Luck/Turmoil versus Nothing.

[chrispedersen]

Originally Posted by sector24

Something I noticed. Chris, when you were determining whether number of provinces affected the number of events, you used a territory size of 4 vs 8.

This was clearly (to me) not a test of whether terrority size affects the number of events. Simply not enough provinces to provide any meaningful results. Then when a test comes along using larger number of provinces that actually might provide an accurate test of the hypothesis, you bash it. But you bash it by comparing it to your 4 vs 8 province test! Everyone else can make up their own minds, but I find a significant lack of objectivity here.

I certainly do not bash the test, and in fact welcome more tests.
And I wouldn't mind being proved wrong on it either.

I certainly don't *mind* if there is an alternate mechanism.

I certainly have no problem with the data,
However having a 1000% increase in the number of provinces yielding an 18% increase in the number of events does not qualify as support saying that events are a function of # of territories. IF we are going to get a valid model here, I think we have to poke holes in all theories.

For curiousity sake, I wonder why do you think that a 4 territory vs an 8 territory is insufficient to determine meaningful results? Especially over 50 turns?

[Finalgenesis]

Thanks for the data TJ

Oh here's a Void summon data for Ryleth lovers, (definitely not enough turns to conclude on anything):

MA Ryleth over 30 turns
a) only summoner(3) is used, priest replaced on lvlup
b) Summon begins on 3rd turn
test 1: 1 order 3 luck 3 magic 3 growth 3 cold
test 2: 3 hot, all other baseline

Test 1: 8 success, 15 spawns total
Test 2: 7 success, 11 spawns total

I can post spawn types on request.

Quality of spawn shows no correlation, all over the place. High summon skill seems to be most important in determining quality. Scales don't seem to have a noticable impact on void summon base on this tiny testing population.

[thejeff]

I was using T3L3, largely because that's what I used on my earlier 300 province test. That does mean it's not directly comparable to your tests.

In both the 30 & 300 province tests, I used map commands to assign provinces to both nations and to put a temple in every province. That's the fastest way I could think of to boost dominion and scales with it. If you can think of a faster way, I'd love to hear it. Events seem to be scarcest in the first 10 turns or so. I'd like to see if that's just scale spreading or something more hardcoded. I think there are a handful of events restricted in the first few turns, but I didn't think it would be that big an effect.

I don't think that more provinces with bigger province terrain masks makes sense. At least not if we're still thinking you start by picking 0 to 4 provinces and checking for events in them. I could see that having a large effect in the small tests, where you might have a province with (terrain/population?) that restricts many events. But scales should have been similar in most provinces and those unlock the largest number of events right?

I suppose I could try modifying a map to have only one terrain type and rerun them against that, but it doesn't seem very useful.

The province number effect is obviously not linear and I don't have an idea for how it's generated, but I'm not sure how I can demonstrate it more clearly.

[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.

[militarist]

It's also possible that it's linear, just Capital is counted as 10-20 provinces for all calculations.

[chrispedersen]

50 turns.

#turmoilevents set to 20.
Nation 1: O3P3 9 events
Nation 2: O3P3 6 events
Nation 3: 03P3 4 events
Controls:
Nation 4: -P3 5 events
Nation 5: P3 22 events.

One territory.

With turmoil events set to 20, this should be a -60% chance of events.

I have no explantion for nation 4, either.

[chrispedersen]

5 nations turmoil events set to 20
50 turns.

8 territories.

ermor O3 5 events
Mar o0 30 events
Saur 03 5 rvents
Agartha o3 5 events
Kailas o0 31 events

Oh, and dominions crashed with turmoil events set to 25.