Re: [Freeciv-Dev] Raw Data from The Battle of 100 Freeciv AIs
Thanks for this! I'll let you know when I get something done :) Mhairi On 3 May 2015 at 20:50, Marko Lindqvist cazf...@gmail.com wrote: On 3 May 2015 at 22:07, Andreas Røsdal andreas.ros...@gmail.com wrote: - With 256 players on large maps, then the client connection to the server is cut very often, due to a buffer overflow error message. By 256 players you mean 246 players + 10 slots reserved for barbarians, or? Was it enough to define max number of players from 160 to 256, or was more changes needed? IIRC from the time I increased max to 160, 256 (or -1 as signed) is a special value so one could have only 256-1x16 = 240 as the max. - ML ___ Freeciv-dev mailing list Freeciv-dev@gna.org https://mail.gna.org/listinfo/freeciv-dev
Re: [Freeciv-Dev] Raw Data from The Battle of 100 Freeciv AIs
Le dimanche 3 mai 2015, 21:07:33 Andreas Røsdal a écrit : I enjoyed your blog about the 100 Freeciv AIs - really cool stuff! Thanks! Would it be possible for you to send me the data you used to make the awesome line graphs? If I managed to get any nice visualisations done I'd send it your way and you could put it on the blog. Freeciv scorelog file: http://play.freeciv.org/blog/battle-of-100-freeciv-ais/data/freeciv-score.lo g.gz Python script to parse scorelog files (modified), run like this python freeciv-score-log.py --size=3200x1800 freeciv-score.log.gz http://play.freeciv.org/blog/battle-of-100-freeciv-ais/data/freeciv-score-lo g.py Hello, I modified the script to output CSV data instead of graphs, to make analysis easier. The modified version will be available at [1] until I run out of space. I made a spreadsheet to have a look at the BNP. It looks like the proportion of weak AIs (less than half the mean BNP) increases over time, but rather slowly. This spreadsheet, the extracted CSVs and graphs are available at [2]. Louis [1] https://webcloud6.zaclys.com/index.php/s/wL0Mcks5nSqkfF4 [2] https://webcloud6.zaclys.com/index.php/s/qxi4QQmGSWGvZtS ___ Freeciv-dev mailing list Freeciv-dev@gna.org https://mail.gna.org/listinfo/freeciv-dev
Re: [Freeciv-Dev] Raw Data from The Battle of 100 Freeciv AIs
On 3 May 2015 at 22:07, Andreas Røsdal andreas.ros...@gmail.com wrote: - With 256 players on large maps, then the client connection to the server is cut very often, due to a buffer overflow error message. By 256 players you mean 246 players + 10 slots reserved for barbarians, or? Was it enough to define max number of players from 160 to 256, or was more changes needed? IIRC from the time I increased max to 160, 256 (or -1 as signed) is a special value so one could have only 256-1x16 = 240 as the max. - ML ___ Freeciv-dev mailing list Freeciv-dev@gna.org https://mail.gna.org/listinfo/freeciv-dev
Re: [Freeciv-Dev] Raw Data from The Battle of 100 Freeciv AIs
On 29 April 2015 at 14:13, Mhairi McNeill mhairi.mcne...@deltadna.com wrote: Hi Freeciv Devs, I enjoyed your blog about the 100 Freeciv AIs - really cool stuff! I'm a data scientist who's really interested in visualisation. Currently I'm learning D3 to do some interactive visualisations. Would it be possible for you to send me the data you used to make the awesome line graphs? If I managed to get any nice visualisations done I'd send it your way and you could put it on the blog. All the best, Mhairi That's about Andreas' freeciv-web blog entry (I'm not sure if it was mentioned in freeciv-dev before, though it made it to Slashdot): http://play.freeciv.org/blog/2015/04/the-battle-of-100-freeciv-ais/ What bothers me in that experiment is that the freeciv revision has not been specified (just mention of post-2.5 svn, I assume that also post-S2_6 trunk has been used). One would want to be able to reproduce that game (so exact configurations would be needed, but also knowledge of freeciv revision to run it in). Another thing about using unstable development code is that there's really no telling what bugs AI has (was it behaving at all sanely - I've got AI stuck to some quite braindead situations while running autogames with recent code). As far as I know the blogged game did not have anything special compared to all the all-AI-autogames we constantly run - setup should be trivial. If one wants to do any kind of research related to this, it's an option (that either makes sense, or not, depending on what you want) to run another game like that instead of using the data from that exact game. - ML ___ Freeciv-dev mailing list Freeciv-dev@gna.org https://mail.gna.org/listinfo/freeciv-dev
[Freeciv-Dev] Raw Data from The Battle of 100 Freeciv AIs
Hi Freeciv Devs, I enjoyed your blog about the 100 Freeciv AIs - really cool stuff! I'm a data scientist who's really interested in visualisation. Currently I'm learning D3 to do some interactive visualisations. Would it be possible for you to send me the data you used to make the awesome line graphs? If I managed to get any nice visualisations done I'd send it your way and you could put it on the blog. All the best, Mhairi ___ Freeciv-dev mailing list Freeciv-dev@gna.org https://mail.gna.org/listinfo/freeciv-dev
Re: [Freeciv-Dev] Raw Data from The Battle of 100 Freeciv AIs
I enjoyed your blog about the 100 Freeciv AIs - really cool stuff! Thanks! Would it be possible for you to send me the data you used to make the awesome line graphs? If I managed to get any nice visualisations done I'd send it your way and you could put it on the blog. Freeciv scorelog file: http://play.freeciv.org/blog/battle-of-100-freeciv-ais/data/freeciv-score.log.gz Python script to parse scorelog files (modified), run like this python freeciv-score-log.py --size=3200x1800 freeciv-score.log.gz http://play.freeciv.org/blog/battle-of-100-freeciv-ais/data/freeciv-score-log.py Savegame at last turn: http://play.freeciv.org/blog/battle-of-100-freeciv-ais/data/savegame.sav.bz2 Server settings: http://play.freeciv.org/blog/battle-of-100-freeciv-ais/data/biggame.serv Requires Freeciv from subversion at date 25.04.2015. Here are some random things I discovered: - Graphs from freeciv-score-log.py don't show which player has which color. - freeciv-score-log.py extended to support more than 16 players. - The Freeciv server takes multiple hours to complete each turn with 100 players on large maps. - With 256 players on large maps, then the client connection to the server is cut very often, due to a buffer overflow error message. Regards, Andreas R. ___ Freeciv-dev mailing list Freeciv-dev@gna.org https://mail.gna.org/listinfo/freeciv-dev
