first and then try to get something out of the data you collected later, instead of forming a scoring method first and then running the experiment
18:40:44 <nielsm> since now you're probably lacking data that could have been useful, like year-by-year growth and amount of infrastructure built, number of unsuccessful routes deleted, and much more that could be interesting to look at
18:43:36 <Samu> that's something I wasn't actively looking at
18:43:54 <Samu> I was only timing first to £10M in company value
18:44:59 <Samu> starting year affects vehicle availability
18:45:12 <Samu> so i had to come up with this bankruptcy thing
18:46:25 <nielsm> how about working by a benchmark instead, say having a "par" time to 10M company value, and scoring the AIs based on how they performed compared to that
18:46:30 <Samu> a company can suck bad when starting in early years and be awesome when starting in the later years (after a bankrupt), which could ruin the time needed to get to £10M
18:47:25 <nielsm> so if you have a par of 20 years to 10M value, and an AI managed to reach it in 17 years you could score it 13, if it reached it in 25 it could score 5, reached in 10 years score 20
18:47:39 <nielsm> and the AIs beaten in a particular game just get zero
18:48:15 <nielsm> or it may not have to be a linear scoring from the comparison, it could be a ratio or a logarithm or something
18:49:32 <Samu> I see, well, I have all the times yet
18:49:45 <Samu> I logged them all, there are ridiculous times as 300 000 days
18:49:47 <nielsm> and score bankrupt AIs by how many years they survived in some manner
18:51:02 <Samu> then there were also the other extreme, with AIs reaching £10M in 900 days
18:51:02 <nielsm> and if you have logged the maximum company value any reached you could use that to scale the bankrupt ones so one reaching a higher max company value and staying alive for long scores much better than one not reaching a good company value but surviving about the same time
18:51:30 <nielsm> yeah that could call for a logarithmmic scale to the scoring
18:52:06 <nielsm> also what is even your goal here? what do you want the score to reflect?
18:52:27 <nielsm> by scoring the AIs you're making a value judgment of them, but what value are you even judging?
18:53:11 <Samu> each round, each AI gets a score, and there were 47 rounds, a round robin tournament, all ais going against all others in a 1 on 1 rush to £10M
18:53:30 <Samu> final score should reflect the best overall AI
18:54:22 <Samu> or at least, the best getting to £10M AI
18:54:34 <nielsm> "best overall" is an extremely vague definition
18:55:07 <Eddi|zuHause> sounds like any social "science" :p
18:55:13 <nielsm> you don't talk about "the best sportsperson in the world", you talk about the best runner at 100 m dash, the best man in hammer throw, etc
19:02:52 <Samu> i was sorting times from lowest to highest
19:03:16 <Samu> lowest time gets the highest score: I decided the highest to be 35
19:03:36 <Samu> lowest time from a round with 24 matches
19:04:06 <Samu> 35, 34, 33, 32... I was scoring like this
19:08:58 <Eddi|zuHause> obviously the best sportsperson is the one with the highest lifetime income
19:11:45 *** Progman has joined #openttd
19:15:25 <Samu> goes from 3093 days to 22043 days
19:15:42 <Samu> then there's 24 other AIs unmentioned, they all lost
19:15:54 <Samu> this is only for round 1
19:16:22 <Samu> different rounds, different days
19:20:05 <Samu> i didn't keep savegames of those
19:20:56 <Samu> ties are games with cause the human company to actually win due to buffer underflow or whatever
19:21:15 <Samu> happens after a very long time, 360000 or something days
19:22:54 <DorpsGek_II> [OpenTTD/OpenTTD] Rk8ley opened issue #7649: Save game can't open (liblzma returned error code)? https://git.io/fjXvL
19:31:08 <Samu> inclined to do it that way
19:34:56 <DorpsGek_II> [OpenTTD/OpenTTD] Eddi-z commented on issue #7649: Save game can't open (liblzma returned error code)? https://git.io/fjXvZ
19:36:56 <DorpsGek_II> [OpenTTD/OpenTTD] nielsmh commented on issue #7649: Save game can't open (liblzma returned error code)? https://git.io/fjXvW
19:38:43 <DorpsGek_II> [OpenTTD/OpenTTD] nielsmh closed issue #7649: Save game can't open (liblzma returned error code)? https://git.io/fjXvL
19:51:39 *** andythenorth has joined #openttd
19:53:03 <andythenorth> I don't see that error in master
19:53:17 <nielsm> production callback running for the wrong economy setting, probably
19:53:54 <nielsm> do the production callbacks leave anything useful behind to trace them in the industry permanent data?
19:58:04 <andythenorth> yes, they fill some registers
19:58:28 <andythenorth> I haven't checked out JGR and compiled it
20:00:03 * andythenorth mostly drawing Horse :)
20:01:26 <andythenorth> also distracted by macos performance ticket
20:01:33 <andythenorth> I'm getting 60fps on ffwd
20:01:44 <andythenorth> in a save that was struggling to do 16fps the other day
20:01:49 <andythenorth> for the same part of the map
20:03:13 <andythenorth> 75% CPU where it was 100%
20:05:05 <andythenorth> I reloaded an earlier version of the save, before I killed 2 AIs
20:05:16 <andythenorth> still getting expected performance
20:05:33 <andythenorth> wondering if macos has something contending for performance
20:05:42 <nielsm> so something external affects it
20:06:02 <andythenorth> my tests are not scientific, they're very subjective
20:06:12 <andythenorth> I haven't tested for more than 30s or so
20:06:38 <andythenorth> current mac laptops will thermally throttle, but that should
20:06:42 <nielsm> if something external affects it, it could indicate a reason perhaps only some users are reporting it
20:07:27 <Dakkus> Argh, I've been world-generating based on heightmaps for 45 minutes now and I never seem to get one with cities on suitable places >.<
20:07:49 <nielsm> then make a scenario based on heightmap and place towns manually?
20:08:07 <Dakkus> That doesn't combine too well with NewGRF, does it?
20:08:22 <Dakkus> Or, maybe start a game, save it, open with editor, save, load?
20:08:37 <nielsm> as long as you set your newgrfs before starting the editor you're def. safe
20:09:45 <Dakkus> Okay, that will make this 530% less annoying. I so wish the heightmap format had a way (say, a 100% black pixel?) to mark the location of a city so that we could have cities in correct places.
20:10:05 <nielsm> yeah that would be useful
20:10:07 <Dakkus> But, of course this workaround will do it as well :)
20:12:32 <nielsm> that should be monitorable
20:13:15 <nielsm> (I need to make a clean version of that picture some time)
20:13:59 <andythenorth> if I could find a clean trigger for this fps issue, that would be interesting
20:16:02 <Samu> will see how it goes when I complete more rounds, see if it still makes sense in my mind
20:22:18 <DorpsGek_II> [OpenTTD/OpenTTD] nielsmh commented on issue #7650: OpenTTD company HQ graphics bug in x-ray mode https://git.io/fjXfk
20:28:56 <DorpsGek_II> [OpenTTD/OpenTTD] Ben79487 commented on issue #7650: OpenTTD company HQ graphics bug in x-ray mode https://git.io/fjXf3
20:33:18 <DorpsGek_II> [OpenTTD/OpenTTD] nielsmh commented on issue #7650: OpenTTD company HQ graphics bug in x-ray mode https://git.io/fjXfW
20:35:55 <DorpsGek_II> [OpenTTD/OpenTTD] Ben79487 commented on issue #7650: OpenTTD company HQ graphics bug in x-ray mode https://git.io/fjXf4
20:43:57 <Samu> it is weird not seeing Admiral in the top, but he was against strong opponents
20:52:48 <DorpsGek_II> [OpenTTD/OpenTTD] Ben79487 commented on issue #7626: Building drive through road stop on town-owned one-way road crashes game. https://git.io/fjXfK
20:54:09 <DorpsGek_II> [OpenTTD/OpenTTD] Ben79487 commented on issue #7619: Game crash from out-of-bounds helicopter, infinite circling with superfast aircraft https://git.io/fjXfP
21:02:55 <DorpsGek_II> [OpenTTD/OpenTTD] Ben79487 opened issue #7651: Too many statues/fountains are built in town centers. https://git.io/fjXfH
21:11:37 <DorpsGek_II> [OpenTTD/OpenTTD] stormcone commented on pull request #7627: Fix #7626: Allow building of drive-through stops over one-way/blocked roads owned by towns (instead of crashing). https://git.io/fjXJe
21:48:50 <DorpsGek_II> [OpenTTD/OpenTTD] stormcone opened pull request #7652: Fix #7635: Game crash on exit scenario editor. https://git.io/fjXJZ
21:59:33 *** ChanServ sets mode: +v tokai
22:39:23 *** Progman has joined #openttd
22:42:23 *** arikover has joined #openttd
23:14:12 <Samu> round 8, some games start in 1985, some in 1950
23:19:18 <Samu> i wonder how many more problems will I find, if any
23:20:51 <Samu> RailwAI was also updated to v19
23:21:26 <Samu> will railwai work on 1.9.1
23:22:48 <Samu> gonna do it on 1.9.1 v18, otherwise I would have to do all other games too, on 1.9.2, which gonna take 3 more months, i dont feel like re-doing that
23:29:24 *** andythenorth has left #openttd
continue to next day ⏵
