[PhilD]

OK, I made my own little test. I started with dominion 1, imprisoned god, no prophet for the whole test.

On the turn before I built a temple, I had a total dominion of 5. On that turn, I built 19 temples, bringing my maximum dominion to 5 and my total temples to 20.

On the same turn, my total dominion jumped to 10 (+5 increase), with dom 2 in my capitol (this at least confirms that dominion spread is after temple building).

Next turn, total dom is 11 (+1), one temple was destroyed by a random event (maximum dominion down to 4)

Next turn, back to 20 temples, total dom is 16 (+5).

Next, down to 18 temples (two destroyed by events!), no increase.

On the 10th turn after I started building temples, my pretender awakened. On that turn, I had a total dom of 33, with 20 converted provinces. My capitol had 5 dom, a few others had more than 1.

At no point during the game did I see a red candle - large map, 2 players (the other was an AI; I should have done it with a second human player, so I could control the second player's dominion).

Now, if one takes into account the fact that the capitol is supposed to spread dominion like a temple, and that I had 2 turns with a maximum dom of 4 instead of 5 (plus the fact that on the last turn, my pretender gave me a guaranteed dominion spread, plus 2 temple checks), my cumulated dominion spread over these 10 turns was 996 - this, with the formula in the manual, should give an expected 99.6 dom increase, and seeing as it's supposed to be the sum of almost 100 independent Bernoulli trials, the total should be pretty close to Gaussian with a standard deviation of 5 - but I only got a total dom increase of 28, which is roughly 14 times the standard deviation below expectation. Statistically, it's a virtual impossibility. No statistician (I'm not one, though I could be described as a probabilist) would accept the hypothesis that dominion spread follows the rules as explained in the manual.

Now, another thing I noticed was that, during the whole test, my dominion was concentrated around my capitol. OK, so it had the equivalent of 2 temples, and should have spread dominion twice as fast as the other provinces, but still - during the whole run, the area covered by my dominion was connected, any provinces with higher than 1 dom were either adjacent or 2 clicks away (except maybe on the final turn), and the only provinces with temples that failed to gain at least 1 dominion were the farthest from home.

All of this supports the idea that capitol-based dom spread might follow the rules as described by the manual, but that the chances for successful temple checks from "real" temples remain based on initial dominion. With such a mechanism, 20 temples and a capitol would give me an expected dominion increase of 2.5 each turn (2 for 20 temples at 10% each, .5 for the home province), which is much, much more consistent with the observation. Or, it could remain based solely on initial dominion, even for the capitol, which would give only 2.1 per turn - in which case getting +28 over 10 turns, even with the pretender at the end, would be a bit high.

Now, I don't know if this will be considered sufficient evidence by IW, but I cannot believe that the dominion spread rules as described in the manual are what is implemented in the code.