--- Log opened Sat Oct 27 00:00:20 2012
20121027 00:31:28-!- zookeeper [~lmsnie@wesnoth/developer/zookeeper] has quit [Ping timeout: 260 seconds]
20121027 02:00:39-!- Cookiee [~quassel@unaffiliated/cookiee] has quit [Quit: No Ping reply in 180 seconds.]
20121027 02:01:04-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 02:01:04-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 02:01:05-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 02:16:23-!- Cookiee [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 02:18:16-!- Cookie [~quassel@unaffiliated/cookiee] has quit [Ping timeout: 252 seconds]
20121027 02:23:46-!- Cookiee [~quassel@unaffiliated/cookiee] has quit [Ping timeout: 245 seconds]
20121027 02:29:43-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 02:29:44-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 02:29:44-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 02:36:42-!- Cookiee [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 02:36:42-!- Cookiee [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 02:36:42-!- Cookiee [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 02:38:26-!- Blueblaze [~Blueblaze@adsl-99-148-246-168.dsl.hstntx.sbcglobal.net] has joined #wesnoth-umc-dev
20121027 02:40:09-!- Cookie [~quassel@unaffiliated/cookiee] has quit [Ping timeout: 252 seconds]
20121027 02:42:17-!- Cookiee [~quassel@unaffiliated/cookiee] has quit [Ping timeout: 255 seconds]
20121027 02:42:32-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 02:42:32-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 02:42:32-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 02:53:22-!- shadowm_laptop [ignacio@wesnoth/developer/shadowmaster] has joined #wesnoth-umc-dev
20121027 02:56:58-!- Cookiee [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 02:56:58-!- Cookiee [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 02:56:58-!- Cookiee [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 03:00:22-!- Cookie [~quassel@unaffiliated/cookiee] has quit [Ping timeout: 268 seconds]
20121027 03:01:56-!- Cookiee [~quassel@unaffiliated/cookiee] has quit [Quit: No Ping reply in 180 seconds.]
20121027 03:21:54-!- ancestral [~ancestral@17.24.116.175] has joined #wesnoth-umc-dev
20121027 03:23:01-!- Blueblaze [~Blueblaze@adsl-99-148-246-168.dsl.hstntx.sbcglobal.net] has quit [Quit: Blueblaze]
20121027 03:23:02-!- loonybot [~loonybot@wesnoth/bot/loonybot] has quit [Remote host closed the connection]
20121027 03:53:55-!- vultraz_laptop is now known as vultraz
20121027 03:57:03-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 03:57:03-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 03:57:03-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 04:05:30-!- Cookie [~quassel@unaffiliated/cookiee] has quit [Ping timeout: 264 seconds]
20121027 04:18:54-!- ancestral [~ancestral@17.24.116.175] has quit [Quit: ancestral]
20121027 04:43:29-!- shadowm_laptop [ignacio@wesnoth/developer/shadowmaster] has quit [Quit: shadowm_laptop]
20121027 05:51:46-!- mattsc [~mattsc@d154-20-32-241.bchsia.telus.net] has quit [Quit: bye]
20121027 07:27:33-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 07:27:34-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 07:27:34-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 07:41:23-!- SeattleDad [~SeattleDa@c-67-171-51-0.hsd1.wa.comcast.net] has joined #wesnoth-umc-dev
20121027 08:30:27-!- zookeeper [~lmsnie@wesnoth/developer/zookeeper] has joined #wesnoth-umc-dev
20121027 08:35:02-!- Cookie [~quassel@unaffiliated/cookiee] has quit [Remote host closed the connection]
20121027 08:56:08-!- faabumc [~vcr@117.136.31.20] has joined #wesnoth-umc-dev
20121027 09:50:40< faabumc> Hi SeattleDad, I am checking the ML recruiter 0.3
20121027 09:51:39< faabumc> While I don't understand much in neural net and such, it is interesting
20121027 09:52:57< faabumc> I have a question: I see in "Winning percentages for Recommended ML Recruiter" that the winning stats for all factions are significantly worst against Northerners
20121027 09:53:04< faabumc> (except Loyalist vs undead)
20121027 09:53:42< faabumc> do you have any idea why?
20121027 10:25:28-!- faabumc [~vcr@117.136.31.20] has quit [Read error: Connection reset by peer]
20121027 10:42:34-!- zookeeper [~lmsnie@wesnoth/developer/zookeeper] has quit [Ping timeout: 240 seconds]
20121027 11:08:54-!- noy [~Noy@wesnoth/developer/noy] has quit [Quit: noy]
20121027 13:18:05-!- skyfaller [~skyfaller@wikipedia/Skyfaller] has quit [Remote host closed the connection]
20121027 14:27:34-!- faabumc [~vcr@117.136.31.18] has joined #wesnoth-umc-dev
20121027 15:16:42-!- loonybot [~loonybot@wesnoth/bot/loonybot] has joined #wesnoth-umc-dev
20121027 15:16:45-!- mode/#wesnoth-umc-dev [+v loonybot] by ChanServ
20121027 15:24:50-!- vultraz [~chatzilla@unaffiliated/vultraz] has quit [Changing host]
20121027 15:24:50-!- vultraz [~chatzilla@wesnoth/developer/vultraz] has joined #wesnoth-umc-dev
20121027 15:28:00-!- faabumc [~vcr@117.136.31.18] has quit [Changing host]
20121027 15:28:00-!- faabumc [~vcr@wesnoth/developer/faabumc] has joined #wesnoth-umc-dev
20121027 15:29:08-!- mattsc [~mattsc@d154-20-32-241.bchsia.telus.net] has joined #wesnoth-umc-dev
20121027 15:29:15-!- faabumc [~vcr@wesnoth/developer/faabumc] has quit [Quit: Ex-Chat]
20121027 15:29:39-!- faabumc [~vcr@117.136.31.18] has joined #wesnoth-umc-dev
20121027 15:30:03-!- faabumc [~vcr@117.136.31.18] has quit [Changing host]
20121027 15:30:03-!- faabumc [~vcr@wesnoth/developer/faabumc] has joined #wesnoth-umc-dev
20121027 15:31:53-!- faabumc [~vcr@wesnoth/developer/faabumc] has quit [Client Quit]
20121027 15:32:23-!- faabumc [~vcr@wesnoth/developer/faabumc] has joined #wesnoth-umc-dev
20121027 15:32:24-!- faabumc [~vcr@wesnoth/developer/faabumc] has quit [Client Quit]
20121027 15:33:35-!- faabumc [~vcr@wesnoth/developer/faabumc] has joined #wesnoth-umc-dev
20121027 15:37:04< SeattleDad> Hi faabumc
20121027 15:37:29< SeattleDad> I'm guess you signed out and I was logged in last night while I was asleep
20121027 15:37:53< SeattleDad> Whoops, I see you just joined, are you there?
20121027 15:38:05< faabumc> Hi SeattleDad
20121027 15:38:45< SeattleDad> To answer your question about the Northerners, what I'm seeing is that the Northerners are the strongest faction across the board
20121027 15:39:17< SeattleDad> On RCA AI vs. RCA AI games, Northerners win more often than anyone else by a pretty wide margin
20121027 15:39:47< mattsc> faabumc, SeattleDad: Crab once told me that the RCA AI was originally designed for Northerners by Dave
20121027 15:39:55< SeattleDad> So the ML Recruiter does better vs. an RCA AI Northerners than RCA AI does, but still loses more often
20121027 15:40:04< mattsc> So it might not be the strongest faction, it might just be that the RCA AI is best at playing them.
20121027 15:40:13< SeattleDad> That could make a lot of sense
20121027 15:40:25< SeattleDad> In campaign games you are often playing against the Northerners
20121027 15:40:46< faabumc> SeattleDad, possibly. Or maybe we could say AI is generally better tooled to play Northerners style
20121027 15:41:07< SeattleDad> I'd agree with that
20121027 15:41:36< faabumc> that makes sense. They are pretty straightforward.
20121027 15:41:45< SeattleDad> The issue with the RCA AI strategy outside of recruiting is that it just charges right at the enemy
20121027 15:41:58< SeattleDad> Not enough emphasis on:
20121027 15:42:05< SeattleDad> A:  capturing villages
20121027 15:42:14< SeattleDad> B:  Retreating injured units
20121027 15:42:39< SeattleDad> although it is quite good at finding the optimal attack in a given situation
20121027 15:43:19< SeattleDad> The logs that you were looking at are a little out of date, by the way, they are for ML Recruiter 0.2
20121027 15:43:43< SeattleDad> I'm in the process of updating the Wiki and will be posting ML Recruiter 0.3 stats today
20121027 15:44:36< faabumc> I talk to you later
20121027 15:44:41< mattsc> Ah, actually, I found the statement in the logs.  It's even the other way around: "since Dave designed them to make it easier for AI to play them"
20121027 15:45:16< SeattleDad> See ya, faabumc
20121027 15:45:36< SeattleDad> by the way, I'm looking for people to try out MLR and give me feedback, by the way, so if you could download MLR 0.3 and try it out, that would be great
20121027 15:45:52< SeattleDad> Hi mattsc
20121027 15:46:00< mattsc> SeattleDad: it's still on the plan that I do that ...
20121027 15:46:35< SeattleDad> Alarantalara is trying it out and seems to be having some success
20121027 15:46:59< SeattleDad> One thing I'm experimenting with right now is integrating MLR with ron and fred
20121027 15:47:12< mattsc> Yeah, he's really taken over the recruiting work for the AI-Demos AIs, so he's the right person 
20121027 15:47:29< mattsc> That would be great
20121027 15:48:16< SeattleDad> I actually have included in MLR 0.3 ai/ais/ml_ai_rush.cfg which is my interpretation of how an integrated system would work
20121027 15:48:52< SeattleDad> If you and Alarantalara could look at that and let me know if that makes sense, that would be great
20121027 15:49:33< SeattleDad> I should caution that ml_ai_rush.cfg plays very badly right now.  I haven't given it enough training data yet
20121027 15:50:02< mattsc> Ok, I'll do that - but I need to fix the AI's that I broke yesterday first ...
20121027 15:50:05< SeattleDad> So it's mostly a proof of concept.  The recruiter is currently much worse than Alarantalara's
20121027 15:50:35< mattsc> Fred and Ron are working again though (<-- skyfaller, Alarantalara)
20121027 15:51:02< mattsc> SeattleDad: ok, noted.  I'll look at the concept only then.
20121027 15:51:29< SeattleDad> I wanted to give you a heads up of a couple of issues I found with ron
20121027 15:52:11< mattsc> please do
20121027 15:52:25< SeattleDad> First of all, I'm getting a lot of crashes on Aethermaw
20121027 15:52:59< SeattleDad> It says 
20121027 15:53:03< SeattleDad> ...h_1.11/data/add-ons/AI-demos/lua/generic-rush_engine.lua:72: bad argument #2 to 'find_path' (invalid location)
20121027 15:53:25< SeattleDad> Aethermaw is kind of funky, if I understand what's going on the map changes in mid-game
20121027 15:53:45< SeattleDad> Maybe this is blowing you up
20121027 15:53:54-!- Jarkko_P [~jake@212-226-75-60-nat.elisa-mobile.fi] has joined #wesnoth-umc-dev
20121027 15:54:16< SeattleDad> Second issue is on speed
20121027 15:54:52< SeattleDad> It's running much slower than the RCA AI
20121027 15:55:31< mattsc> I just checked l.72.  That's in a function that Alarantalara wrote, so we should wait for him with that (or if he doesn't show up, I can have a look later)
20121027 15:55:59< mattsc> As for speed: are you talking about Ron or Fred?  
20121027 15:56:07< SeattleDad> I'm talking about ron
20121027 15:56:11< mattsc> Because Fred is particularly slow at the moment, but I am working on that.
20121027 15:56:13< mattsc> ok.
20121027 15:56:51< SeattleDad> With ron, there seems to be an issue across the board, but it's especially pronounced on the cynsaun battlefield
20121027 15:57:03< mattsc> Part of this is that both Fred and Ron are doing additional things, so they have to be slower.
20121027 15:57:31< SeattleDad> That's understandable
20121027 15:57:44< mattsc> I've been concentrating on Fred recently and haven't checked out Ron in a while.
20121027 15:58:09< SeattleDad> I'm not sure if this is up to the point where it's human noticeable, but it definitely impacts games running in nogui mode
20121027 15:58:23< mattsc> Let me do a few trial runs later today and see if I can find some obviously slow CA.
20121027 15:58:37< SeattleDad> cynsaun battlefield is the worst, if you wanted to look at that
20121027 15:58:44< mattsc> Ok, will do.
20121027 15:58:51< SeattleDad> It's often taking over 5 minutes to run a game
20121027 15:58:53< mattsc> It's possible though that I won't be able to do much about it.
20121027 15:59:22< mattsc> Ok, 5 min should not be necessary for ron, unless it's a 50-turn game or so.
20121027 15:59:40< SeattleDad> Okay.  I think I might put a timer in ai_test2.py to kill a game if it's taking over, say, 5 minutes
20121027 15:59:56< SeattleDad> I'm already killing it if it goes over 50 turns
20121027 16:00:17< SeattleDad> On fred, I was thinking of experimenting with that too
20121027 16:00:25< mattsc> I see what I can do.  I haven't worked on Ron in a while, so it's possible there's still some old code in there that can be optimized.
20121027 16:01:01< mattsc> Don't try with Fred yet.  Alarantalara recently found a single move that took 800 seconds.  :P
20121027 16:01:14< SeattleDad> Okay, I'll hold off
20121027 16:01:21< mattsc> Although I have that one down to 15s yet, but it's still work in progress.
20121027 16:01:30< mattsc> s/yet/by now
20121027 16:01:44< SeattleDad> Okay
20121027 16:02:10< mattsc> That's the reason why I am breaking the AIs at the moment...
20121027 16:02:33< SeattleDad> Just so you know where I'm going with this, I'm aiming at producing a scientific paper out of this . . .
20121027 16:02:44< mattsc> Anyway, I'll have a look at Ron on CB, probably tonight, to see if I can find something obvious.
20121027 16:03:08< mattsc> SD: That would be neat!
20121027 16:03:19< SeattleDad> and what would be cool would be to show that ML Recruiter can outperform the fred or ron recruiter when dropped into fred and ron
20121027 16:04:06< mattsc> right
20121027 16:04:26< mattsc> I better get on with fixing things then...
20121027 16:04:49< SeattleDad> Thanks.  I'm aiming for a conference called "Computational Intelligence in Games":  http://eldar.mathstat.uoguelph.ca/dashlock/CIG2013/ 
20121027 16:05:02< SeattleDad> Which is going to be held in Niagara Falls, Canada
20121027 16:05:36< SeattleDad> Due date for the paper is March 1, 2013
20121027 16:05:47< mattsc> Paper submission: 1 March : does that mean full papers or abstracts?
20121027 16:05:56< SeattleDad> full papers
20121027 16:06:04< mattsc> ok, so that sets the timeline.
20121027 16:07:01< SeattleDad> I think the results so far may be enough to get accepted, but showing that it works on a hybrid system with ron and/or fred would make it stronger
20121027 16:07:47< SeattleDad> What I would hope to see would be that it recruits a different blend (faster units?  more poisoners?) with those AI's than with the RCA AI
20121027 16:08:18< mattsc> That makes sense.
20121027 16:09:29< mattsc> Of course, at least Fred is still changing constantly, but Ron's been pretty stable for a long time.
20121027 16:10:16< SeattleDad> What would be interesting would be to show that as Fred evolves, MLR evolves
20121027 16:11:39< SeattleDad> For instance, if you put in logic that used healers more, it would start recruiting more healers
20121027 16:11:46< mattsc> Yeah.  Well, as I said, I am doing some basic refacturing at the moment with the goal of speeding up some things.
20121027 16:11:57-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has joined #wesnoth-umc-dev
20121027 16:12:14< mattsc> I introduced "real" counter attack calculations recently, and that takes a lot of computation.
20121027 16:12:15< SeattleDad> That would be great
20121027 16:12:36< mattsc> I'll let you know when that is done.
20121027 16:13:05< SeattleDad> Speed is pretty important to what I'm doing since the faster you are, the more simulations I can run and the more data I can gather for the neural net
20121027 16:13:13< mattsc> I'll also have a look at Ron's speed and the file you mentioned up there sometime this weekend.  I'll try today, but no promises.
20121027 16:13:42< SeattleDad> Okay, this all sounds good
20121027 16:13:54< SeattleDad> One question:
20121027 16:14:16< mattsc> Don't expect things to be as fast as the RCA AI though.  Esp. Fred does a lot of extra stuff, and that just takes time.
20121027 16:14:56< SeattleDad> That's okay, I can work with slower speed.  I just won't have as many datapoints
20121027 16:15:15< SeattleDad> In Ron and Fred, I think I might be able to help in ways other than recruiting
20121027 16:15:52< SeattleDad> I think I could give you a probability of a side winning in a given situation, would this be helpful?
20121027 16:16:31< mattsc> I think so ...   (hesitation because I have not thought about that before)
20121027 16:16:53< mattsc> I guess it could help with deciding on an offesnive vs. defensive strategy
20121027 16:17:11< SeattleDad> If you're losing, would you take bigger risks?
20121027 16:17:40< mattsc> That's one option - but you can do it the other way around too.
20121027 16:18:03< mattsc> If you're winning anyway, you might take a bigger risk because the loss of a couple units won't hurt you as much.
20121027 16:18:05< SeattleDad> An attack that has only a 25% chance of succeeding is a good idea if your odds of winning are only 10% at that time . . .
20121027 16:18:29< mattsc> Yep.  It could go either way.
20121027 16:18:45< SeattleDad> Well, it it's an opportunity to kill the enemy leader, then it's worth the risk if you're losing, but not if you're winning
20121027 16:18:48< mattsc> Could you use this for only a sub-set of the units as well?
20121027 16:19:20< SeattleDad> Like this subset of units is losing?
20121027 16:19:52< SeattleDad> I'm not sure I follow
20121027 16:20:04< mattsc> E.g: are the AI units in the eastern part of the map stronger than the enemy units in that same part of the map.
20121027 16:20:47< mattsc> Should I rush with them, or pull them back and combine with all the other units?  That kind of question.
20121027 16:21:34< SeattleDad> Here's something that may help:
20121027 16:22:29< SeattleDad> the other way I was thinking of helping is that I could compute gold yield on a going-forward basis given a units current hitpoints
20121027 16:22:43< SeattleDad> and given their current experience poitns
20121027 16:23:02< SeattleDad> Gold yield is the metric I'm using in MLR 0.3
20121027 16:23:34< mattsc> yeah, that might be useful.
20121027 16:23:39< SeattleDad> It's described at http://wiki.wesnoth.org/Machine_Learning_Recruiter#Gold_Yield 
20121027 16:24:08< mattsc> I've seen that before but not really worked my way through it.
20121027 16:24:36-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121027 16:24:37-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121027 16:24:37-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121027 16:25:12< SeattleDad> The core idea is with the "Basic Damage Metric" shown there
20121027 16:25:38< SeattleDad> If I destroy 25% of the hitpoints of a unit that costs 20 gold, then that's worth 5 gold
20121027 16:26:11< SeattleDad> So I credit the unit that destroyed the 25% of hitpoints for 5 gold
20121027 16:26:33< SeattleDad> Poisoning, healing, and slowing are handled analagously
20121027 16:27:19< SeattleDad> What the neural net is predicting when it's making its choice is how much gold yield a unit is likely to do given the current game state
20121027 16:27:55-!- zookeeper [~lmsnie@wesnoth/developer/zookeeper] has joined #wesnoth-umc-dev
20121027 16:34:08< mattsc> Sounds good.  Maybe that could be included in some of the ratings.
20121027 16:34:41< mattsc> I have to go now though, I'll get back to you.
20121027 16:35:58< SeattleDad> Okay, see ya
20121027 16:40:34< faabumc> SeattleDad, what of the other thing I find quite surprising is the low rating for the mage for loyalist given the high effect of damage in the "gold yield"
20121027 16:41:10< faabumc> would it be that the damage is cumulative during the life of the unit and mage are dying fast?
20121027 16:41:56< faabumc> *one of the other thing
20121027 16:50:35< SeattleDad> faabumc:  Sorry, I stepped away for a minute
20121027 16:50:59< SeattleDad> I've been focusing on the Loyalists vs. Undead a lot
20121027 16:51:48< SeattleDad> It recruits more Mages when fighting the Undead
20121027 16:52:07< SeattleDad> Currently it's heavily oriented towards spearmen
20121027 16:52:32< faabumc> which is good :)
20121027 16:52:48< SeattleDad> It's heavily oriented towards spearmen in general, that is, but swings more strongly to Mage and Heavy Infantryman when playing the Undead
20121027 16:53:09< SeattleDad> Glad you think spearman is a good option :)
20121027 16:53:10< faabumc> how heavily it relies on archer and merman is more confusing
20121027 16:53:29< SeattleDad> Just a sec, let me check current stats
20121027 16:53:49< faabumc> 25.7%	3814	Bowman
20121027 16:54:43< faabumc> it does mean that I have 1/4 chance that the AI is recruiting an archer, right? 
20121027 16:57:07< SeattleDad> Current breakdown is 43% Spearman, 15.6% Bowman
20121027 16:57:20< SeattleDad> And 8.1% mage
20121027 16:58:10< SeattleDad> The thing is, in a given situation, it can vary dramatically
20121027 16:58:26< faabumc> ok, you don't get these numbers from  http://wiki.wesnoth.org/Machine_Learning_Recruiter, do you?
20121027 16:59:01< SeattleDad> I'll update those number.  Give me a sec.
20121027 16:59:22< SeattleDad> Those are numbers for MLR 0.2, I'm on 0.3
20121027 16:59:23-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has quit [Quit: i go nstuf kthxbai]
20121027 17:09:41< SeattleDad> Sorry, faabumc, this is taking me longer than I thought
20121027 17:10:08< SeattleDad> I'm going to have to put it in tabular format to make it look right
20121027 17:10:20< SeattleDad> I'll have an update in an hour or so
20121027 17:10:43< faabumc> it is ok. It is not like I need these numbers right now.
20121027 17:11:08< faabumc> I was just checking I looked at the right table
20121027 17:11:53< SeattleDad> Just to give you a quick answer, Loyalists are recruiting 8.1% Mage overall, but against the Undead they are recruiting 28.8% Mage
20121027 17:12:18< faabumc> so if a mage is hitting 40 HP in 2 turns and a bowman is hitting 40 HP in 4 turns, that gives the same "gold yield"?
20121027 17:12:52< SeattleDad> Similarly, Heavy Infantryman jumps from 9.6% overall to 28.6
20121027 17:13:02< SeattleDad> *to 28.6% vs. the Undead
20121027 17:13:41< SeattleDad> Yes, assuming that the Bowman lives for four turns
20121027 17:13:50< faabumc> I see, that seems like a sane trend :)
20121027 17:13:53< SeattleDad> The idea is that it's the total gold yield through the life of the unit
20121027 17:14:24< SeattleDad> this tends to reward units like Troll Whelp and Wose that regenerate
20121027 17:14:46< SeattleDad> the ML Recruiter likes these units a lot for Northerners and Rebels
20121027 17:15:47< SeattleDad> Northerners currently recruit 45.1% Troll Whelp and Rebels recruit 46.9% Wose
20121027 17:15:59< SeattleDad> For MLR 0.3, that is
20121027 17:16:38< SeattleDad> Okay, got to go
20121027 17:16:56-!- SeattleDad [~SeattleDa@c-67-171-51-0.hsd1.wa.comcast.net] has quit []
20121027 17:42:35-!- faabumc [~vcr@wesnoth/developer/faabumc] has quit [Quit: Ex-Chat]
20121027 17:55:47-!- mattsc [~mattsc@d154-20-32-241.bchsia.telus.net] has quit [Quit: I'm asleep]
20121027 17:59:16-!- mattsc [~mattsc@d154-20-32-241.bchsia.telus.net] has joined #wesnoth-umc-dev
20121027 18:09:22-!- noy [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
20121027 18:10:28< AI0867> mattsc: maybe I've said this before, but I'm currently running a CIA-irker proxy on my server, so commit notifications can be fixed even though CIA is dead
20121027 18:11:14< AI0867> what you need to do is edit the CIA service hook and set the address to http://ai0867.net:8000/RPC2
20121027 18:11:22< AI0867> also, remind me what you have as 'project'
20121027 18:11:30< AI0867> I need to do some local configuring too
20121027 18:12:26< mattsc> AI0867: thanks, I was going to ask you about it.  I'm on my way out the door, will do this when I get back.
20121027 18:12:44< AI0867> k
20121027 18:25:31-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has joined #wesnoth-umc-dev
20121027 18:32:22-!- Blueblaze [~Blueblaze@99.148.246.168] has joined #wesnoth-umc-dev
20121027 19:16:51-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has quit [Quit: i go nstuf kthxbai]
20121027 19:56:45-!- Blueblaze [~Blueblaze@99.148.246.168] has quit [Quit: Blueblaze]
20121027 19:56:51< mattsc> AI0867: I set up the hook and used AI-Demos as project name.
20121027 19:59:21-!- Blueblaze [~Blueblaze@99.148.246.168] has joined #wesnoth-umc-dev
20121027 20:00:39-!- shadowm_laptop [ignacio@wesnoth/developer/shadowmaster] has joined #wesnoth-umc-dev
20121027 20:55:36-!- Alarantalara [~Adium@CPEc0c1c09e8055-CM00252eac6d62.cpe.net.cable.rogers.com] has joined #wesnoth-umc-dev
20121027 20:58:52< mattsc> Hi Alarantalara.  SeattleDad says he gets crashes with Ron, on Aethermaw.
20121027 20:59:28< mattsc> 20121027 15:53:03< SeattleDad> ...h_1.11/data/add-ons/AI-demos/lua/generic-rush_engine.lua:72: bad argument #2 to 'find_path' (invalid location)
20121027 20:59:44< Alarantalara> I see. I haven't tried that map, but I can guess why it might occur.
20121027 21:00:12< mattsc> Ok, so will you look into it?
20121027 21:00:27< Alarantalara> Of course
20121027 21:00:51< mattsc> Thanks.  I will continue to work on the speed issues with Ron and Fred.
20121027 21:00:53< Alarantalara> The crash is only in nogui mode of course
20121027 21:01:28< mattsc> I don't know if you've read the logs of the last 1.5 days or so, I've been whining at you a bit about my own stupidity.  :)
20121027 21:01:47< Alarantalara> I've read the logs
20121027 21:02:15< mattsc> The positive side effect of all that is that the ai_helper functions are much cleaner now.
20121027 21:02:29< Alarantalara> Oh, the reason that remove is so slow is that it copies all the higher numbered items down after each remove, thus the extra time
20121027 21:02:48< mattsc> Well, some of them, I'm not quite done yet.  After that, I'll try to get the remaining 15 seconds down to a much shorter time.
20121027 21:03:09< mattsc> Ok, I assumed that it was something like that.
20121027 21:03:58< mattsc> I use table.remove in other places too, but I think this was the only one where the arrays are that big.  I'll check for that as part of the clean-up process.
20121027 21:05:11< mattsc> Anyways, so unfortunately I am not yet ready to work on tweaking the behavior itself, but the current refactoring will make that easier in the end (again).
20121027 21:06:01< mattsc> Oh, and all the AIs should be working (at least) as good as before again now.
20121027 21:09:28< Alarantalara> Excellent.
20121027 21:09:53< Alarantalara> I've fixed the problem with Aethermaw (trying to consider movement to keeps on the map border)
20121027 21:10:13< mattsc> Perfect.
20121027 21:11:20< mattsc> I'll work on Ron's speed issues for the ML recruiter this afternoon then.
20121027 21:16:50< Alarantalara> Ron doesn't have many CAs. I suspect that most speed issues may be my fault
20121027 21:17:23< Alarantalara> After all, I've added a complicated recruit CA and the vllage hopping one
20121027 21:18:53-!- noy [~Noy@wesnoth/developer/noy] has quit [Quit: noy]
20121027 22:06:50< AI0867> mattsc: AI-Demos should now report to #wesnoth-umc-dev and #commits
20121027 22:46:44< mattsc> AI0867: Thanks.  I just made a commit that didn't show up, but then I don't see the irker online here.
20121027 22:48:15-!- irker421 [~irker@ai0867.net] has joined #wesnoth-umc-dev
20121027 22:48:16< irker421> AI-Demos: mattsc master * r5a59388 / lua/generic-rush_engine.lua : Ron: update stats CA - http://git.io/mkMmkg
20121027 22:48:42< mattsc> Cool!  Thanks.
20121027 22:49:03< AI0867> sh3.rs.github.com - - [27/Oct/2012 22:47:16] "POST /RPC2 HTTP/1.1" 200 <-- it can take up to a minute for irker to connect
20121027 22:49:43-!- Alarantalara [~Adium@CPEc0c1c09e8055-CM00252eac6d62.cpe.net.cable.rogers.com] has quit [Quit: Leaving.]
20121027 22:53:55< mattsc> AI0867: ok, thanks.  Actually, the first comment was my bad.  I'd done a local commit only.   :\ 
20121027 23:05:39< irker421> AI-Demos: mattsc master * r112091f / (lua/generic-recruit_engine.lua lua/generic-rush_engine.lua): Ron: add eval and exec debug messages - http://git.io/XzPBcQ
20121027 23:09:22-!- Crendgrim [~crend@port-92-204-55-61.dynamic.qsc.de] has joined #wesnoth-umc-dev
20121027 23:55:14< irker421> AI-Demos: mattsc master * rf20ebf9 / (lua/generic-recruit_engine.lua lua/generic-rush_engine.lua): Ron: fix a couple bugs in the debug messages - http://git.io/-4oYNw
--- Log closed Sun Oct 28 00:00:17 2012