ORIGINAL: Numdydar
Of course given my track record in understanding how the game works is very poor, I could be just as wrong about this as R&D acceleration [:D]. But I would like to see some actual data to back up the claims that the total factory size matters versus just the damaged ones.
Numdydar, looks like you fight your battles to the very end. Respect! [:)]
As for the proof for total FactorySize as factor in r&d repair source code is the only true authority and source code is Michael's kingdom. So if you ask him kindly and if he is not too busy squashing bugs or simply too tired with this very topic he will give you your desired definite answer. All I can say with 99.9(9)% certainty level is that in classic WitP everyday repair test check is based on the condition
Random(NumberOfDaysTillPlaneArrival) <= FactorySize
where factory size is sum of active and damaged devices. Use it or loose it as you wish. As I've said before part of having fun with the game is doing as you see fit which includes making mistakes and learning form them.
I know that the highlighted item is definately not acccurate. I have 0(2) factories for aircraft in late 45/early 46 some of which have fully repaired while others have not repaired a single point. I also have factories that are under 10 that have gotten their factories fully repaired while larger factories are still waiting to be fully repaired over the same time frame for planes being available within the same six moth window. I will be glad to show you the data to support this if you want.
I think you don't pay enough attention to "on average" phrase. Remember that r&d factory repair from fully damaged to fully repaired is stochastic process - chain of random events. Think of it as repeated coin flipping where we want to know how many flips is needed to achieve certain fxed number of "heads". Only in our case the coin is heavy biased against us (probability of single success is very low in all practical cases) and that bias is not constant (probability of success grows slowly with every flip). For such process you need to use statistics approach and certainly you need more then few test runs to draw definitive conclusions.
Beeing a true Bayesian Hero ( [:D]) maybe I would be able to calculate in a blink of an eye both expected average as well as three-sigma interval and then use them to evaluate them against your sample results. But as a poor, little chap who tries to stay away from fat statistics textbooks I could offer you only (I did in the other thread) an intution why the fact that probability of single success is in proportion to factory size results in average of total repair times which do not depend on that very factory size (with silent but obvious assumption that NumberOfDaysTillArrival >> FactorySize).
Now I may supplement that with second intuition: though averages of times of full repair does not depend on FactorySize their deviations (and thus confidence intervals) are getting smaller with the the size of factory. With 1(0) factory one lucky flip with our heavily biased coin is enough to get done, with 30(0) factory there is enough time for lucky and unlucky results to even up a bit. Surely deviation still will be there and running say ten 30(0) factories you will see some of them finish repair well before average time whereas others will lag far behind. Again: that's life!