2011-01-02T00:04:38 antimatroid: draw a graph differently? 2011-01-02T00:15:09 sigh: you index a single variable graph (x,y) but x is like the columns axes, and y the row, and row goes up for increasing rather than down 2011-01-02T00:15:37 i am pedantic about things being consistent :P 2011-01-02T00:15:57 it caused me a lot of confusion when learning matrices that it was switched though 2011-01-02T00:17:05 even now, i sometimes come across people that want to index 2d vectors as [x][y] (graph) rather than [i][j] (matrix), and for god knows what reason, my graphics lecturer insisted we use xzy instead of xyz when drawing coordinate systems in 3d 2011-01-02T00:17:12 oh, your mean like graphs that you plot with 2011-01-02T00:17:15 that is confusing as hell when i'm used to xyz 2011-01-02T00:17:18 yes 2011-01-02T00:17:37 thought you meant the other kind, since there was a discussion about graph theory before 2011-01-02T00:17:52 yeah sorry, i do dislike those two having the same name too :P 2011-01-02T00:18:12 well, kind of 2011-01-02T00:29:06 * antimatroid is hating on c++ templates 2011-01-02T00:29:44 are you trying to implement a tensor template again? :P 2011-01-02T00:30:04 i'm playing around with my entire game theory library (doing a big clean up) 2011-01-02T00:30:18 ah, fair enough 2011-01-02T00:30:27 i hate that i can't declare stuff in headers then define it all in the cpp file 2011-01-02T00:30:35 everything gets so messy 2011-01-02T00:30:53 and i hate not being allowed to use std namespace 2011-01-02T00:30:53 yeah, because c++ wanted to be compatible with c 2011-01-02T00:31:02 essentially a pr move :P 2011-01-02T00:32:30 is it considered bad if i just have my own namespace for a group of files 2011-01-02T00:32:38 cause then one doesn't have to worry about linking crap 2011-01-02T00:32:51 you don't? 2011-01-02T00:32:53 what do you mean? 2011-01-02T00:33:17 nope, if i just add everything to a namespace and include that namespace in my main file, then no linking is required 2011-01-02T00:33:29 it's cool, but i'm not sure if that's considered bad or not 2011-01-02T00:34:22 I'm not sure how namespaces relate to linking *shrug* 2011-01-02T00:35:05 i have no idea, i just know it makes ones life a lot easier 2011-01-02T00:48:22 *** delt0r___ has joined #aichallenge 2011-01-02T00:49:33 *** Utkarsh has quit IRC (Read error: Connection reset by peer) 2011-01-02T00:49:49 *** Utkarsh has joined #aichallenge 2011-01-02T00:50:08 *** delt0r_ has quit IRC (Ping timeout: 264 seconds) 2011-01-02T01:18:01 *** yasith_ has quit IRC (Ping timeout: 276 seconds) 2011-01-02T01:29:11 *** yasith_ has joined #aichallenge 2011-01-02T01:31:33 anyone here do much c++ programming? 2011-01-02T01:31:37 dmj111: you about? :P 2011-01-02T02:13:57 *** bhasker has quit IRC (Quit: bhasker) 2011-01-02T02:36:01 *** Rubicon-|-Cross has quit IRC (Quit: Leaving) 2011-01-02T02:38:28 *** bhasker has joined #aichallenge 2011-01-02T02:49:21 *** wh1teside_ has joined #aichallenge 2011-01-02T02:49:57 *** wh1teside has quit IRC (Ping timeout: 240 seconds) 2011-01-02T03:06:12 *** boegel has joined #aichallenge 2011-01-02T04:03:17 antimatroid: Lots of C++ 2011-01-02T04:03:34 there is a reason why I avoid C/C++ when i can ;) 2011-01-02T04:03:55 However most of my C++ was before the "boost" days... 2011-01-02T04:04:33 also when i started compilers will still having a hard time with templates 2011-01-02T04:10:30 *** Palmik has joined #aichallenge 2011-01-02T04:11:59 *** mega1 has joined #aichallenge 2011-01-02T04:13:01 delt0r___: i haven't been using boost anyway 2011-01-02T04:13:23 i wanted to keep it so someone could use it without installing additional libraries 2011-01-02T04:16:10 hmmm i am starting to really despise c++ 2011-01-02T04:16:24 maybe i should just jump ship to haskell 2011-01-02T04:16:44 *** yasith_ has quit IRC (Ping timeout: 272 seconds) 2011-01-02T04:17:23 well for a reference implementation i would recomend C++/Java/Python 2011-01-02T04:17:42 i couldn't use java 2011-01-02T04:17:49 and python doesn't have function overloading does it? 2011-01-02T04:17:54 Haskell quite difficult to do properlly like mondas 2011-01-02T04:17:54 can i overload operators at least? 2011-01-02T04:17:59 yes it does 2011-01-02T04:18:01 well, at least define them 2011-01-02T04:18:18 and why not java... it copies its syntax from C/C++ 2011-01-02T04:18:44 can i define operators and overload functions/ make variables for different scopes etc.? 2011-01-02T04:18:52 i couldn't go back on being able to do that 2011-01-02T04:19:26 well that would make you a not a very good programmer 2011-01-02T04:19:37 i never claimed to be a good programmer :) 2011-01-02T04:19:43 switching syles and lanuages is the mark of programming 2011-01-02T04:20:05 An "operator" is just a badly named function 2011-01-02T04:20:06 yes, but i like being able to reuse stuff, otherwise i'd end up with lots of different names for things 2011-01-02T04:20:14 ?? 2011-01-02T04:20:15 what about operators for matrices? 2011-01-02T04:20:17 wtf 2011-01-02T04:20:34 i don't want to type mult(A, B), A*B is nicer 2011-01-02T04:20:45 I think your problem is that you just need to write more code... I mean C++ didn't really work out that well for you so far 2011-01-02T04:21:22 You seem to have all these ideas about whats good and bad, that really have little to do with what really is good or bad 2011-01-02T04:21:24 even better would be if i could set operator precedence, so I could write A*B^4 rather than A*(B^4) 2011-01-02T04:21:39 my idea of good vs bad is the amount i have to type 2011-01-02T04:22:11 and i need to be able to understand what i wrote easily 2011-01-02T04:22:24 then perl is perfect for you 2011-01-02T04:22:41 it's much harder to understand an equation of the form 2011-01-02T04:22:41 mult(A, pow(B, 4)) for example 2011-01-02T04:22:45 typing amounts to less that .01% of the time coding 2011-01-02T04:22:50 and even worse as it gets more complex 2011-01-02T04:23:00 saving typing at the expense of other things is silly 2011-01-02T04:23:12 i am a very silly person 2011-01-02T04:23:17 also in C++ you cannot alter the copy symatics 2011-01-02T04:23:43 where if you use methods you can have things done in place if you need it for performance reasons 2011-01-02T04:23:56 for operators 2011-01-02T04:24:08 :s? 2011-01-02T04:24:37 lisp does it right (+ (* 3 4) (* 7 8) (* (- a b))) 2011-01-02T04:25:09 would i prefer lisp or perl to haskell? 2011-01-02T04:25:17 everyone says mathsy people love haskell 2011-01-02T04:25:21 and so far it seems good 2011-01-02T04:25:27 but haven't really learnt heaps yet 2011-01-02T04:25:29 perl is a horrible lang ... even the coder can't read their own code 2011-01-02T04:25:47 often folks joke about "write once, read never" 2011-01-02T04:25:49 code 2011-01-02T04:26:06 haskell is not a very "popular" lanuage... 2011-01-02T04:26:18 so "people love it" is very relative... 2011-01-02T04:26:31 is worth learning... just to teach you to think differently 2011-01-02T04:26:44 i don't need popular, most things i will ever program no one else will ever read 2011-01-02T04:26:48 but otherwise.... getting real things done... it has problems 2011-01-02T04:27:13 unless you want to learn arrows/monads which is still have problems with 2011-01-02T04:27:58 You want popular if you want someone else you be able to look at your code... or you want compiliers that are mature for different platforms 2011-01-02T04:28:16 but haskell pretty much has just one "complier" these days 2011-01-02T04:28:18 the GHC 2011-01-02T04:29:02 while look at lisp... there are not just many free versions, there are quite a few commercial ones as well 2011-01-02T04:29:30 And trust me... if others don't look at your code... bugs will lurk... 2011-01-02T04:29:56 Its very -------Humiliating keeping your source code open :D 2011-01-02T04:30:29 yeah, my code is horrible so people get a good chuckle when i show them anything 2011-01-02T04:30:38 and I thought we are talking about a reference impl for this comp. In which case a "popular" language is a really really good idea.. 2011-01-02T04:30:58 nah, what i'm playing around with atm is my own personal game theory library 2011-01-02T04:31:15 i think they want to use python for the contest 2011-01-02T04:31:22 i really ought to learn more of that sometime too 2011-01-02T04:31:28 writing a library is very hard to do 2011-01-02T04:31:57 Its one area where i would be very strict with "test drivien design" 2011-01-02T04:32:33 my friend and i have already been working on it for over a year, but i'm at the point where i don't want to work on algo's anymore before getting an actual backend structurally sound 2011-01-02T04:32:54 and i've never been happy with it in c++ 2011-01-02T04:33:08 if i use templates everything has to be in header files, which is just messy 2011-01-02T04:34:36 C++ instance on many files for a single class is very annoying... After useing java, you realize how silly it is... it was a throw back to the C header files that serve a real purpose.. 2011-01-02T04:35:22 Of course if you are not doing templates then it can still serve a purpose... 2011-01-02T04:37:02 antimatroid: Sorry earlier i sounded pretty harsh (about good programing)... It was not my intention...Sounded nice out loud the way i said it out loud 2011-01-02T04:37:44 *** fgump has quit IRC (Quit: Leaving) 2011-01-02T04:41:09 delt0r___: no offense taken :) 2011-01-02T04:41:29 oh --good 2011-01-02T04:41:31 i don't intend to work as a programmer and have terrible programming practices 2011-01-02T04:47:24 *** boegel has quit IRC (Quit: Leaving) 2011-01-02T05:04:43 antimatroid: You still can improve them... they will save you time.. even for small scripts 2011-01-02T05:18:21 i am still learning as i go along 2011-01-02T05:18:30 delt0r___: perl is not that bad... you can write incomprehensible code in any language 2011-01-02T05:18:48 but i'm mostly interested in doing stuff how i find it easiest, rather than what's considered better for lots of people working on something etc. 2011-01-02T05:19:17 yea.. but in perl you can't *not* write incomprehensible code -- 2011-01-02T05:19:19 antimatroid: I suggest playing around with python more, then 2011-01-02T05:19:27 Sorry worked with perl too much 2011-01-02T05:19:29 delt0r___: sure you can 2011-01-02T05:19:30 hate the thing 2011-01-02T05:19:42 well find me a perl coder that does? 2011-01-02T05:19:43 I've worked with perl a bit too :P 2011-01-02T05:19:51 all the libs are awfull 2011-01-02T05:20:06 the comercial sector is moving away from perl now 2011-01-02T05:20:07 is there a python tutorial as entertaining as learn you a haskell? 2011-01-02T05:20:28 all programming tutorials should be of that nature 2011-01-02T05:20:39 bioinformatics (my field kinda) still beats that dead horse 2011-01-02T05:21:51 does that cover phenology at all? 2011-01-02T05:21:55 yes 2011-01-02T05:22:07 there are some maths/physics people at my uni who seem to be working on that kind of stuff 2011-01-02T05:22:23 I do population genetics... after doing a physics Masters 2011-01-02T05:22:30 yea 2011-01-02T05:22:57 is it worth learning lie algebras? 2011-01-02T05:23:14 Unless you do global warming, CERN or quantum computing... Physics funding is thin on the ground 2011-01-02T05:23:27 i haven't done any physics anyway 2011-01-02T05:23:40 my undergrad majors were cs, economics and maths, atm i think i'm still planning a phd in economics 2011-01-02T05:23:41 while bioinformatics can't give the money away fast enough 2011-01-02T05:23:50 although will probably have thesis using a fair bit of game theory 2011-01-02T05:24:41 i have at least a year before i need to make such a decision though 2011-01-02T05:24:52 i really should do some stuff to get ready for the maths gre this summer too 2011-01-02T05:25:18 apparently they had a lot more analysis than algebra this year :( 2011-01-02T05:25:29 last year* 2011-01-02T05:31:11 so what is a lot of units... 2011-01-02T05:31:35 was thinking of 8 fighters, 8 gun ships, 8 arty, 2 destroys 2011-01-02T05:31:57 how long is a piece of string 2011-01-02T05:32:09 well for "coordination" ? 2011-01-02T05:32:19 don't want to make it too hard for newbs 2011-01-02T05:38:17 how much do the maps vary? 2011-01-02T05:39:14 it needs to be hard enough that decent people will put a good effort in 2011-01-02T05:39:26 hard/interesting 2011-01-02T05:39:35 it sounds interesting enough 2011-01-02T05:49:24 it will be very hard to play well 2011-01-02T05:49:34 perhaps even humans will do well 2011-01-02T05:49:38 at least for a while 2011-01-02T05:49:57 The nebula affects speed and weapon range 2011-01-02T05:50:04 and sight distance 2011-01-02T05:50:17 so working with the map can make a massive difference 2011-01-02T05:50:55 but all dust is "traversalable" so you don't need a A* or anything to have a bot work 2011-01-02T05:51:12 as for the maps... not so random right now... 2011-01-02T06:01:24 *** delt0r_ has joined #aichallenge 2011-01-02T06:02:51 *** delt0r___ has quit IRC (Ping timeout: 255 seconds) 2011-01-02T06:08:17 delt0r_: but we will be able to have a random map generator yeah? 2011-01-02T06:08:27 yes 2011-01-02T06:08:32 But thats easy 2011-01-02T06:08:41 if you gave me an example map i could knock a map generator up for you 2011-01-02T06:11:28 haha, you love making map generators 2011-01-02T06:11:54 Its fine... 2011-01-02T06:12:14 I want them to be a bit structured... so i already have some code ready to go 2011-01-02T06:12:45 * antimatroid would prefer random maps for the actual contest 2011-01-02T06:13:20 and i think people want maps like they'll get at the end right near the start (from the outcry about the new map generator in planet wars) 2011-01-02T06:13:37 sigh: it's kind of fun writing them :P 2011-01-02T06:14:13 the programming is easy, it's a matter of working out how to do it, which is usually fun for anything 2011-01-02T06:14:35 if i spend considerable amount of time working out how to program what i want to do, then i get bored 2011-01-02T06:35:38 *** Palmik has quit IRC (Remote host closed the connection) 2011-01-02T06:35:58 *** Palmik has joined #aichallenge 2011-01-02T06:38:26 *** Frontier has joined #aichallenge 2011-01-02T06:51:27 *** Naktibalda has joined #aichallenge 2011-01-02T06:54:29 *** Frontier has left #aichallenge 2011-01-02T07:31:58 *** amstan has quit IRC (Ping timeout: 255 seconds) 2011-01-02T08:12:59 *** yasith has joined #aichallenge 2011-01-02T09:27:06 *** sigh has quit IRC (Remote host closed the connection) 2011-01-02T10:07:07 *** mega1 has quit IRC (Ping timeout: 240 seconds) 2011-01-02T10:19:48 antimatroid: haskell allows you to set operator precedence, btw 2011-01-02T10:21:52 and i think the difficulty of monads is overblown, and monads are given too much emphasis in the first place. the only thing about Monad that's special is that it has some optional syntax sugar 2011-01-02T10:34:50 *** mceier has joined #aichallenge 2011-01-02T10:55:59 jmcarthur: either way many people struggle with arrows and Monads. Many "practical" coding examples use them 2011-01-02T11:13:33 *** amriedle has joined #aichallenge 2011-01-02T11:18:20 *** yasith has quit IRC (Ping timeout: 265 seconds) 2011-01-02T11:29:08 *** BtbN has quit IRC (Quit: leaving) 2011-01-02T11:29:18 *** BtbN has joined #aichallenge 2011-01-02T11:30:12 *** yasith has joined #aichallenge 2011-01-02T11:46:38 *** boegel has joined #aichallenge 2011-01-02T12:02:41 delt0r_: i think they struggle with it because they expect to, mainly 2011-01-02T12:04:31 and as far as "practical" coding examples go, i assume you are talking about IO. as long as you can understand "do this, then do this," which should be easy coming from an imperative language, i don't see why that would be so difficult 2011-01-02T12:05:16 delt0r_: arrows are completely unnecessary. i've never seen coding example for beginners use Arrow 2011-01-02T12:10:14 do this, then do this is very unsatisfying if you really want to understand it. 2011-01-02T12:10:54 and folks that take haskell are typically not just there for the credits (mostly) 2011-01-02T12:42:45 delt0r_: i _think_ by "do this, then do this", jmcarthur was talking about sequencing operations, not cookbook programming. 2011-01-02T12:43:42 Oh 2011-01-02T12:43:44 right 2011-01-02T12:43:45 yea 2011-01-02T12:43:59 I pretend that there is state.... ;) 2011-01-02T12:44:19 yeah. me too. 2011-01-02T12:44:20 there *is* state 2011-01-02T12:44:32 IO is an imperative DSL 2011-01-02T12:44:47 yea-- but its hidden 2011-01-02T12:44:54 how is it hidden? 2011-01-02T12:46:16 there are still no "side effects" thats the point off monads (or so I thought...) Perhaps i need to go over it again... haven't done much haskell for quite a while 2011-01-02T12:46:43 this is just a product of giving monads too much special attention. they are not at all special 2011-01-02T12:46:54 and the only thing special about IO is that it's not implementable in pure haskell 2011-01-02T12:48:45 i have flirted with haskell several times, but have not written much of it. I really enjoy reading wadler's papers. 2011-01-02T12:49:10 there are some cool benefits to how haskell models IO, but you can treat IO as a normal imperative language 2011-01-02T12:50:13 Most of the stuff I looked at and played with only used IO for ... IO. I was more interested in lower level calculation and data structure type things. 2011-01-02T12:50:24 i daresay that it's just a sin bin. "i don't really know how to model this in pure code. whatever, i'll just throw it in IO" 2011-01-02T12:50:47 dmj111: as opposed to using IO for... not IO? ;) 2011-01-02T12:51:30 :) yep. instead of using it for imperative programming, using it to print stuff to the screen. 2011-01-02T12:51:44 that's mostly what it's used for in haskell anyway 2011-01-02T12:51:57 it's very rare to use it for imperative algorithms 2011-01-02T12:52:28 normally when you have an imperative algorithm to implement you would do it in ST because then you can make a pure function out of it and use it in pure code 2011-01-02T12:53:32 i used ST for arrays once, but don't really learn how to do things on the first try :) 2011-01-02T12:53:52 the hard thing about ST is its use of higher ranked types 2011-01-02T12:54:11 I really think Either is magic. Monad's might not be special, but Either looks cool every time I look at it. 2011-01-02T12:54:19 lol 2011-01-02T12:54:24 it's not special either 2011-01-02T12:54:53 no :) but I love how the idea simplifies some problems. 2011-01-02T12:55:01 that is the magic part, I guess. 2011-01-02T12:55:21 i find that Maybe gets used a lot more often 2011-01-02T12:55:46 Yeah, it seems a lot more "common case". It is magic too. 2011-01-02T12:55:49 :) 2011-01-02T12:55:56 things get "magical" when you start relating the type system to other areas of math 2011-01-02T12:56:19 that gets beyond my haskell expertise. 2011-01-02T12:57:14 Either a b => a+b; (a, b) => a*b; a->b => b^a 2011-01-02T12:57:37 if you think of types as sets then those formulas compute the number of elements in the set 2011-01-02T12:57:58 so if a has m members and b has n members, Either a b has a+b members 2011-01-02T12:58:13 and the familiar algebraic laws apply 2011-01-02T12:58:17 as far as equivalences 2011-01-02T12:59:04 I'm not sure I have read any of that before. pretty neat. 2011-01-02T12:59:12 for example, a*b+a*c ~= a*(b+c) 2011-01-02T12:59:31 meaning there is an equivalence relation between Either (a,b) (a,c) and (a, Either b c) 2011-01-02T13:00:13 that is, you can always write a function to switch between either representation 2011-01-02T13:00:22 currying is also expressible in this notation 2011-01-02T13:01:38 cool. learning haskell definitely feels more mathematical than other languages (whose names I won't mention to protect the presumed innocent) 2011-01-02T13:02:05 a->b->c ==> (c^b)^a ==> b^(a*b) ==> (a,b)->c 2011-01-02T13:02:17 i messed that up 2011-01-02T13:02:26 third step should be c^(a*b) 2011-01-02T13:02:58 that's one of the ways to think of it in math 2011-01-02T13:03:05 the other way is much more useful in my opinon 2011-01-02T13:03:08 *opinion 2011-01-02T13:03:36 types represent logic propositions, and values represent their constructive proofs 2011-01-02T13:04:12 that's the basis of theorem provers and programming languages that allow you to make compiler-checked proofs about your code 2011-01-02T13:04:28 that's a bit beyond what haskell is capable of though 2011-01-02T13:05:11 for a simple example, the type of function composition is modus ponens 2011-01-02T13:05:26 if A implies B and B implies C then A implies C 2011-01-02T13:05:49 (.) :: (b -> c) -> (a -> b) -> (a -> c) 2011-01-02T13:06:30 Either is logical or, (,) is logical and, (->) is logical implication 2011-01-02T13:08:32 maybe stuff like what i was just saying is what gives haskell it's overly academic/difficult reputation 2011-01-02T13:09:05 maybe :) but still, more interesting then much other stuff that is out there. 2011-01-02T13:15:23 *** antimatroid1 has joined #aichallenge 2011-01-02T13:17:06 *** antimatroid has quit IRC (Ping timeout: 240 seconds) 2011-01-02T13:25:17 *** yasith has quit IRC (Ping timeout: 272 seconds) 2011-01-02T13:30:00 *** Cold-Phoenix has joined #aichallenge 2011-01-02T14:03:27 *** delt0r___ has joined #aichallenge 2011-01-02T14:04:43 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2011-01-02T14:07:43 *** Utkarsh has quit IRC (Ping timeout: 272 seconds) 2011-01-02T14:12:00 *** Utkarsh has joined #aichallenge 2011-01-02T14:16:35 *** Utkarsh has quit IRC (Ping timeout: 272 seconds) 2011-01-02T14:32:45 *** dmj111 has quit IRC (Ping timeout: 240 seconds) 2011-01-02T14:39:33 http://mitpokerbots.com/competition 2011-01-02T14:39:41 looks awesome, but i think it's for mit students only 2011-01-02T14:45:42 i've thought that texas holdem would be a fun game to make AI for 2011-01-02T14:46:00 not the most exciting to watch though 2011-01-02T14:46:24 i like the idea of bluffing, in AI 2011-01-02T14:46:47 i've been trying to think of how one could introduce bluffing in a simpler game that's more fun to watch 2011-01-02T14:49:24 if you imagine that each bot is assigned two properties randomly chosen from a set of three possible, where one is visible to the other bot and the other isn't, and where the combination of the two properties affects your optimal strategy, then i can see bluffing becoming a part of a game 2011-01-02T14:49:48 the randomness isn't strictly necessary, but the assignment must be invisible to your opponent at least 2011-01-02T14:50:08 can't you bluff with normal poker? 2011-01-02T14:50:14 why go to texas holdem? 2011-01-02T14:50:25 yeah, but texas holdem draws it out more 2011-01-02T14:50:33 gives you more chances to change your mind 2011-01-02T14:50:40 (and lose more money) 2011-01-02T14:50:45 i'm pretty bummed that this contest isn't open 2011-01-02T14:50:58 what's "normal" poker anyway? ;) 2011-01-02T14:51:21 stud? two-card draw? 2011-01-02T14:52:20 meh, i haven't played poker in years :P 2011-01-02T14:52:34 and even then i would have been a kid, so probably wasn't doing it right 2011-01-02T14:52:56 ah 2011-01-02T14:53:21 texas holdem is really just the current hotness in the poker world 2011-01-02T14:53:29 but it also happens to be my favorite variant 2011-01-02T15:09:13 *** tapwater has joined #aichallenge 2011-01-02T15:18:33 *** Accoun has quit IRC () 2011-01-02T15:28:31 *** mega1 has joined #aichallenge 2011-01-02T15:31:06 *** Accoun has joined #aichallenge 2011-01-02T15:31:48 *** boegel has quit IRC (Quit: Leaving) 2011-01-02T15:41:37 *** dmj111 has joined #aichallenge 2011-01-02T15:56:22 *** Naktibalda has quit IRC (Remote host closed the connection) 2011-01-02T16:03:45 *** Migi32 has joined #aichallenge 2011-01-02T16:21:04 *** fmeyer has joined #aichallenge 2011-01-02T16:22:02 *** fmeyer has joined #aichallenge 2011-01-02T16:30:38 *** Utkarsh has joined #aichallenge 2011-01-02T17:12:20 you know what would be awesome? real time stratego 2011-01-02T17:13:08 that is, an rts where you know where all your opponents units are but not what kind they are until you fight them 2011-01-02T17:14:10 the flag thing might be tough to adapt to real time though 2011-01-02T17:14:14 flags and bombs can't move 2011-01-02T17:14:50 maybe it could be a "king" instead of a flag 2011-01-02T17:14:53 so it can move 2011-01-02T17:16:45 could be extra tough if not only can you not tell what kind of unit a particular unit is but there's also a fog of war 2011-01-02T17:18:10 because then even if you know what kind of unit it is, if it moves out of your vision you might not be able to figure it out the next time you see it 2011-01-02T17:20:32 hmm... might be nice to stick with flags and bombs actually since a stationary unit could be your opponent's flag, a bomb, *or* a moveable unit that's just pretending to be immovable 2011-01-02T17:21:43 *** choas has joined #aichallenge 2011-01-02T17:22:04 *** Palmik has quit IRC (Remote host closed the connection) 2011-01-02T17:25:59 *** choas has quit IRC (Client Quit) 2011-01-02T17:31:01 *** amstan has joined #aichallenge 2011-01-02T17:31:01 *** ChanServ sets mode: +o amstan 2011-01-02T17:36:42 amstan! real time stratego! it would be awesome, and you know it 2011-01-02T17:37:03 i have no idea about it yet lol 2011-01-02T17:37:17 let me unpack/repack and move first 2011-01-02T17:37:22 just came from florida 2011-01-02T17:37:26 bah 2011-01-02T17:37:43 i wasn't saying you must make a decision or something 2011-01-02T17:49:36 *** Accoun has quit IRC (*.net *.split) 2011-01-02T17:49:36 *** fmeyer has quit IRC (*.net *.split) 2011-01-02T17:49:36 *** Migi32 has quit IRC (*.net *.split) 2011-01-02T17:49:37 *** delt0r has quit IRC (*.net *.split) 2011-01-02T17:49:37 *** ChanServ has quit IRC (*.net *.split) 2011-01-02T17:49:37 *** greghaynes has quit IRC (*.net *.split) 2011-01-02T17:49:37 *** CIA-112 has quit IRC (*.net *.split) 2011-01-02T17:49:37 *** SmkMnstr has quit IRC (*.net *.split) 2011-01-02T17:49:37 *** acieroid has quit IRC (*.net *.split) 2011-01-02T17:49:38 *** pgpaskar_ has quit IRC (*.net *.split) 2011-01-02T17:49:38 *** AlliedEnvy has quit IRC (*.net *.split) 2011-01-02T17:49:39 *** wh1teside_ has quit IRC (*.net *.split) 2011-01-02T17:49:39 *** smellyhippy has quit IRC (*.net *.split) 2011-01-02T17:49:39 *** nann has quit IRC (*.net *.split) 2011-01-02T17:49:39 *** Zannick has quit IRC (*.net *.split) 2011-01-02T17:49:40 *** chl_ has quit IRC (*.net *.split) 2011-01-02T17:49:40 *** Utkarsh has quit IRC (*.net *.split) 2011-01-02T17:49:40 *** Sylph has quit IRC (*.net *.split) 2011-01-02T17:49:40 *** UndiFineD has quit IRC (*.net *.split) 2011-01-02T17:49:40 *** drodil has quit IRC (*.net *.split) 2011-01-02T17:49:41 *** spacebat has quit IRC (*.net *.split) 2011-01-02T17:49:41 *** mceier has quit IRC (*.net *.split) 2011-01-02T17:49:41 *** yellowseed has quit IRC (*.net *.split) 2011-01-02T17:49:41 *** jmpespxoreax has quit IRC (*.net *.split) 2011-01-02T17:49:41 *** davidd has quit IRC (*.net *.split) 2011-01-02T17:49:41 *** jbroman has quit IRC (*.net *.split) 2011-01-02T17:49:41 *** BtbN has quit IRC (*.net *.split) 2011-01-02T17:49:42 *** amstan has quit IRC (*.net *.split) 2011-01-02T17:49:42 *** dmj111 has quit IRC (*.net *.split) 2011-01-02T17:49:42 *** mega1 has quit IRC (*.net *.split) 2011-01-02T17:49:42 *** tapwater has quit IRC (*.net *.split) 2011-01-02T17:49:42 *** Cold-Phoenix has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** javagamer has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** jmcarthur has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** Cyndre has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** politik_ has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** amriedle has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** iFire has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** janzert has quit IRC (*.net *.split) 2011-01-02T17:49:43 *** delt0r___ has quit IRC (*.net *.split) 2011-01-02T17:49:44 *** antimatroid1 has quit IRC (*.net *.split) 2011-01-02T17:49:44 *** Error323 has quit IRC (*.net *.split) 2011-01-02T17:51:33 *** politik_ has joined #aichallenge 2011-01-02T17:51:33 *** Cyndre has joined #aichallenge 2011-01-02T17:51:33 *** jmcarthur has joined #aichallenge 2011-01-02T17:51:33 *** javagamer has joined #aichallenge 2011-01-02T17:51:33 *** Cold-Phoenix has joined #aichallenge 2011-01-02T17:51:33 *** tapwater has joined #aichallenge 2011-01-02T17:51:33 *** mega1 has joined #aichallenge 2011-01-02T17:51:33 *** dmj111 has joined #aichallenge 2011-01-02T17:51:33 *** Utkarsh has joined #aichallenge 2011-01-02T17:51:33 *** Migi32 has joined #aichallenge 2011-01-02T17:51:33 *** Accoun has joined #aichallenge 2011-01-02T17:51:33 *** delt0r___ has joined #aichallenge 2011-01-02T17:51:33 *** antimatroid1 has joined #aichallenge 2011-01-02T17:51:33 *** BtbN has joined #aichallenge 2011-01-02T17:51:33 *** amriedle has joined #aichallenge 2011-01-02T17:51:33 *** mceier has joined #aichallenge 2011-01-02T17:51:33 *** wh1teside_ has joined #aichallenge 2011-01-02T17:51:33 *** iFire has joined #aichallenge 2011-01-02T17:51:33 *** Sylph has joined #aichallenge 2011-01-02T17:51:33 *** yellowseed has joined #aichallenge 2011-01-02T17:51:33 *** smellyhippy has joined #aichallenge 2011-01-02T17:51:33 *** Error323 has joined #aichallenge 2011-01-02T17:51:33 *** UndiFineD has joined #aichallenge 2011-01-02T17:51:33 *** janzert has joined #aichallenge 2011-01-02T17:51:33 *** greghaynes has joined #aichallenge 2011-01-02T17:51:33 *** nann has joined #aichallenge 2011-01-02T17:51:33 *** AlliedEnvy has joined #aichallenge 2011-01-02T17:51:33 *** delt0r has joined #aichallenge 2011-01-02T17:51:33 *** jbroman has joined #aichallenge 2011-01-02T17:51:33 *** jmpespxoreax has joined #aichallenge 2011-01-02T17:51:33 *** CIA-112 has joined #aichallenge 2011-01-02T17:51:33 *** Zannick has joined #aichallenge 2011-01-02T17:51:33 *** pgpaskar_ has joined #aichallenge 2011-01-02T17:51:33 *** ChanServ has joined #aichallenge 2011-01-02T17:51:33 *** acieroid has joined #aichallenge 2011-01-02T17:51:33 *** SmkMnstr has joined #aichallenge 2011-01-02T17:51:33 *** chl_ has joined #aichallenge 2011-01-02T17:51:33 *** spacebat has joined #aichallenge 2011-01-02T17:51:33 *** drodil has joined #aichallenge 2011-01-02T17:51:33 *** davidd has joined #aichallenge 2011-01-02T17:51:33 *** gibson.freenode.net sets mode: +o ChanServ 2011-01-02T18:07:45 *** dmj111 has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** mega1 has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** tapwater has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** Cold-Phoenix has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** javagamer has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** jmcarthur has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** Cyndre has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** politik_ has quit IRC (*.net *.split) 2011-01-02T18:07:46 *** Accoun has quit IRC (*.net *.split) 2011-01-02T18:07:48 *** BtbN has quit IRC (*.net *.split) 2011-01-02T18:07:48 *** Migi32 has quit IRC (*.net *.split) 2011-01-02T18:07:49 *** delt0r has quit IRC (*.net *.split) 2011-01-02T18:07:49 *** ChanServ has quit IRC (*.net *.split) 2011-01-02T18:07:49 *** greghaynes has quit IRC (*.net *.split) 2011-01-02T18:07:50 *** CIA-112 has quit IRC (*.net *.split) 2011-01-02T18:07:50 *** SmkMnstr has quit IRC (*.net *.split) 2011-01-02T18:07:50 *** acieroid has quit IRC (*.net *.split) 2011-01-02T18:07:50 *** pgpaskar_ has quit IRC (*.net *.split) 2011-01-02T18:07:50 *** AlliedEnvy has quit IRC (*.net *.split) 2011-01-02T18:07:51 *** wh1teside_ has quit IRC (*.net *.split) 2011-01-02T18:07:51 *** smellyhippy has quit IRC (*.net *.split) 2011-01-02T18:07:52 *** nann has quit IRC (*.net *.split) 2011-01-02T18:07:52 *** Zannick has quit IRC (*.net *.split) 2011-01-02T18:07:52 *** chl_ has quit IRC (*.net *.split) 2011-01-02T18:07:52 *** Utkarsh has quit IRC (*.net *.split) 2011-01-02T18:07:52 *** Sylph has quit IRC (*.net *.split) 2011-01-02T18:07:52 *** UndiFineD has quit IRC (*.net *.split) 2011-01-02T18:07:53 *** drodil has quit IRC (*.net *.split) 2011-01-02T18:07:53 *** spacebat has quit IRC (*.net *.split) 2011-01-02T18:07:53 *** mceier has quit IRC (*.net *.split) 2011-01-02T18:07:53 *** yellowseed has quit IRC (*.net *.split) 2011-01-02T18:07:53 *** jmpespxoreax has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** davidd has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** jbroman has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** amriedle has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** iFire has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** janzert has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** delt0r___ has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** antimatroid1 has quit IRC (*.net *.split) 2011-01-02T18:07:54 *** Error323 has quit IRC (*.net *.split) 2011-01-02T18:08:28 *** BtbN has joined #aichallenge 2011-01-02T18:08:28 *** politik_ has joined #aichallenge 2011-01-02T18:08:28 *** Cyndre has joined #aichallenge 2011-01-02T18:08:28 *** jmcarthur has joined #aichallenge 2011-01-02T18:08:28 *** javagamer has joined #aichallenge 2011-01-02T18:08:28 *** Cold-Phoenix has joined #aichallenge 2011-01-02T18:08:28 *** tapwater has joined #aichallenge 2011-01-02T18:08:28 *** mega1 has joined #aichallenge 2011-01-02T18:08:28 *** dmj111 has joined #aichallenge 2011-01-02T18:08:28 *** Utkarsh has joined #aichallenge 2011-01-02T18:08:28 *** Migi32 has joined #aichallenge 2011-01-02T18:08:28 *** Accoun has joined #aichallenge 2011-01-02T18:08:28 *** delt0r___ has joined #aichallenge 2011-01-02T18:08:28 *** antimatroid1 has joined #aichallenge 2011-01-02T18:08:28 *** amriedle has joined #aichallenge 2011-01-02T18:08:28 *** mceier has joined #aichallenge 2011-01-02T18:08:28 *** wh1teside_ has joined #aichallenge 2011-01-02T18:08:28 *** iFire has joined #aichallenge 2011-01-02T18:08:28 *** Sylph has joined #aichallenge 2011-01-02T18:08:28 *** yellowseed has joined #aichallenge 2011-01-02T18:08:28 *** smellyhippy has joined #aichallenge 2011-01-02T18:08:28 *** Error323 has joined #aichallenge 2011-01-02T18:08:28 *** UndiFineD has joined #aichallenge 2011-01-02T18:08:28 *** janzert has joined #aichallenge 2011-01-02T18:08:28 *** greghaynes has joined #aichallenge 2011-01-02T18:08:28 *** nann has joined #aichallenge 2011-01-02T18:08:28 *** AlliedEnvy has joined #aichallenge 2011-01-02T18:08:28 *** delt0r has joined #aichallenge 2011-01-02T18:08:28 *** jbroman has joined #aichallenge 2011-01-02T18:08:28 *** jmpespxoreax has joined #aichallenge 2011-01-02T18:08:28 *** CIA-112 has joined #aichallenge 2011-01-02T18:08:28 *** Zannick has joined #aichallenge 2011-01-02T18:08:28 *** pgpaskar_ has joined #aichallenge 2011-01-02T18:08:28 *** ChanServ has joined #aichallenge 2011-01-02T18:08:28 *** acieroid has joined #aichallenge 2011-01-02T18:08:28 *** SmkMnstr has joined #aichallenge 2011-01-02T18:08:28 *** chl_ has joined #aichallenge 2011-01-02T18:08:28 *** spacebat has joined #aichallenge 2011-01-02T18:08:28 *** drodil has joined #aichallenge 2011-01-02T18:08:28 *** davidd has joined #aichallenge 2011-01-02T18:08:28 *** gibson.freenode.net sets mode: +o ChanServ 2011-01-02T18:08:31 *** mega1 has quit IRC (Ping timeout: 240 seconds) 2011-01-02T18:10:25 *** Cold-Phoenix has quit IRC () 2011-01-02T18:10:31 http://en.wikipedia.org/wiki/Dark_Chess 2011-01-02T18:22:00 *** amstan has joined #aichallenge 2011-01-02T18:22:00 *** ChanServ sets mode: +o amstan 2011-01-02T18:32:11 *** dmj111 has quit IRC (*.net *.split) 2011-01-02T18:32:11 *** tapwater has quit IRC (*.net *.split) 2011-01-02T18:32:11 *** javagamer has quit IRC (*.net *.split) 2011-01-02T18:32:11 *** jmcarthur has quit IRC (*.net *.split) 2011-01-02T18:32:12 *** Cyndre has quit IRC (*.net *.split) 2011-01-02T18:32:12 *** politik_ has quit IRC (*.net *.split) 2011-01-02T18:32:12 *** Accoun has quit IRC (*.net *.split) 2011-01-02T18:33:54 *** Accoun has joined #aichallenge 2011-01-02T18:33:54 *** dmj111 has joined #aichallenge 2011-01-02T18:33:54 *** tapwater has joined #aichallenge 2011-01-02T18:33:54 *** javagamer has joined #aichallenge 2011-01-02T18:33:54 *** jmcarthur has joined #aichallenge 2011-01-02T18:33:54 *** Cyndre has joined #aichallenge 2011-01-02T18:33:54 *** politik_ has joined #aichallenge 2011-01-02T18:49:43 *** Migi32 has quit IRC (Remote host closed the connection) 2011-01-02T18:52:14 *** nann has quit IRC (Ping timeout: 240 seconds) 2011-01-02T19:25:55 *** amstan_ has joined #aichallenge 2011-01-02T19:25:55 *** ChanServ sets mode: +o amstan_ 2011-01-02T19:28:23 *** amstan has quit IRC (Ping timeout: 240 seconds) 2011-01-02T19:56:28 i want to do the poker contest god damn you mit 2011-01-02T19:57:28 jmcarthur: dark chess looks cool, but I really like the simultaneous games 2011-01-02T20:06:35 yeah me too 2011-01-02T20:06:53 i kind of like my real time stratego idea 2011-01-02T20:07:39 i'm starting an engine. for now it's to try out my idea, if it gets anywhere, but even if it doesn't i might be able to use it to make another engine later 2011-01-02T20:13:14 *** mceier has quit IRC (Quit: leaving) 2011-01-02T20:16:32 delt0r is working on an rts like game atm 2011-01-02T20:16:43 which i think is nearly at a stage where he cane show us an example 2011-01-02T20:16:57 you have seenthe ants game in action yeah? 2011-01-02T20:18:53 yeah 2011-01-02T20:49:54 *** Apophis_ has joined #aichallenge 2011-01-02T20:49:54 *** sigh has joined #aichallenge 2011-01-02T21:06:33 jmcarthur: think there is any chance of getting the ratings changed to trueskill? 2011-01-02T21:13:01 *** amstan__ has joined #aichallenge 2011-01-02T21:13:01 *** ChanServ sets mode: +o amstan__ 2011-01-02T21:15:26 *** amstan_ has quit IRC (Ping timeout: 240 seconds) 2011-01-02T21:16:04 dmj111: what would be cool is to run the planet wars bots a few more times with different ranking methods /pairing methods 2011-01-02T21:18:43 *** amstan__ has quit IRC (Ping timeout: 265 seconds) 2011-01-02T21:20:33 dmj111: i don't really have control over the backend 2011-01-02T21:20:47 i too would like to see that chance, you've probably guessed 2011-01-02T21:20:50 *change 2011-01-02T21:21:37 antimatroid1: yeah, an objective experiment like that would probably be great 2011-01-02T21:24:04 *** bhasker has joined #aichallenge 2011-01-02T21:24:11 jmcarthur: yeah, I guess I didn't mean that you would decide that. It was kind of a random comment. 2011-01-02T21:24:27 I think there are several advantages to using trueskill for this. 2011-01-02T21:24:39 yeah 2011-01-02T21:25:00 BayesElo may be more "correct" given a set of games, but I think trueskill during the run-up would be more intuitive to everyone. 2011-01-02T21:25:29 and it can be updated one game at a time... 2011-01-02T21:25:36 trueskill also has a nice pairing algorithm, to boot 2011-01-02T21:25:42 and every game a bot plays is part of its rating. 2011-01-02T21:26:00 and your score won't change retroactively 2011-01-02T21:26:06 ^^ yes. 2011-01-02T21:27:19 I played with the pairing thing, but only with the match quality stuff. In my simplified version, a small number players ended up with small sigmas, and then were picked most of the time by the algorithm. 2011-01-02T21:28:28 after how many games? 2011-01-02T21:28:33 antimatroid1: yeah, I would kind of like to try that. I _think_ trueskill should give the same top 3, and I hope even more than that. 2011-01-02T21:28:48 I forget... and don't have the code handy, but it was thousands. 2011-01-02T21:29:02 that's actually a kind of good thing as much as it is a bad thing since it leads to other bots converging more quickly anyway 2011-01-02T21:29:04 ah 2011-01-02T21:29:10 then it's not going to happen anyway 2011-01-02T21:29:59 i was running rounds where each bot gets to pick an opponent. The opponents selection was a weighted score based on the match quality. 2011-01-02T21:30:26 ah 2011-01-02T21:30:34 i think there is probably a better way 2011-01-02T21:30:36 dmj111: i think people would cry foul if a significantly different ranking system was used pre contest 2011-01-02T21:30:38 i think so. 2011-01-02T21:30:51 the trueskill matching algorithm was designed with players that aren't playing *all the time* in mind 2011-01-02T21:30:53 antimatroid1: yeah, that is why I didn't suggest it for the final. 2011-01-02T21:30:59 jmcarthur: yep. 2011-01-02T21:31:25 so i think there should only be a pool of bots that are candidates to play more games at any given time 2011-01-02T21:31:51 it works out anyway since we want everybody to get a fair chance to play 2011-01-02T21:31:58 i think the system they used this time, with playing bots close to you more often, but other people some times, may still work with trueskilll. 2011-01-02T21:32:10 i thought the ranking went quite well for planet wars 2011-01-02T21:32:14 but i have nothing to really base that off 2011-01-02T21:32:18 yeah it was much better than tron 2011-01-02T21:32:18 other than intuition 2011-01-02T21:33:06 ah, let me be clear. I think it went well, and that bayeselo is a very good solution. but,. ... 2011-01-02T21:33:38 bayeselo is an *accurate* solution, but leads to confusing changes in rank at arbitrary times 2011-01-02T21:33:39 for intuition, and speed of updating, it could be improved. 2011-01-02T21:33:53 Yes. Its ok if you understand, but many people did not. 2011-01-02T21:34:18 even understanding it, i didn't really like it before the tournament 2011-01-02T21:34:26 and it leads to people going "WTF!!!" when someone wins 1 game against a top 20, and ends up with a huge elo score. 2011-01-02T21:34:42 well, that would sort of happen with trueskill as well 2011-01-02T21:35:18 dmj111: surely trueskill has to do that too? 2011-01-02T21:35:27 jmcarthur: the mean - 3 * sigma should help though, shouldn't it? 2011-01-02T21:35:32 oh that 2011-01-02T21:35:34 yes 2011-01-02T21:35:43 it would dampen weird effects like that significantly 2011-01-02T21:35:53 but that could have been applied to bayeselo scores too 2011-01-02T21:35:54 antimatroid1: this part should fix that ^^ 2011-01-02T21:36:21 what if there was a page explaining the rankigns better (that should be done regardless actually) 2011-01-02T21:36:33 jmcarthur: yes, we could just do that and it would help. the other things I don't like are the retroactive changes. 2011-01-02T21:36:53 antimatroid1: that would probably be a good idea. 2011-01-02T21:37:05 work needs to start on the new site soon 2011-01-02T21:37:08 antimatroid1: i think people would be much more willing to accept a score that changes incrementally over a score that jumps around, regardless of whether there is a good explanation for it available 2011-01-02T21:37:22 i'm quite interested in getting training grounds set up with old contest games (and possible a few other example games) 2011-01-02T21:37:38 it would end a lot of the "ELO is broken" arguments. Not all, of course, but a lot. 2011-01-02T21:37:47 *** bhasker has quit IRC (Ping timeout: 240 seconds) 2011-01-02T21:38:38 a nice property of the mean - 3 * sigma adjustment is that your bot would gradually rise in rank over time rather than jumping up and down and then converging. that would be much more intuitive to people. at the same time, even while your rank is very low it will attempt to match you with bots that it estimates will *eventually* be on the same level as you 2011-01-02T21:39:13 i don't like bots retaining their rank when being resubmitted 2011-01-02T21:39:29 they wouldn't, actually 2011-01-02T21:39:31 i think you should have to climb the ranks 2011-01-02T21:39:34 antimatroid1: hmm. I don't mean to do that. a submission still has its own history. 2011-01-02T21:39:39 their visible rank would drop 2011-01-02T21:39:43 yeah no, i'm referring to what was done with planet wars 2011-01-02T21:39:46 their hidden mean would be the same 2011-01-02T21:40:05 but with bayeselo, whenever opponents resubmitted, your old games against them were dropped from the ratings history. 2011-01-02T21:40:13 bots didn't retain their ranks in planet wars, did they? 2011-01-02T21:40:18 it was all reset at the end, but during the lead up, you kept your old rank when resubmitting, which was very easy to game and also left people out of their true position 2011-01-02T21:40:28 ? this is news to me 2011-01-02T21:40:39 Yeah, it was like antimatroid1 says 2011-01-02T21:40:49 must have happened after i dropped out 2011-01-02T21:41:06 it didn't make the final contest unfair in anyway, but i think some people did worse than they expected 2011-01-02T21:41:22 well, i disagree with that then 2011-01-02T21:41:23 particularly the people who uploaded numerous times a day 2011-01-02T21:41:28 but it wouldn't be that way with trueskill 2011-01-02T21:41:42 all you have to do is make sigma very large when you resubmit 2011-01-02T21:41:59 jmcarthur: that was a hack that I think the "true skill" method would fix. 2011-01-02T21:42:04 which would drop your visible score and pair you with a wider variety of opponents until you stabilize again 2011-01-02T21:42:05 i don't have a problem with someones rank being all over the show when they resubmit 2011-01-02T21:42:17 in fact i think it is kind of fun to watch it overshoot its mark 2011-01-02T21:42:36 it may be fun, but i do think it's less informative 2011-01-02T21:43:21 it's harder to see how quickly you're converging than it is to see how quickly you're rising 2011-01-02T21:43:34 i also like the idea of the mu - 3 * sigma, but create matches using your mu. The problem with the early matching methods is that it put you in the median of the rankings. With the default mu thing, a resubmit won't have to play in the "long tail", which is where the median ranking probably is. 2011-01-02T21:43:46 that was confusing. 2011-01-02T21:43:49 right 2011-01-02T21:43:55 besides, the main server during the lead up is more of a random pissing contest than anything 2011-01-02T21:44:12 if you want to get real testing done, you'll be on tcp, and it's intended for that to be made more official in the future 2011-01-02T21:44:37 awesome 2011-01-02T21:44:57 what would be best is if the official gui released was like jbotmanager 2011-01-02T21:44:59 i'm still disappointed that we never got daily/weekly replay dumps to download 2011-01-02T21:45:05 making it pretty damn easy for people to connect to tcp 2011-01-02T21:45:15 true, but the rankings confuse a lot of people. and, I think it led people to resubmit a lot, increasing the load on the servers for everyone else. 2011-01-02T21:45:29 dmj111: and lowering the accuracy even more 2011-01-02T21:45:34 yes. 2011-01-02T21:46:12 people will definitely be less inclined to resubmit if it makes their score drop instantly ;) 2011-01-02T21:46:29 *to resubmit often 2011-01-02T21:46:51 the problem is, resubmitting like that, even now, is the worst way to work out how good your bot is 2011-01-02T21:46:59 you are better off uploading it and leaving it a few days 2011-01-02T21:47:05 yeah 2011-01-02T21:47:08 perhaps we just need to communicate that to people better? 2011-01-02T21:47:27 we can communicate as much as we want, but sometimes there needs to be an incentive too. 2011-01-02T21:47:28 people will always go with their gut, despite any advice 2011-01-02T21:48:09 i just want a game with an active server to start making a bot again with :( 2011-01-02T21:48:13 i personally think it was a bad idea to prioritize new submissions in the last contest 2011-01-02T21:48:27 i dunno about that 2011-01-02T21:48:40 it incentivized resubmitting a lot 2011-01-02T21:48:46 for little gain 2011-01-02T21:49:01 maybe count number of games since first submission in the last 24 hours or something? 2011-01-02T21:49:06 want more replays? just resubmit! 2011-01-02T21:49:06 i think we need some unit tests on compile. I think that was in the forum too. 2011-01-02T21:49:11 but once you upload a new bot, it's nice to get some games right away 2011-01-02T21:49:41 yeah, it is nice, i agree, but you really shouldn't be uploading so often anyway 2011-01-02T21:49:57 if you need the instant feedback, that's what the tcp server is for 2011-01-02T21:50:04 if there are more people on tcp, there is less need for people to resubmit to the main server, both because you have more variety to play on tcp and more people aren't just testing on the main server 2011-01-02T21:50:18 IMO, the official server should be only for testing purposes until the tournament, to make sure your bot will run on it 2011-01-02T21:50:35 for getting lots of games tcp is superior anyway 2011-01-02T21:51:04 do you think it is likely that I could get a dump of game results from the finals to experiment with? i have a two player true skill program that I could try it out on. 2011-01-02T21:51:18 if the results are different, that would probably kill the idea. 2011-01-02T21:51:22 dmj111: i've been asking since planet wars started 2011-01-02T21:51:38 hm. ok :( 2011-01-02T21:51:42 dmj111: also, it's tough to test trueskill accurately in isolation from its matching algorithm 2011-01-02T21:52:06 dmj111: janzert is probably the best person to ask 2011-01-02T21:52:29 i think it could be used as an elo replacement, without the matching algorithm. just like glicko or something. 2011-01-02T21:52:32 without its matching algorithm, trueskill would almost certainly be worse than bayeselo 2011-01-02T21:53:25 i still think the best thing to do if we want to change ranking systems is to run a bunch of smaller tournaments 2011-01-02T21:53:31 do you guys have any annoyances with the engines from the last couple of contests that we should avoid this time around? 2011-01-02T21:53:45 i have two 2011-01-02T21:53:45 since i'm writing an engine now i figure i might as well try to improve on them 2011-01-02T21:54:08 jmcarthur: i don't understand it enough for that to make sense to me. I took it as two pieces. a rating algorithm, and a system for creating matchups to reduce sigmas. They were mostly independent though, I thought. 2011-01-02T21:54:12 1. send players information about the maps so people can't complain about hard coding them 2011-01-02T21:54:12 2. send players information when the game is finished ("finished\n" is sufficient) 2011-01-02T21:54:35 antimatroid1: oh, i mean the engines you use locally for testing 2011-01-02T21:55:03 what do you mean? 2011-01-02T21:55:08 on the other hand, I do think bayeselo will likely always be more accurate. 2011-01-02T21:55:16 dmj111: the matching algorithm attempts to pair people who would stand a better chance at lowering sigma 2011-01-02T21:55:27 i see. 2011-01-02T21:55:41 antimatroid1: like the crappy java things that get distributed with the starter packs 2011-01-02T21:55:52 anyway, i have to go now 2011-01-02T21:55:58 later. 2011-01-02T21:56:01 yeah, but that engine handles the format of input output with bots too 2011-01-02T21:56:04 anyway, later 2011-01-02T21:56:04 leave suggestions if any and i'll read later 2011-01-02T21:56:40 ie. bots should have been passed the number of planets at the start of a game in planet wars 2011-01-02T21:57:01 in ants you would get game board dimensions and base locations on the grid 2011-01-02T21:57:30 antimatroid1: i did find that kind of funny that 23 got hard coded in some programs. 2011-01-02T21:57:53 i guess i am just mean. 2011-01-02T22:03:33 *** delt0r_ has joined #aichallenge 2011-01-02T22:04:18 dmj111: anyone that did that screwed themselves over 2011-01-02T22:05:19 *** delt0r___ has quit IRC (Ping timeout: 272 seconds) 2011-01-02T22:07:27 *** amstan__ has joined #aichallenge 2011-01-02T22:07:27 *** ChanServ sets mode: +o amstan__ 2011-01-02T22:09:17 if someone is interested in trying out ranking algorithms I'd encourage you to look into whole history rating by Remi Coulom as well 2011-01-02T22:09:53 also the game *results* for planetwars are and have been available as pgn files from http://ai-contest.com/ranking-data/ 2011-01-02T22:10:23 the full game replay data is a different matter and rather harder to make easily available 2011-01-02T22:13:04 one note is that the pgn files above contain a lot of overlapping game data, in particular once the finals started they are almost completely cumulative 2011-01-02T22:15:04 janzert: thanks! 2011-01-02T22:15:34 played my first arimaa game an hour ago. not really sure what I was doing 2011-01-02T22:16:08 :) that's the way to start though 2011-01-02T22:16:55 the wiki book is pretty good if you want a resource on how basic strategy http://en.wikibooks.org/wiki/Arimaa 2011-01-02T22:17:13 i skimmed through part of it, but decided to take the plunge and play a game. 2011-01-02T22:17:28 thanks for the link, though, that is an impressive wikibook. 2011-01-02T22:17:55 arimaa looked kind of fun, but i much prefer the simultaneous games 2011-01-02T22:18:05 yeah, I think playing and then reading is really the best way 2011-01-02T22:18:09 i want a game where i can try something other than game tree searching 2011-01-02T22:18:32 antimatroid1: I agree Arimaa is probably not a good candidate for the ai contest 2011-01-02T22:18:37 too deep ;) 2011-01-02T22:18:54 exactly. this contest is not supposed to scare people away. 2011-01-02T22:19:01 -- from trying, that is. 2011-01-02T22:19:09 i think arimaa is a great game for an ai contest 2011-01-02T22:19:20 just not this one 2011-01-02T22:19:20 the deeper the game the better, provided people can get started easily 2011-01-02T22:19:27 *** amstan__ is now known as amstan 2011-01-02T22:19:30 janzert : that pgn file is what I wanted. 2011-01-02T22:19:35 yeah, i think these ones should stick to simultaneous games 2011-01-02T22:19:38 it doesn't have the visual fast pace that we really need 2011-01-02T22:20:01 and ease of people understanding what is going on just by looking at it 2011-01-02T22:20:04 i prefer the strategies involved and the games are easier to watch at a fast pace 2011-01-02T22:20:16 dmj111: great 2011-01-02T22:20:22 i'm quite interested in the game delt0r is working on 2011-01-02T22:21:06 i think the ants game would make for an interesting contest too though 2011-01-02T22:21:38 * janzert goes afk again, the sherbet is too much tempation 2011-01-02T22:22:32 janzert, when you get back ... 2011-01-02T22:22:32 i looked at the WHR stuff before... but recall not being able to find code. 2011-01-02T22:23:07 a quick google didn't reveal anything. his paper looks interesting, but a bit much for me to code for fun. 2011-01-02T22:55:37 *** dmj111 has quit IRC (Remote host closed the connection) 2011-01-02T23:03:45 *** dmj111 has joined #aichallenge 2011-01-02T23:03:51 antimatroid1: i wasn't talking about map format. i meant issues with the engine itself, like crappy performance, bad things about the interface, features you wished it had, etc. 2011-01-02T23:05:30 in my own engine, I added the ability to dump the output to a named file, and to output a "score" as the game ran. 2011-01-02T23:05:53 i had a separate engine to run tournaments between two bots over a map directory. 2011-01-02T23:06:10 jbotmanager was much nicer, i am sure. 2011-01-02T23:06:50 that's the kind of stuff i mean, yeah 2011-01-02T23:07:12 with tron we could use cerr to output to the terminal, but not in planet wars 2011-01-02T23:07:17 i'm definitely doing the output stuff 2011-01-02T23:07:18 it could potentially be nice to be able to do that again 2011-01-02T23:07:28 those are the only "features" I added to my test code. 2011-01-02T23:07:46 i ended up just doing all my debugging to a file though 2011-01-02T23:07:58 mine captures stderr from both bots and i plan to allow you to say where you want each error stream to go 2011-01-02T23:08:13 i don't think i'll use that again anyway 2011-01-02T23:08:26 i much prefer just opening up a file stream and throwing everything at that 2011-01-02T23:09:19 the ultimate benefit here is that i plan to have it run little tournaments between different version of your bot on different maps and it can keep all the data relating to each game grouped together or something 2011-01-02T23:09:42 i wouldn't do that in the engine 2011-01-02T23:09:45 leave that for the gui 2011-01-02T23:09:58 this engine would also spit out replays with that data 2011-01-02T23:10:29 so it would be the map, error logs from each bot, and game replay 2011-01-02T23:11:54 antimatroid1: the engine is the only place where it makes sense to decide where to store the log files if you want it to be consistent between bots 2011-01-02T23:12:14 well, tournament manager maybe, if it's separate 2011-01-02T23:12:53 to me the gui should pretty much just take replay files and render them for you and that's it 2011-01-02T23:13:16 did you use jbotmanager at all? 2011-01-02T23:13:19 delt0rs child 2011-01-02T23:13:21 it was awesome 2011-01-02T23:13:42 no i didn't 2011-01-02T23:13:50 i think that came about after i stopped competing too 2011-01-02T23:14:54 it enabled you to load up bots from a location, play them against each other (don't think it enabled you to run tournaments between your bots, but surely wouldn't be hard), play on tcp, and play one of them yourself in human mode, you could also save the games to file etc. 2011-01-02T23:15:35 sounds like the python engine from tron 2011-01-02T23:15:44 i loved that engine 2011-01-02T23:16:02 nneonneo made that 2011-01-02T23:16:04 i don't remember an engine like that from tron 2011-01-02T23:16:11 amstan: jbotmanager? 2011-01-02T23:16:17 no.. the tron one 2011-01-02T23:16:24 ah okay, cool 2011-01-02T23:16:37 amstan: how's the moving going? / hurry up :P 2011-01-02T23:16:48 the python one didn't have actual integration with the tcp stuff i guess 2011-01-02T23:16:48 tomorrow 2011-01-02T23:17:02 but it worked very well with the standalone tcp client 2011-01-02T23:17:22 before jbotmanager i was just running tcp games from the command line 2011-01-02T23:17:22 * amstan signs off 2011-01-02T23:17:31 and jbotmanager let you watch tcp games live 2011-01-02T23:17:37 rather than waiting for them to appear on the site 2011-01-02T23:18:23 ran a true skill experiment on the pgn game data... here is the top 100: http://pastebin.com/4a2h7WqD 2011-01-02T23:18:46 well i didn't change place 2011-01-02T23:19:10 why are mu and sigma the same? 2011-01-02T23:19:21 and what is "rating"? 2011-01-02T23:19:22 hmm. my bad. 2011-01-02T23:19:35 FlagCapper went up 10 places, interesting 2011-01-02T23:19:51 is rating the lower confidence bound? 2011-01-02T23:20:15 jmcarthur: yes on the rating. 2011-01-02T23:20:33 the mu and sigma thing was a fat finger on my part making the table. fixing now... 2011-01-02T23:21:05 there are a few people who moved, i'm curious as to what the different is there? 2011-01-02T23:21:10 most seems to have stayed the same 2011-01-02T23:21:12 i actually think those ratings look pretty decent for not even being used with the matching algorithm 2011-01-02T23:21:46 i think the matching algorithm will just reduce the sigmas overall, but besides that, with this much data, not change much else. 2011-01-02T23:21:46 antimatroid1: the main difference is that with trueskill the order of the games matters 2011-01-02T23:22:02 woah, isn't that bad? 2011-01-02T23:22:04 dmj111: reducing the sigmas would also raise the final scores 2011-01-02T23:22:05 any interest in more rankings than 100? 2011-01-02T23:22:15 antimatroid1: it is both good and bad 2011-01-02T23:22:33 i think ones ranking should be independent of the order in which they win their games 2011-01-02T23:22:44 and that was not the case for this most recent contest 2011-01-02T23:23:05 antimatroid1: bayeselo doesn't consider the order of the games, I don't think. 2011-01-02T23:23:26 dmj111: no, but the pairing algorithm skewed it so that if you lost early, you had a harder time climbing up the ranks 2011-01-02T23:23:44 i think it was worth having the change to pairing, but it certainly removed the independence between ranking and order of games 2011-01-02T23:23:54 antimatroid1: it's bad because it's nondeterministic in the ordering. it's good because it means you don't jump around so much. it also doesn't matter so much when paired with the good matchup algorithm 2011-01-02T23:24:13 how does the matchup algorithm work? 2011-01-02T23:24:44 i'm almost preferring my mergesort ranking idea again 2011-01-02T23:24:47 http://pastebin.com/FQPVsC0Z 2011-01-02T23:25:07 antimatroid1: it matches players randomly, weighted by the probability that the matchup will lead to a draw 2011-01-02T23:25:08 continue splitting the leaderboard up into sections after rounds, then resetting rankings inside those blocks to continue running bots that are closer in rank 2011-01-02T23:25:35 giving more games to the bots in the higher ranked blocks 2011-01-02T23:25:37 antimatroid1: that is a different problem then this one. even given the exact same results, but in different order, trueskill can give different answers. bayeselo won't, I believe. 2011-01-02T23:25:42 antimatroid1: it takes certainty of score into account in that probability, too 2011-01-02T23:26:01 right, bayeselo will always give the same results regardless of what order the games are played in 2011-01-02T23:26:05 dmj111: yes, different problem but still removes independence between order of games and ranking in the final contest 2011-01-02T23:26:12 which i think is an indeal criteria 2011-01-02T23:26:12 but *our* use of bayeselo *does* depend on order 2011-01-02T23:26:29 because the intermediate scores affect our matchup algorithm 2011-01-02T23:26:58 ideal* 2011-01-02T23:28:16 trueskill and the trueskill matching algorithm is probably superior for real time leaderboards 2011-01-02T23:28:26 not so for tournaments, most likely 2011-01-02T23:28:28 the only real problem i have with my *mergesort* ranking thing is that you then don't have final rankins (other than vague regions) determined from bots outside that block of rankings 2011-01-02T23:29:19 your mergesort idea sounds like a pretty standard tournament bracket to me 2011-01-02T23:29:24 another option is to have rounds (somewhat like codejam) and just reset the leaderboard each time for the upper half, freezing the rankings of bots that just got thrown out 2011-01-02T23:29:36 there are well-studied tournament algorithms out there 2011-01-02T23:30:00 the continously running swiss tournament was thrown around for a while, but I worry that people will play the same people too much that way. 2011-01-02T23:30:02 it's not meant to be anything new, i'm curious as to whether that would work better though? 2011-01-02T23:30:24 final blocks could consist of 100 or so bots 2011-01-02T23:30:35 you aren't going to run too many games amoungst them in a day 2011-01-02T23:31:21 elimination tournaments are probably not ideal, as i understand it, but i'm no expert on tournaments 2011-01-02T23:32:02 i personally lean toward the swiss tournament, based on my limited knowledge 2011-01-02T23:32:36 dmj111: ah, those sigmas are much better :) 2011-01-02T23:32:58 the biggest flaw I found with the swiss idea, is that it would conflict with the whole median matchup problem. 2011-01-02T23:33:01 yeah, much better. 2011-01-02T23:33:36 dmj111: the best thing about your results is that, at least in the top bots, the confidence bounds don't seem to overlap at all 2011-01-02T23:33:51 the trueskill algorithm will limit the sigma to some minimum value, but I don't know how to calculate it (yet). 2011-01-02T23:33:52 2011-01-02T23:34:00 a little further down the do overlap 2011-01-02T23:34:03 *they do 2011-01-02T23:34:24 that's an optional part of trueskill 2011-01-02T23:34:40 bocsimacko kicked everyones ass so hard 2011-01-02T23:34:46 honestly i think sigma should only be artificially increased after a resubmission 2011-01-02T23:34:53 yeah... at first glance, I think these results are fine. Further down the leaderboard, there must be groupings that could be reordered without qualms. 2011-01-02T23:35:57 jmcarthur: i think trueskill has a sqrt(sigma1 **2 + sigma2**2 + beta**2), factor in there somewhere, so the smallest possible sigma is beta. (or something like that...) 2011-01-02T23:36:27 dmj111: i'm pretty sure it has an additional step for raising sigma between games 2011-01-02T23:37:04 that is the idea, but it doesn't have a separate step to do that. 2011-01-02T23:37:10 it is part of the math. 2011-01-02T23:37:18 ah 2011-01-02T23:37:21 looking at it now to try to figure it out. 2011-01-02T23:38:26 here is the code https://bitbucket.org/dmj111/planetwars_test_tools/src/38084fa3299a/stuff/ 2011-01-02T23:38:58 its ugly... 2011-01-02T23:41:12 ok, I cannot back up my claim. I thought I read that in one of the papers, but I don't see it yet. 2011-01-02T23:41:57 i personally don't like the idea of artificially affecting rankings just so they appear more stable 2011-01-02T23:42:17 and the only important time for rankings to be as accurate as possible is for the final contest 2011-01-02T23:43:22 okay, my interpretation of what i read could be wrong too, though 2011-01-02T23:43:26 "Before starting to determine the new skill beliefs of all participating players for a new game outcome, the TrueSkill ranking system assumes that the skill of each player may have changed slightly between the current and the last game played by each player. The mathematical consequence of making such an assumption is that the skill uncertainty σ will be slightly increased, the amount of which 2011-01-02T23:43:28 is, in principle, a configurable parameter of the TrueSkill ranking system. It is this parameter that both allows the TrueSkill system to track skill improvements of gamers over time and ensures that the skill uncertainty σ never decreases to zero ("maintaining momentum")." 2011-01-02T23:43:43 antimatroid1: there's nothing artificial about it 2011-01-02T23:43:45 yeah, maybe I am confusing this with glicko. 2011-01-02T23:44:15 but, I did not change sigmas besides using the update equation in that run I did. 2011-01-02T23:44:16 antimatroid1: it's only more stable by virtue of reporting more conservative results