[0.9.2] Production stats way off

[FrozenOne]

Also, mind the bugs in my original thread https://forums.factorio.com/forum/vie ... f=7&t=1914
(for example on the screenshot there the colors of graph are random/not matching, and others described there...)

[slpwnd]

Also, mind the bugs in my original thread https://forums.factorio.com/forum/vie ... f=7&t=1914
(for example on the screenshot there the colors of graph are random/not matching, and others described there...)

Ok, so there was actually a fundamental bug in the production statistics calculation. The consumption and production were added up in the same variable (with different signs) for given frame. This could really skew the numbers especially for longer times (when the frame is big enough for numbers to cancel out - for 5 s interval when 1 frame is 1 tick the effect would probably be negligible). This is because of the historical reasons - the production statistics and energy statistics use the same code and using same variable for the energy statistics was an intentional optimisation. Anyway, this is now fixed for the 0.9.3.

[Holy-Fire]

Also, mind the bugs in my original thread https://forums.factorio.com/forum/vie ... f=7&t=1914
(for example on the screenshot there the colors of graph are random/not matching, and others described there...)

Ok, so there was actually a fundamental bug in the production statistics calculation. The consumption and production were added up in the same variable (with different signs) for given frame. This could really skew the numbers especially for longer times (when the frame is big enough for numbers to cancel out - for 5 s interval when 1 frame is 1 tick the effect would probably be negligible). This is because of the historical reasons - the production statistics and energy statistics use the same code and using same variable for the energy statistics was an intentional optimisation. Anyway, this is now fixed for the 0.9.3.

Thanks, but there's still a bug. Production stats for 5s interval are too high.

This is easy to see in a toy setup with only one furnace.

(There's probably some problem with the way this is sampled that causes an overestimate, which has the strongest effect in a short interval).