[thejeff]

Well, it's obvious from experience and a casual look at the tests that any relationship isn't linear. If it was, either you'd get almost no events with 1 province or you'd max out at 4 a turn well under 300 provinces.
That's partly what I was trying to confirm.

One potential issue with the 4 vs 8 test is that, assuming your theory about terrain masks limiting the number of events holds, it's much more likely that the smaller number won't be a representative sample of terrain types.
If you're testing for the effects of number of provinces, number of turns is less important. Going for a big difference magnifies any effects so they're more easily noticeable. If the effect is a 1.18 multiplier for every 10-fold increase, a simple doubling might not be noticeable in the random noise.
I'd done the 300 first for other reasons, then tried the 2. When I found so many less events in 2, I tried 30 to see if it would fall neatly in between, which it did.

And, frankly, if it's repeatable, a 18% increase in events for a 10 fold increase in province is precisely a function based on number of territories. What's needed to confirm it is more tests. See if the pattern holds. It's quite possible the randomness is so high, it will be hard to pick any actual meaning without huge amounts of data. And since there isn't any easy way to automate this, it's boring, and I'm not being paid, I'm not planning to run hundreds of tests.