--- Log opened Sun Oct 28 00:00:17 2012
20121028 00:21:49-!- zookeeper [~lmsnie@wesnoth/developer/zookeeper] has quit [Ping timeout: 252 seconds]
20121028 00:40:46-!- ancestral [~ancestral@50-78-227-230-static.hfc.comcastbusiness.net] has joined #wesnoth-umc-dev
20121028 01:06:24-!- ancestral [~ancestral@50-78-227-230-static.hfc.comcastbusiness.net] has quit [Quit: i go nstuf kthxbai]
20121028 01:30:45-!- Crendgrim [~crend@port-92-204-55-61.dynamic.qsc.de] has quit [Read error: Operation timed out]
20121028 01:43:57-!- noy [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
20121028 01:46:34< Jarkko_P> quit
20121028 01:46:36-!- Jarkko_P [~jake@212-226-75-60-nat.elisa-mobile.fi] has quit [Quit: leaving]
20121028 02:10:48< irker421> wesnoth-umc-dev: doofus-01 * r16121 /trunk/Tales_of_the_Setting_Sun/ (4 files in 3 dirs):
20121028 02:10:49< irker421> replacing [set_variable]rand=... (which does not work like it used to in BfW 1.10.0) with LUA tag [random_number].  The skirmishes should now work again.
20121028 02:22:00< irker421> AI-Demos: mattsc master * rb839d4e / lua/generic-rush_engine.lua : Ron: First attempt to speed up grab_villages CA - http://git.io/LRZjcA
20121028 02:23:42-!- Lukc [~Lukc@natsu.upyum.com] has quit [Remote host closed the connection]
20121028 02:53:02-!- faabumc [~vcr@wesnoth/developer/faabumc] has joined #wesnoth-umc-dev
20121028 02:58:29-!- noy [~Noy@wesnoth/developer/noy] has quit [Quit: noy]
20121028 02:35:42-!- faabumc [~vcr@wesnoth/developer/faabumc] has quit [Quit: Ex-Chat]
20121028 02:43:37-!- loonybot [~loonybot@wesnoth/bot/loonybot] has quit [Remote host closed the connection]
20121028 02:50:03-!- skyfaller [~skyfaller@wikipedia/Skyfaller] has joined #wesnoth-umc-dev
20121028 03:07:09< irker421> AI-Demos: mattsc master * r61b5f26 / lua/generic-rush_engine.lua : Ron grab_villages: use ai.get_attacks, rather than ai_helper.get_attacks - http://git.io/eDzx5g
20121028 03:08:17< mattsc> skyfaller, Alarantalara: with this ^ commit, there's no way back to 1.10 any more
20121028 03:18:04< irker421> AI-Demos: mattsc master * r7dece05 / ais/ai_generic-rush.cfg : Ron: add CA max_score values - http://git.io/sYOdew
20121028 03:27:59-!- Cookie [~quassel@unaffiliated/cookiee] has quit [Quit: No Ping reply in 180 seconds.]
20121028 03:28:27-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121028 03:28:27-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121028 03:28:27-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121028 03:38:34-!- Blueblaze [~Blueblaze@99.148.246.168] has quit [Quit: Blueblaze]
20121028 03:42:40-!- Blueblaze [~Blueblaze@adsl-99-148-246-168.dsl.hstntx.sbcglobal.net] has joined #wesnoth-umc-dev
20121028 03:53:26< skyfaller> good riddance
20121028 03:59:36-!- Cookie is now known as CatwomenCookie
20121028 04:05:16-!- Blueblaze [~Blueblaze@adsl-99-148-246-168.dsl.hstntx.sbcglobal.net] has quit [Quit: Blueblaze]
20121028 04:42:00-!- CatwomenCookie is now known as CatwomanCookie
20121028 04:50:51< irker421> AI-Demos: mattsc master * r967ff13 / CHANGELOG : Update changelog - http://git.io/u7kiEg
20121028 04:54:25-!- SeattleDad [~SeattleDa@c-67-171-51-0.hsd1.wa.comcast.net] has joined #wesnoth-umc-dev
20121028 05:00:54< mattsc> SeattleDad: I did what I could to speed up Ron.
20121028 05:01:32< mattsc> He should be faster than before, but things will always be slower than the RCA AI because of the added functionality.
20121028 05:02:02< SeattleDad> Awesome.
20121028 05:02:24< SeattleDad> So I should download it from GitHub?
20121028 05:02:29< mattsc> Also, Cynsaun Battlefield is a big map with lots of villages and units, so it just takes some time to evaluate and play that map.
20121028 05:02:50< mattsc> Yeah, the current github version is probably about as fast as it will get.
20121028 05:03:01< SeattleDad> Okay, I'll give it a try
20121028 05:04:08< SeattleDad> I think I'll add something to my analyze_log.py to measure time per game, which should give us a better take on time usage
20121028 05:04:23< mattsc> I could probably get the grab_villages CA (the biggest offender) another 20-30% faster with some effort, but the overall speed improvement would be much less than that and it's not really worth the effort, IMO.
20121028 05:04:52< SeattleDad> If you feel like you've harvested the low-hanging fruit, that's all I can really ask
20121028 05:06:18< mattsc> I do.  Fred, on the other hand, is going to be a lot slower than that for quite some while, I'm afraid.
20121028 05:06:39< SeattleDad> Okay
20121028 05:06:49< SeattleDad> Just a question about Fred . . .
20121028 05:07:05< SeattleDad> Do I understand that it is specific to Freelands, side 1
20121028 05:07:06< mattsc> I'm still in the middle of the development there, and I don't really want to spend a lot of time optimizing things that I'll just revert later anyway.
20121028 05:07:15< mattsc> That is correct.
20121028 05:07:21< SeattleDad> Totally the right thing to do
20121028 05:07:45< SeattleDad> One of my favorite quotes is "premature optimization is the root of all evil"
20121028 05:07:55< mattsc> :)
20121028 05:08:13< SeattleDad> Fred is Northerners only or works across all factions?
20121028 05:08:40< mattsc> That said, I do optimize things that take way too long, but that's measured by human player standards, not by running lots of iterations against another AI.
20121028 05:09:02< SeattleDad> Sounds good
20121028 05:09:04< mattsc> Northernes, Freelands and Side 1 only.
20121028 05:09:19< SeattleDad> But against any of the six standard factions?
20121028 05:09:57< mattsc> In principle Fred can play against any faction, but he still has real trouble against the Undead.
20121028 05:10:25< SeattleDad> Is this a recruiting issue?  Against the Undead, I think you'd want to recruit a lot of Troll Whelps
20121028 05:11:07< mattsc> That's correct on both accounts.
20121028 05:11:20< mattsc> Also, assassins are essentially useless.
20121028 05:12:20< SeattleDad> I'm taking a quick look at  ai_grunt-rush-Freelands-S1.cfg
20121028 05:12:34< SeattleDad> Where is the CA that does recruitment
20121028 05:12:58< mattsc> We started with this specialized behavior because we thought it would be easier to figure out.  A lot of things are already getting generalized to work on different maps and/or with different factions, but we're not quite there yet
20121028 05:13:29< mattsc> At the very end: recruit_orcs_eval()
20121028 05:13:41-!- shadowm_laptop [ignacio@wesnoth/developer/shadowmaster] has quit [Quit: shadowm_laptop]
20121028 05:13:51< SeattleDad> Whoops, I see that now
20121028 05:14:22< SeattleDad> So if I was to experiment with plugging the ML Recruiter, I could just comment out that CA and substitute my own, right?
20121028 05:14:40< mattsc> Right.
20121028 05:15:10< mattsc> Comment it out in ais/ai_grunt-rush-Freelands-S1.cfg though.
20121028 05:15:40< SeattleDad> The interesting thing about Fred is that since it's so specific, I may not need as much training data to train a decent Recruiter
20121028 05:16:07< mattsc> Good point
20121028 05:16:17< SeattleDad> Yep, comment it out in the .cfg
20121028 05:17:05< SeattleDad> I think your strategy with Fred makes sense--get it working on a specific case, and then generalize
20121028 05:17:34< SeattleDad> Do you have any idea how you're doing vs. human opponents?
20121028 05:18:01< SeattleDad> Or how you're doing vs. the RCA AI?
20121028 05:18:05< mattsc> Oh, yeah, he's getting clobbered...
20121028 05:18:42< mattsc> against human opponents that is, although he usually scores a victory against people who don't suspect what's going to happen.
20121028 05:19:08< mattsc> The problem with humans is that they learn all the weaknesses extremely quickly, so even noobs usually win on the second game.
20121028 05:19:28< mattsc> I am still working on making that a little harder for them.
20121028 05:19:50< mattsc> Against the RCA AI, if you give Fred 100 gold and the RCA 150, Fred wins >80% of the time.
20121028 05:20:04< SeattleDad> Wow, that's really god
20121028 05:20:15< SeattleDad> really good
20121028 05:21:08< SeattleDad> How would we know if the ML Recruiter is making Fred better?
20121028 05:21:26< SeattleDad> Maybe by improving the win ratio on 100 gold vs. 150?
20121028 05:21:38< mattsc> Hmm, right, we cannot play Fred against Fred...
20121028 05:21:46< skyfaller> sure, that would be a start... it's hard to quantify improvement against humans
20121028 05:21:57< skyfaller> mattsc: yeah, I really want to get Fred to play side 2 ASAP
20121028 05:22:09< skyfaller> that would be my next priority once we finish Fred in his current state
20121028 05:22:29< mattsc> Fred has a pretty hard time against Ron (plays barely better than 50/50 last time I checked), so that might be a good test.
20121028 05:22:51< SeattleDad> Okay, that makes sense
20121028 05:22:51< mattsc> skyfaller: agreed.  And believe me, I am working on it.
20121028 05:23:07< mattsc> I'm not doing all this refactoring because I like refactoring so much.  :)
20121028 05:23:12< skyfaller> lol
20121028 05:23:31< skyfaller> are you sure? I hear refactoring is a blast
20121028 05:23:50< SeattleDad> I could gather training data on Ron vs. Fred.  That would give me training data for both sides
20121028 05:24:04< mattsc> Well, sometimes when you're stuck it's easier than trying to figure out the next step, that's true...
20121028 05:24:21< SeattleDad> You know, I've built a nice little unit testing architecture for refactoring around ai_test2.py 
20121028 05:24:38< mattsc> SeattleDad: that sounds good.  I think Fred has a bit of an edge by now, but I don't know by how much.
20121028 05:24:48< mattsc> Certainly not the same margin as against the RCA AI.
20121028 05:25:38< mattsc> SeattleDad: I'm not sure I understand your last comment to its full extent...
20121028 05:26:13< SeattleDad> So my ai_test2.py is a script for running large batches of Wesnoth games . . .
20121028 05:26:25< mattsc> right
20121028 05:26:39-!- artisticdude_iOS [~artisticd@152.sub-70-192-198.myvzw.com] has joined #wesnoth-umc-dev
20121028 05:26:41< SeattleDad> There's an existing version of it in utils, but it's completely bugged in the mainline code base
20121028 05:26:45-!- artisticdude_iOS [~artisticd@152.sub-70-192-198.myvzw.com] has left #wesnoth-umc-dev []
20121028 05:27:48< SeattleDad> You can configure my ai_test2.py to run only a single games, run it with --rng-seed 0, and run with a specific map, side, and faction
20121028 05:28:25< SeattleDad> It outputs the units recruited along with all the "features" that went into make the decisions
20121028 05:28:49< SeattleDad> If this feature file is identical from run a to run b, then you're 99% sure that the games were identical
20121028 05:29:07< SeattleDad> This is what you want when refactoring
20121028 05:29:43< mattsc> SeattleDad: I see.  That makes sense.
20121028 05:30:26< mattsc> Besides the rng-seed, I'm doing a lot of command-line stuff with outputs also.
20121028 05:30:43< mattsc> However, for the refactoring, I usually use a different mechanism.
20121028 05:31:33< mattsc> I have a test scenario where I can call individual CAs (or even parts of individual CAs) with a right-click option.
20121028 05:32:25< mattsc> But what you say makes a lot of sense for overall in-context testing.
20121028 05:33:01< mattsc> [I have no real coding background, I just make things up along the way as I need them  :) ]
20121028 05:33:12< SeattleDad> I have something like six different config files to test different combinations when I'm doing refactoring
20121028 05:33:31< SeattleDad> Then I recently set up a bash script to iterate through all of them
20121028 05:33:47< SeattleDad> Well, it sounds like you're doing pretty well . . .
20121028 05:34:01< SeattleDad> Lots of people have ended up in coding who started in other professions
20121028 05:34:07< mattsc> That sounds pretty good.
20121028 05:35:04< SeattleDad> It seems a little odd that Fred is only about 50/50 vs. Ron
20121028 05:35:16< mattsc> (your testing method, I mean)
20121028 05:35:42< skyfaller> Basically Fred is not sufficiently aggressive, Ron maintains the RCA's aggression while benefiting from a few key improvements
20121028 05:35:48< mattsc> Well, as I said, I think he's a bit better by now.
20121028 05:36:34< SeattleDad> One thing that's struck me about RCA AI is that it doesn't retreat units when injured
20121028 05:36:41< SeattleDad> Or at least not enough
20121028 05:36:51< SeattleDad> Lots of attacks with very weak units
20121028 05:37:04< SeattleDad> So in that sense I'd say RCA is overly aggressive
20121028 05:38:11< mattsc> SeattleDad: agreed.  However, in Wesnoth there is one huge advantage of attacking over defending: choice of weapon.
20121028 05:38:37< skyfaller> that, and units that die don't get a chance to counter-attack
20121028 05:39:15< mattsc> So, if you want a (comparatively) simple AI (not that the RCA is all that simple), attacking under almost all circumstances is a pretty good strategy.
20121028 05:39:21< skyfaller> so if you can hit hard enough in your first strike, the defenders can't adequately punish you
20121028 05:40:06< mattsc> But yes, getting this balance between attack and retreat right is what I've been working on with Fred for a long time now.  And I still haven't quite figured it out.
20121028 05:40:28< skyfaller> mattsc: I still want to play some games with you sometime ;-)
20121028 05:40:53< mattsc> skyfaller: you just want some easy wins!  :D
20121028 05:40:57< skyfaller> :D
20121028 05:41:08< skyfaller> if you don't play, how can you teach Fred to play?!
20121028 05:41:19< skyfaller> knowing is half the battle!
20121028 05:41:19< SeattleDad> I'm always getting clobbered in MP games by more experienced players too . . .
20121028 05:41:28< mattsc> But yes, we should do that.  Sometime.
20121028 05:41:49< mattsc> I know.  I can still easily beat Fred myself though...
20121028 05:42:02< skyfaller> fair enough :)
20121028 05:42:21< skyfaller> but you wrote Fred :P
20121028 05:42:22< mattsc> But seriously, let's do that sometime.  There just never seems enough time.
20121028 05:42:29< mattsc> Right, that helps.  A lot.
20121028 05:42:46< skyfaller> well what are you doing now?
20121028 05:42:57< skyfaller> let's do a fast game and then let SeattleDad play the winner ;-)
20121028 05:43:10< SeattleDad> Sounds good to me.  
20121028 05:43:39< SeattleDad> I just played Aethermaw the first time today so that I could understand what was going on there
20121028 05:43:48< skyfaller> hah, that map is a nightmare for AI
20121028 05:44:02< SeattleDad> Yes, very weird
20121028 05:44:10< skyfaller> I think Fred will have an advantage over any AI that isn't map-specific once we get to Aethermaw
20121028 05:44:34< SeattleDad> Although it doesn't seem to impact the ML Recruiter much
20121028 05:44:40< skyfaller> really?
20121028 05:44:50< SeattleDad> The thing about recruiting is that the recruiter cheats . . .
20121028 05:45:02< SeattleDad> It can see what its opponent has recruited 
20121028 05:45:18< skyfaller> that's true, but does it know how to take advantage of positioning and when the map opens?
20121028 05:45:21< skyfaller> all AIs cheat
20121028 05:45:28< skyfaller> I'm saying that Fred will cheat better once we get there :)
20121028 05:45:44< SeattleDad> Sounds good
20121028 05:46:35< mattsc> skyfaller: sorry, can't do tonight.  In fact, I really should have signed off about 15min ago...
20121028 05:46:45< SeattleDad> By the way, I'm thinking that replacing AI_CA_HEALING might be an initial target for using a neural net on something other than recruiting . . .
20121028 05:47:06< SeattleDad> mattsc, I guess we might have to let you go
20121028 05:47:25< mattsc> SeattleDad: don't run tests of Fred vs. Ron right now.  I am in the middle of changing things and Fred's not playing at his best at the moment.
20121028 05:47:28< skyfaller> lol
20121028 05:47:42< skyfaller> SeattleDad: well, I could still play you now :)
20121028 05:47:49< SeattleDad> Okay, I'll concentrate my tests on Ron
20121028 05:48:14< mattsc> For several weeks now I've been thinking that within a couple days I'll have him up to better than ever, but that moment always stays a couple days away...
20121028 05:48:33< SeattleDad> I know how it is . . .
20121028 05:49:07< mattsc> skyfaller: about this time of day, sometime during the week next week would probably work
20121028 05:49:20< mattsc> not Tuesday, though
20121028 05:49:33< skyfaller> ok :) we'll see whether our schedules mesh, I guess
20121028 05:50:04< mattsc> yeah, sorry, "real life's" been very busy lately...
20121028 05:50:50< SeattleDad> skyfaller, I could play a game
20121028 05:50:55< mattsc> Well, I better sign off before this interesting discussion picks up again.  :)
20121028 05:50:56< skyfaller> mattsc: what needs to happen before Fred can play at his best again?
20121028 05:51:05< skyfaller> ha, ok, tell me some other time
20121028 05:51:11< mattsc> well, briefly:
20121028 05:51:39< mattsc> I need to finish cleaning up some of the ai_helper functions and make sure individual moves don't take 800 seconds.  ;)
20121028 05:51:45< mattsc> That's mostly done.
20121028 05:52:25< mattsc> Then I need to absorb a couple more things into zone_control (attacks on weak enemy, poison attacks, etc.)
20121028 05:52:43< mattsc> And at that point it's down to balancing, tuning, whatever you want to call it.
20121028 05:53:06< skyfaller> mattsc: ok, that doesn't sound too bad
20121028 05:53:14< skyfaller> I guess let me know when it is safe to start creating test cases again
20121028 05:53:26< mattsc> No.  As I said, it should only be a couple days, right?  :P
20121028 05:53:42< skyfaller> I really want to have a battery of tests we can run Fred through every time we update him to check for regressions in play
20121028 05:54:01< skyfaller> is Fred "passing" the one test case I've made, currently?
20121028 05:54:44< mattsc> I haven't tested recently.  I don't think he's perfect with that one yet, but I think he does better than he used to at least.
20121028 05:54:55< skyfaller> SeattleDad: I'm on the Wesnoth server, join up and we can duel it out :D
20121028 05:55:14< SeattleDad> Okay, I'll get on
20121028 05:55:17< mattsc> Cool, I'll let you do that and go on to the next order of business...
20121028 05:55:33< skyfaller> mattsc: yeah, any move where he retakes two villages that turn is a passing grade, other improvements are bonuses
20121028 05:56:00< skyfaller> failing to reclaim villages is a fatal error in that situation, which is why I created that test case first
20121028 05:56:23< mattsc> Agreed.
20121028 05:56:39< mattsc> Well, have fun!  Let's play sometime next week and bye for now.
20121028 05:56:43< skyfaller> ok, good night!
20121028 05:57:01< mattsc> Not yet...  :(
20121028 05:58:00< SeattleDad> I'm on the server
20121028 05:58:06-!- mattsc [~mattsc@d154-20-32-241.bchsia.telus.net] has quit [Quit: bye]
20121028 05:58:18< skyfaller> ok
20121028 05:59:00< skyfaller> SeattleDad: do you want a friendly match or a ladder game? :D
20121028 05:59:28< SeattleDad> I guess I don't know what a ladder game is . . .
20121028 05:59:33< skyfaller> ha
20121028 05:59:38< skyfaller> it's a competitive ranking
20121028 05:59:44< skyfaller> like the ELO system in chess
20121028 05:59:51< SeattleDad> Okay, that's what I thought
20121028 06:00:03< SeattleDad> I'd be happy to do a ladder game
20121028 06:00:04< skyfaller> the only problem is that there are currently two incompatible ranking systems / websites at the moment
20121028 06:00:21< SeattleDad> I'm on the official server
20121028 06:00:24< skyfaller> which has resulted in a bit of a schism in the ladder community
20121028 06:00:55< SeattleDad> How do you specify a game is laddered?
20121028 06:00:59< skyfaller> I'm going to recommend you start by creating an account at http://wesnoth.gamingladder.info/
20121028 06:01:12< skyfaller> that is the "old ladder"
20121028 06:01:19< skyfaller> and it is easier to register
20121028 06:01:46< skyfaller> the new ladder, "ladder8" is at http://ladder8.herokuapp.com/ and has a more complicated registration process that is supposed to make it harder to cheat
20121028 06:02:09< skyfaller> but I think it just discourages people from joining :P
20121028 06:02:22< skyfaller> that said, it's much nicer and better maintained
20121028 06:02:24< SeattleDad> Okay, working on it
20121028 06:02:30< skyfaller> so basically it's a bloody disaster
20121028 06:02:39< skyfaller> having two ladders and neither will surrender
20121028 06:04:05< skyfaller> ok, basically once you succeed in getting an account on the ladder, I'll title the game as a ladder game and the winner will report the game to the ladder website
20121028 06:04:44< SeattleDad> Okay, so I signed up as SeattleDad
20121028 06:04:51< skyfaller> sweet
20121028 06:04:57< SeattleDad> Should work now
20121028 06:05:08< skyfaller> ok
20121028 06:05:37< skyfaller> ok, I created a game for you, join it :)
20121028 06:05:46< SeattleDad> Got it.  Joining now
20121028 06:07:03< skyfaller> oh crap
20121028 06:07:05< skyfaller> something went wrong
20121028 06:56:24-!- CatwomanCookie is now known as Cookie
20121028 07:02:50-!- Netsplit *.net <-> *.split quits: skyfaller
20121028 07:09:32-!- SeattleDad [~SeattleDa@c-67-171-51-0.hsd1.wa.comcast.net] has quit []
20121028 07:27:34-!- Cookie [~quassel@unaffiliated/cookiee] has quit [Remote host closed the connection]
20121028 07:43:13-!- Netsplit *.net <-> *.split quits: @ChanServ, +shikadibot
20121028 07:48:55-!- Netsplit over, joins: +shikadibot, @ChanServ
20121028 08:25:09-!- zookeeper [~lmsnie@wesnoth/developer/zookeeper] has joined #wesnoth-umc-dev
20121028 08:43:01-!- Netsplit *.net <-> *.split quits: @ChanServ, +shikadibot
20121028 09:03:07-!- Crendgrim [~crend@port-92-204-24-230.dynamic.qsc.de] has joined #wesnoth-umc-dev
20121028 09:30:15-!- Netsplit over, joins: +shikadibot, @ChanServ
20121028 09:31:55-!- noy [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
20121028 09:42:07-!- Crendgrim is now known as Guest16617
20121028 09:58:18-!- skyfaller [~skyfaller@wikipedia/Skyfaller] has joined #wesnoth-umc-dev
20121028 09:58:47-!- irker421 [~irker@ai0867.net] has quit [Quit: transmission timeout]
20121028 10:05:59-!- noy [~Noy@wesnoth/developer/noy] has quit [Quit: noy]
20121028 10:08:33-!- Netsplit *.net <-> *.split quits: skyfaller
20121028 10:13:53-!- skyfaller [~skyfaller@wikipedia/Skyfaller] has joined #wesnoth-umc-dev
20121028 10:34:27-!- Netsplit *.net <-> *.split quits: skyfaller, @ChanServ, +shikadibot
20121028 10:38:28-!- Netsplit over, joins: skyfaller, +shikadibot, @ChanServ
20121028 11:13:03-!- Guest16617 [~crend@port-92-204-24-230.dynamic.qsc.de] has quit [Ping timeout: 260 seconds]
20121028 11:21:34-!- Guest16617 [~crend@port-92-204-24-230.dynamic.qsc.de] has joined #wesnoth-umc-dev
20121028 11:22:50-!- Guest16617 is now known as Crendgrim
20121028 13:15:11-!- loonybot [~loonybot@ppp94-29-12-11.pppoe.spdop.ru] has joined #wesnoth-umc-dev
20121028 13:15:11-!- loonybot [~loonybot@ppp94-29-12-11.pppoe.spdop.ru] has quit [Changing host]
20121028 13:15:11-!- loonybot [~loonybot@wesnoth/bot/loonybot] has joined #wesnoth-umc-dev
20121028 13:15:14-!- mode/#wesnoth-umc-dev [+v loonybot] by ChanServ
20121028 14:44:39-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has joined #wesnoth-umc-dev
20121028 15:08:28-!- Alarantalara [~Adium@CPEc0c1c09e8055-CM00252eac6d62.cpe.net.cable.rogers.com] has joined #wesnoth-umc-dev
20121028 15:48:35-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has joined #wesnoth-umc-dev
20121028 15:48:35-!- Cookie [~quassel@60-240-54-150.tpgi.com.au] has quit [Changing host]
20121028 15:48:35-!- Cookie [~quassel@unaffiliated/cookiee] has joined #wesnoth-umc-dev
20121028 15:56:33-!- vultraz [~chatzilla@wesnoth/developer/vultraz] has quit [Read error: Connection reset by peer]
20121028 15:57:48-!- vultraz [~chatzilla@124.109.10.167] has joined #wesnoth-umc-dev
20121028 16:02:06-!- vultraz [~chatzilla@124.109.10.167] has quit [Changing host]
20121028 16:02:06-!- vultraz [~chatzilla@wesnoth/developer/vultraz] has joined #wesnoth-umc-dev
20121028 16:43:39-!- noy [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
20121028 17:03:22-!- noy [~Noy@wesnoth/developer/noy] has quit [Quit: noy]
20121028 17:05:35-!- mattsc [~mattsc@d154-20-32-241.bchsia.telus.net] has joined #wesnoth-umc-dev
20121028 17:12:10< mattsc> Alarantalara: FYI, there's no problem with your CAs (computation time wise) in Ron.
20121028 17:12:41< mattsc> The time sink is grab_villages, and I've reduced that about as much as I can if I want to keep the current functionality.
20121028 17:12:42< Alarantalara> excellent
20121028 17:35:07-!- irker834 [~irker@ai0867.net] has joined #wesnoth-umc-dev
20121028 17:35:08< irker834> AI-Demos: mattsc master * r37b08ce / (lua/ai_helper.lua lua/grunt-rush-Freelands-S1_engine.lua): ai_helper: rename attack_combo functions - http://git.io/Mtf8kg
20121028 17:47:59< irker834> AI-Demos: mattsc master * r609693a / lua/ai_helper.lua : ai_helper.get_attack_combos_full(): clean up comments - http://git.io/8tS71A
20121028 18:16:29< irker834> AI-Demos: mattsc master * r8b65a0b / lua/ai_helper.lua : ai_helper.get_attack_combos_full(): change return value format - http://git.io/0L2tNQ
20121028 18:20:12< vultraz> hm...people working on projects...
20121028 18:20:19 * vultraz is tempted to go work on NX
20121028 18:35:38< irker834> AI-Demos: mattsc master * re71c478 / (lua/ai_helper.lua lua/grunt-rush-Freelands-S1_engine.lua): ai_helper.get_attack_combos(): modify how cfg parameter is passed - http://git.io/J0YCbA
20121028 18:38:37-!- shadowm_laptop [ignacio@wesnoth/developer/shadowmaster] has joined #wesnoth-umc-dev
20121028 19:10:40< vultraz> what was I last working on, anway....
20121028 19:10:53< vultraz> I think it was refactoring the item and inventory code
20121028 19:12:02< vultraz> er..why was I doing that...
20121028 19:12:05< vultraz> oh yeah
20121028 19:12:18< vultraz> adding the functionality to the other buttons
20121028 20:05:21< irker834> AI-Demos: Alarantalara master * r4feb786 / lua/generic-rush_engine.lua : Only print information for sides with a leader - http://git.io/1o10bQ
20121028 20:05:22< irker834> AI-Demos: Alarantalara master * r40a5359 / lua/generic-recruit_engine.lua : Account for opponents' recruit lists in recruiting - http://git.io/RxN_RQ
20121028 20:05:40< irker834> wesnoth-umc-dev: doofus-01 * r16122 /trunk/Tales_of_the_Setting_Sun/ (8 files in 2 dirs):
20121028 20:05:41< irker834> First pass at upper Agoratera scenarios (702b group)
20121028 20:27:22-!- noy [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
20121028 20:35:33-!- Crendgrim [~crend@port-92-204-24-230.dynamic.qsc.de] has quit [Quit: Konversation terminated!]
20121028 20:35:42-!- Crendgrim [~crend@port-92-204-24-230.dynamic.qsc.de] has joined #wesnoth-umc-dev
20121028 20:53:41-!- noy [~Noy@wesnoth/developer/noy] has quit [Quit: noy]
20121028 21:03:24-!- noy [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
20121028 21:11:06-!- Blueblaze [~Blueblaze@2602:306:394f:6a89:6233:4bff:fe0a:827b] has joined #wesnoth-umc-dev
20121028 21:22:49-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has quit [Quit: i go nstuf kthxbai]
20121028 21:23:12< irker834> AI-Demos: Alarantalara master * r37e9822 / lua/generic-recruit_engine.lua : Increase importance of absolute hp - http://git.io/M17_ww
20121028 21:23:13< irker834> AI-Demos: Alarantalara master * rfc4eecc / lua/generic-recruit_engine.lua : Adjust relative importance of attacks - http://git.io/pQUxDQ
20121028 21:23:14< irker834> AI-Demos: Alarantalara master * r771e130 / lua/generic-recruit_engine.lua : Increase value of "slow" - http://git.io/FUVbqA
20121028 21:23:51-!- noy_ [~Noy@d50-98-14-63.bchsia.telus.net] has joined #wesnoth-umc-dev
20121028 21:23:51-!- noy_ [~Noy@d50-98-14-63.bchsia.telus.net] has quit [Changing host]
20121028 21:23:51-!- noy_ [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
20121028 21:26:41-!- noy [~Noy@wesnoth/developer/noy] has quit [Ping timeout: 256 seconds]
20121028 21:26:41-!- noy_ is now known as noy
20121028 21:39:19< irker834> AI-Demos: Alarantalara master * r4af0063 / lua/generic-recruit_engine.lua : Adjust importance of movement to map size - http://git.io/AJRALw
20121028 21:43:18-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has joined #wesnoth-umc-dev
20121028 21:48:23< irker834> AI-Demos: Alarantalara master * rad1e58d / CHANGELOG : Make note of bug fix in changelog - http://git.io/7HvfuA
20121028 22:02:55< mattsc> Alarantalara: I'd previously gotten your final move evaluation time down from 815s to 15s.  I now managed to reduce it to 5.5s.
20121028 22:03:06< Alarantalara> Excellent!
20121028 22:04:05< mattsc> Well, yes and no.
20121028 22:04:20< Alarantalara> I've hopefully just finished another round of edits to the recruiter, and while 20 games isn't a good sample, I have a 70% win rate against the ml recruiter so far
20121028 22:04:31< mattsc> That's about as far as I can get it down, I think, given that there are something like 300,000 unique combinations.
20121028 22:04:39< mattsc> Cool!
20121028 22:05:47< Alarantalara> Is it possible to reject some combinations without evaluating them?
20121028 22:05:54< mattsc> 5.5s to evaluate a move that obviously (to the human eye) is no good is still a long time.  But I guess I can only get it down further if I put in an a priori rejection...
20121028 22:06:12< mattsc> We're thinking the same thing there...
20121028 22:07:02< Alarantalara> Well, from the human point of view, no move is good. The only reason humans would be faster at that point is that the only considerations would be minimizing damage taken and possible escaping
20121028 22:07:04< mattsc> So, yeah, probably we can do that under some circumstances.  But the will always be some that simply need to be evaluated brute force.
20121028 22:07:54< mattsc> Anyways, I'm going take off for a while, and tonight I'll clean that change up and commit it.
20121028 22:08:26< mattsc> After that, I think I'll go back to trying to work on the actual behavior again.  Fred's consistenly losing to Ron these days, and that's bugging me.  :)
20121028 22:09:25< Alarantalara> I can imagine, given how much less Ron does
20121028 22:10:26< mattsc> I still think that what I've been doing recently are actual improvements, I just need to get the weights right...
20121028 22:38:35-!- ancestral [~ancestral@75-161-178-116.mpls.qwest.net] has quit [Quit: i go nstuf kthxbai]
20121028 22:39:17< skyfaller> mattsc: hey Matt, is there anything that needs testing currently?
20121028 22:42:01< skyfaller> Alarantalara: what's new? :)
20121028 22:42:46< Alarantalara> Ron's recruiting is noticeably better
20121028 22:43:26< skyfaller> is that affecting Fred at all?
20121028 22:44:11< Alarantalara> Not yet, but if the test series I'm currently running against the ML recruiter is good enough, I will probably switch Fred over as well
20121028 22:44:28< skyfaller> what have you been doing to Ron's recruiting?
20121028 22:45:06< Alarantalara> Normalization of results to avoid having to deal with extremes caused by some factions
20121028 22:46:51< Alarantalara> Ron's favourite units by faction are now: glider/augur, dwarf fighter, spearman, grunt, elf fighter, and DA
20121028 22:49:26-!- Crendgrim [~crend@port-92-204-24-230.dynamic.qsc.de] has quit [Quit: Konversation terminated!]
20121028 22:50:39< Alarantalara> Things I don'tlike: Too many ghosts, goblins guardsmen, not enough assassins, thunderers
20121028 22:51:02< skyfaller> glider/augur is actually a reasonable combination in many circumstances
20121028 22:51:27< skyfaller> the main problem is not having any tanks
20121028 22:51:40< Alarantalara> Next most recruited there is clasher
20121028 22:51:44< skyfaller> I also question dwarf fighter spam because dwarf fighters usually suck at attacking
20121028 22:52:10< skyfaller> ok, as long as there are clashers to hold the front line I guess augurs+gliders is plausible
20121028 22:54:46< skyfaller> Alarantalara: wanna play a game? :)
20121028 22:54:52< Alarantalara> sure
20121028 22:54:57< skyfaller> ok
20121028 22:55:06< Alarantalara> 1.10?
20121028 22:55:09< skyfaller> sure
20121028 22:55:16< skyfaller> do you want to play a ladder game, or a friendly game?
20121028 22:55:32< Alarantalara> It doesn't matter to me
20121028 22:55:44< skyfaller> ok, let's do a ladder game, I'll make
20121028 22:58:12< skyfaller> http://ladder8.herokuapp.com/
20121028 23:00:49< skyfaller> Alarantalara: http://ladder8.herokuapp.com/players/signup is where you sign up
20121028 23:06:23-!- noy [~Noy@wesnoth/developer/noy] has quit [Quit: noy]
20121028 23:08:09-!- zookeeper [~lmsnie@wesnoth/developer/zookeeper] has quit [Ping timeout: 252 seconds]
20121028 23:44:45-!- noy [~Noy@wesnoth/developer/noy] has joined #wesnoth-umc-dev
--- Log closed Mon Oct 29 00:00:20 2012