Topic: productivity percentage of buildings is wrong

The code for this has not been touched in the four revisions. Have you had a closer look at the issue or is it just a "feeling", something you noticed while playing? What happens when you try to reproduce my experiment with the mill?

"Only few people know how much one has to know in order to know how little one knows." - Werner Heisenberg

Top

Quote

i am doing the same thing, running empty mines, with barbarians now, and they work at 5% on average, unlike the atlanteans ones. so it works differently for barbarians than for atlanteans.

actually, the average over 90 mines is 3%, so even here it gives half the productivity it should.

Top

Quote

I decided to make some more testing with it, because I know I am not crazy and am not imagining things. no, really, that pink elephant is really flying out of my window, you have to believe me.

Anyway, I decided to test imperial mills, because they take one product and make one product, so they are easy to work with. I set up a game on ice wars, because it is a big map with flat terrain, without any oopponent. I made ten mills. I mined enough iron and coal to be able to produce schites for a long time. I made a stock of donkeys. Then I stopped all the other grain-consuming buildings, and I gradually added farms. At each added farm, I'd see how the productivity of the mills would stabilize. I set the economy target to 1000 flour, I directed all of them to a single warehouse, when the warehouse got nearly full I destroyed it, and directed all the flour to a nearby warehouse I had prepared. Then i rebuilt the destroied warehouse, ready to take flour when I would destroy the other warehouse. That way, i was able to keep an uninterructed flour production. Farm productivity was always above 90%, and generally above 95%. There had never been any road congestion. I looked the productivity for a goood minute (at 7x speed) to make sure it stabilized.

So here are the data, number of farms to the left, productivity (it had some statistical variability, so I give a window for it) to the right.

• 6 3-5%
• 7 4-5%
• 8 4-6%
• 9 5-7%
• 10 6-8%
• 11 7-9%
• 12 8-10%
• 13 8-10%
• 14 9-12%
• 15 10-13%
• 16 11-15%
• 17 12-15%
• 18 12-16%
• 19 13-20%
• 20 17-21%
• 21 18-22%
• 22 20-22%
• 23 20-24%
• 24 24-28%
• 25 23-29%
• 26 28-32%
• 30 35-38%
• 35 50-65%
• 40 70-90%
• 42 90-98%
• 43 100%

With those numbers, it is easy to see that I was right. There should be a linear correlation with the productivity, and indeed there is for some intervals, but the general code is broken. If 43 farms are enough to push the 10 mills to 100% (with very slow accumulation of grain, i checked over half an hour of gametime) Then 22 farms should give a productivity of 50%, and 10 farms around 25%. 30 farms should be around 70%. You can see that the buildings return a wrong productivity value for a productivity below 100%. I don't know where the error in the code is, but this definitely proves it is there. I also have kept the saved game. Shall I open a bug report?

P.S. Since I had 43 farms, I decided to calculate the performance, since the help window always lists performance as "calculation needed". I slowed down the game speed to 1x to minimize human error in keeping track of the time, looked at the stock of flour, and counted 5 minutes with a chronometer. 158 flour were produced in that interval, giving an estimate of one grain every 82 seconds per farm. Seems high, but I have always knows empiire needs tons of farms. It also follows that a fully stocked mill can make one flour every 19 seconds.

Top

Quote

QCS wrote:

Were your farms close enough so farmers reuse plants planted by someone else? That usually increases productivity if done right...

To produce one grain, a farmer needs to plant the field, then to cut the field. it does not matter that two different farmers are doing it, it's always one grain per two farmer actions.

Although i noticed that a farm with plenty of free space sometimes will plant many more fields than it can work, and some of those will rot, with the paradox that a farm with much space will actually produce less. I intend to investigate this issue in the next few days. anyway, they were close enough. it's difficult to fit 43 farms without putting them close. to be exact, they were at the corners of a 4x4 grid.

Edit: That said, productivity is reported per building per cycle, not for economy in time measures, so trying to find a correlation between those two will fail, I guess...

I don't understand your point. productivity is by definition how much the building is producing divided by how much it could produce in the most optimal conditions. so if a mill can produce flour every 19 seconds but but it is supplied with grain only every 25 seconds, its productivity should be (19/25)%. there is no difference between "productivity per building per cycle" and "productivity for economy in time measures" (whatever those are supposed to mean), except for some statistical random variation, but i made 10 mills exactly to get an average. they are simply different methods to calculate productivity, and they must give the same result. If they don't give the same result, it means that one method is wrong.

Top

Quote

wl-zocker wrote:

Did you consider that the mill has an animation of 10s, and a sleep time of 5s? If the mill cannot work, it will try to start work again earlier. This could be the reason for the low productivity with few farms. Remove the animation time and see if the productivity grows linearly with the number of farms.

Yes, good idea. I tried now. I could nt do it with 10 mills, because without animation time they are so fast the grain does not have time to reach them. but i could do it with one mill and 8 farms, close enough to it that they could keep it supplied. 8 farms should give a productivity of around 90%, since 4.3 farms are needed for one mill and i doubled the mill's speed by removing the animation. with 8 farms, it stabilized at 80%. I stoopped 4 farms, and it stabilized at 30-35%. Not eexactly half the yield, but close enough. From that experiment I draw two conclusions

1) the reason for the wrong productivity is that the program will count the failed attempts much faster than the succesful ones, leading to an overestimation of skipped work.

2) animation time is the main component of this effect, but it is not the only one.

For buildings where the worker has to move around the map (wooodcutters, hunters) this cannot be avoided, but for every other building the animation time could be included in the count of how much time it passes before it fills its string with a "failed work"

Top

Quote