On 23.09.2012 03:28, Michael T. Pope wrote:
> On Sat, 22 Sep 2012 07:59:27 PM Michael T. Pope wrote:
>    
>> If it stands up there there will be greater cause for optimism.
>>      
> OK, we have greater cause for optimism.  Here are the results for the last two
> overnight runs.  What is being measured is mostly AI mission completions, with
> the addition of colony number and average colony size.  I do 10 runs with 7
> active AIs playing to 1700 and average the results, using a small random map
> with Very Hard settings (small-VH) and a Caribbean map with normal settings
> (debug-N), using the freecol ruleset in both cases.  The names in parentheses
> are the applied patch sets, with "st" being the starvation patch.  iter2 was
> committed yesterday so it is present in the latter runs, animate is a logging
> boost to try to catch a Heisenbug in animation and headless is the patch to
> make headless mode work, both of which should be irrelevant to production.
>
> 20120922-small-7-VH x 10 (animate,headless,iter2)
> Builds         24.800 ~   3.250         Cashins         0.000 ~   0.000
> Colony#        22.300 ~   3.407         ColonySize      3.429 ~   0.375
> Defences      230.200 ~   9.600         Missions        1.200 ~   0.748
> Native Demands 26.400 ~  11.473         Native Gifts  493.200 ~  94.330
> Pioneerings    22.300 ~   4.406         Piracies       19.300 ~   8.439
> Scoutings      66.900 ~  18.807         Seek+Destroys 201.800 ~  45.299
> Transports    113.400 ~  17.200         Wishes         33.700 ~   3.874
>
> 20120922-debug-N x 10 (animate,headless,iter2)
> Builds         23.500 ~   2.377         Cashins         0.000 ~   0.000
> Colony#        22.400 ~   1.855         ColonySize      3.330 ~   0.274
> Defences      139.400 ~  22.155         Missions        2.400 ~   0.490
> Native Demands  8.000 ~   2.828         Native Gifts  182.100 ~  49.780
> Pioneerings    26.400 ~   5.044         Piracies       87.800 ~  13.898
> Scoutings      22.300 ~   6.694         Seek+Destroys 104.300 ~  22.334
> Transports    106.500 ~  20.319         Wishes         27.100 ~   5.262
>
> 20120922-small-7-VH x 10 (animate,headless,st)
> Builds         29.200 ~  10.303         Cashins         0.000 ~   0.000
> Colony#        24.600 ~   6.931         ColonySize      3.267 ~   0.310
> Defences      238.600 ~  18.938         Missions        1.500 ~   1.118
> Native Demands 23.800 ~  10.778         Native Gifts  446.400 ~  74.247
> Pioneerings    26.300 ~   3.035         Piracies       11.300 ~  10.383
> Scoutings      74.700 ~  16.565         Seek+Destroys 217.900 ~  46.664
> Transports     98.800 ~  17.411         Wishes         34.800 ~   9.097
>
> 20120923-debug-N x 10 (animate,headless,st)
> Builds         23.200 ~   3.280         Cashins         0.000 ~   0.000
> Colony#        22.100 ~   3.448         ColonySize      3.889 ~   0.432
> Defences      132.600 ~  22.668         Missions        2.700 ~   0.900
> Native Demands  8.300 ~   3.716         Native Gifts  177.300 ~  54.798
> Pioneerings    26.100 ~   4.784         Piracies       94.800 ~  22.063
> Scoutings      23.300 ~   5.967         Seek+Destroys 106.800 ~  18.027
> Transports    118.000 ~  22.401         Wishes         30.300 ~   6.084
>
>
> I am not seeing any difference big enough to suspect any adverse effect of the
> starvation patch.  Indeed the variation of the numbers are within what I
> normally see from day to day.  Now this is not an ideal test in that it was
> designed to spot AI regressions and is not directly measuring production, but
> I would expect to see a secondary effect on the colony number/size and
> Wishes/Transports if production was significantly different.
>
> The other interesting result is that I added code to ProductionTree.remove
> that logs all instances of negative production.  In all the runs, this only
> happened for food and only for the "(goods.getType() == root.getType())"
> case.
>    

This is expected, since food is the only production that needs a tree, 
i.e. the only goods type that can be produced by two different raw 
materials (grain and fish).


> Of course we still should be wary about what we conclude here, as the AI does
> try to be productive:-), and has a narrower range of behaviour than a typical
> human player.  However I think we at least have evidence that the patch does
> not cause immediate showstoppers, so I am lead to recommend committing it and
> let us see what humans can find.
>
> Cheers,
> Mike Pope
>    

Good. I will commit it.


Regards

Michael

------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://ad.doubleclick.net/clk;258768047;13503038;j?
http://info.appdynamics.com/FreeJavaPerformanceDownload.html
_______________________________________________
Freecol-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freecol-developers

Reply via email to