Currently Online

Latest Posts

Topic: Measuring working time of productionsites

hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2019-05-20, 22:48

Tibor wrote:

einstein13 wrote:

Tibor wrote:

total_duration = 5 * 60 * 1000 seconds

I guess that it is 560 seconds or 560*1000 miliseconds, but that is minor thing (rather a bit funny to me face-smile.png )

No it is 5 minutes, but value is arbitrary...

curent_result_weight (...) If "something" took few seconds it will have lesser impact and so on...

I don't understand everything (total duration is a constant here?), but as I know you, I believe that this formulas are good. So I will have no objections for your work any more! face-smile.png Go for it and merge to make the AI even better!

What I dont like (too) about current main statistics, that it covers varying range of time. From 20x skipped to 20x producing, this can be from 1 minute to more then 20 minutes...

the current statistics for display in the UI does not count skipped cycles it simply ignores them. Only thing that got updated was the AI crude_percent value with a scaling factor of 98%. only failing and completing is added in the register of the last 20 program cycles. As explained in the branch review I think a task based productivity count like it is implemented now fits much better the informational need of the player as well as of the AI. For me the question is not how long has the building been working in the last 20 minutes. The question is rather how often has it been succesful in the last 20 attempts. A nice side effect of the current implementation is that the falling time is much shorter due to failing is much shorter than completing a task. So we really get a quick alert if a building has a problem.


Top Quote
Tibor

Topic Opener
Joined: 2009-03-23, 23:24
Posts: 1377
Ranking
One Elder of Players
Location: Slovakia
Posted at: 2019-05-20, 23:08

OK - next argument:

We have a question - "how much can we increase the production if we supply the site properly?". Official statistics can be misleading here. E.g. 35 % indicated that we can increase output by 200 %. But actually it can be 20%


Top Quote
einstein13
Avatar
Joined: 2013-07-29, 00:01
Posts: 1118
Ranking
One Elder of Players
Location: Poland
Posted at: 2019-05-21, 15:33

einstein13 wrote:

Tibor wrote:

total_duration = 5 * 60 * 1000 seconds

I guess that it is 560 seconds or 560*1000 miliseconds, but that is minor thing (rather a bit funny to me face-smile.png )

sorry, the markdown compiler skipped "*" between 5 and 60. It should be:

I guess that it is 5 * 60 seconds or 5 * 60 * 1000 miliseconds, but that is minor thing (rather a bit funny to me face-smile.png )


einstein13
calculations & maps packages: http://wuatek.no-ip.org/~rak/widelands/
backup website files: http://kartezjusz.ddns.net/upload/widelands/

Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2019-05-21, 17:42

Tibor wrote:

OK - next argument:

We have a question - "how much can we increase the production if we supply the site properly?". Official statistics can be misleading here. E.g. 35 % indicated that we can increase output by 200 %. But actually it can be 20%

For this question the attempt based statistics is not very well suited. Correct but the current crude_statistics isn't as well. As isn't your new formula. For this question we would need to store the times and the results and the produced wares / attempted wares of the attempts of the last x minutes.
However this is a different question to the question you wanted to solve according to the discussion on Launchpad "The question to answer here is - do we need another building of the type"


Top Quote
Tibor

Topic Opener
Joined: 2009-03-23, 23:24
Posts: 1377
Ranking
One Elder of Players
Location: Slovakia
Posted at: 2019-05-21, 18:07

My formula is only approximate. I call it 'weighted' - older attempts have lesser weight then latest ones... But we do no need mathematical accuracy, just something more accurate than official statistics...


Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2019-05-21, 18:41

Tibor wrote:

My formula is only approximate. I call it 'weighted' - older attempts have lesser weight then latest ones... But we do no need mathematical accuracy, just something more accurate than official statistics...

I tried to implement all formulas in excel and it does show some weird results maybe I made a mistake but it seems that in your formula there might be still an error.
I think I'll create a bug for this issue to be able to upload my excel. Here it is https://bugs.launchpad.net/widelands/+bug/1829898

Edited: 2019-05-21, 18:46

Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2019-05-21, 20:03

hessenfarmer wrote:

I tried to implement all formulas in excel and it does show some weird results maybe I made a mistake but it seems that in your formula there might be still an error.

Sorry I made a mistake now it seems to be correct. Have to rethink and experiment a bit. Maybe I change my mind and we should have this after all.

New Version of the table now uploaded to https://bugs.launchpad.net/widelands/+bug/1829898


Top Quote
Tibor

Topic Opener
Joined: 2009-03-23, 23:24
Posts: 1377
Ranking
One Elder of Players
Location: Slovakia
Posted at: 2019-05-21, 20:22

As I posted there, I like the result of my formula in your table, thanks for your effort


Top Quote
Tibor

Topic Opener
Joined: 2009-03-23, 23:24
Posts: 1377
Ranking
One Elder of Players
Location: Slovakia
Posted at: 2019-05-21, 20:27

Also, as I said few times, there will be no change for players... this statistics is only for AI use...


Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2019-05-21, 20:30

ok I changed my mind about this just neede to see the excel to get the full concept. I now see that the new formula is assymptotically goint to the real relation between working time of a succesful cycle to the time to supply the building with all wares. And this is regardless of time to fail. And this is the major point. UI would be only correct if fail time = succestime. As we have a lot of different fail times now and even very small ones, we might try this also for the UI stats as the values calculated here show better how much more wares are needed per time to suplly this building fully. We just need to think about how to calculate the trend then.

Just forgot to say sorry for the trouble based on my ignorance

Edited: 2019-05-21, 20:31

Top Quote