Luck/Turmoil versus Nothing.

[rdonj]

I use mainly the oni when playing as yomi, and the bakemono are more like special forces. So I would always take sloth when playing yomi.

Keep in mind that all the nations I listed were supposed to be playable with both sloth 3 and a combination of order3/luck3. Mictlan is probably the hardest to fit into that category with their bless requirements, but with order 3 they can take a little less dominion, since they should be able to fort up pretty quickly.

[Wrana]

Originally Posted by rdonj

I use mainly the oni when playing as yomi, and the bakemono are more like special forces. So I would always take sloth when playing yomi.

Keep in mind that all the nations I listed were supposed to be playable with both sloth 3 and a combination of order3/luck3. Mictlan is probably the hardest to fit into that category with their bless requirements, but with order 3 they can take a little less dominion, since they should be able to fort up pretty quickly.

Yes, that's why I wondered about Mictlan in this list, but not about Yomi or monkey nations. But maybe - as with Order/Luck they should have money for castles and can take less extreme bless probably - just using sacreds together with massed slingers...

[rdonj]

Yeah, that was the thought. It might make more sense to say ma mictlan specifically, since they are not as likely to be doing heavy bloodhunting, so the order scales would have more benefit.

[13lackGu4rd]

well, MA Mictlan works perfectly fine with O3L3, for EA and LA Mictlan T3L3 is a no brainer...

[chrispedersen]

I disagree completely. Using Turmoil3 Luck 3 for EA/LA mictlan needs consideration.

It is not that Mictlan can't be played that way - it can.
I prefer O3 L-2 for mictlan. Mictlan needs to expand as FAST as possible to get a commanding advantage in the early to mid game. This is best accomplished by fort/temple production.

[militarist]

I've checked all threads about order/luck, but still can't understand all mechanics. What is a base % of probability that event will happen this turn? What basic % that it will be good? Does it depend on amount of provinces?

IS there was such a thing as a "basic" probability of event and we knew base prob.for good/bad events, than we could measure the value of each luck scale increase by doing a sum of each event output with taking to account it's probability: (gold value/gems event brings)*probability

I also so somewhere someone made a test, comparing output from 2 different order/luck combination during 100 turns. Could anyone point me to this?

[Finalgenesis]

the 100 turn test is not going to be enough data base anything on, you need a lot more data points then that to draw a reasonable average.

if anything going the other direction: a hard-crunch of event probabilities factoring with average good/gem/unit gain or loss seem best, modified by events certain combo like death scale will unlock. Using Edi's event list as a base is a good start.

Note: The following includes a lot of my own assumption and extrapolation/ answer I got from various posts in this forum. I have not code-dived or tested these assumptions (Not that I'd know how), please feel free (and please do) to correct me if you see any wrong assumption on mechanics, I'd love to know.

Relevant to the hard crunch method: The more province you have the more events happens (the event generation goes through all your province until a max of 3-4 events have been generated. So having 10 provinces will generate 1 event check for each province for a total of 10, increasing effect of luck/misfortune until you consistently get 3-4 events a turn, also note that if an event is generated, the same province gets to do another event check i think, since I've seen 2-3 events on a single province assuming I didn't hallucinate).

Finally, the local luck at each province is a huge factor, even if you have luck +3, if most of your province are under enemy dominion with misfortune scale, your essentially wasting your luck scale. E.g. if you have 10 province, 4 of which has +3 luck, 6 of which has -2 misfortune from enemy dominion, 4 of your event check will get +3 luck bonus while other other 6 gets -2. Also note that you do not get +luck scales if it has enemy dominion on the same province. This is another often overlooked downside to using luck scale, having to keep your dominion up in order to use it, whereas for misfortune scale you could care less if enemy dominion brings you negative luck (hell you may even get neutral luck out of your opponent's positive luck scales).

For the hard-crunch, I think using one solo province as base is best. Things we need to know before we can do it is the base event occurence chance and good/bad event chance. I assume the good/bad event base chance is 50/50, and say if the base event occurence chance is 20% (made-up number), we have a good start. Next thing we need to know is actual % in Edi's event list, as there are rarity 1 and 2 events (2 being the rarer), what is the ratio of rarity 2 event vs rarity 1 event? (I've wanted to know this for the magic site too for a long time) Say rarity 1 event is 4 times as common as rarity 2, we can now calculate the average pay out of any given scales given a single province. 20% event chance -> 50/50 good/bad event list -> 80% rarity 1 20% rarity 2 event -> mix in modifiers like luck/turmoil scales.

If someone can give me those base % chances, I'd be happy to crunch out average pay out under some common scenarios, i envision it something like this:

3T3L3G3M scale - 48% event occurence, avg payout: 132.2 gold, 12.7 fire/water gem, 2.3 slaves ...etc, 6%heroes, 3.1% free units, 1.7% artifact lvl 2...etc

3O1G2U1M scale - ...etc

It'll be very rough calculation to be sure due to the extreme randomness and various factors like dominion / province # in an actual game, but useful nonetheless for reference.

[militarist]

The idea main goal I want to reach is to measure the value of each step of luck scale, or just of each % of luck, basing on events which cause generation/loss of money/gems/items. Ignoring raiding events and other, which are impossible to compare with something.


So, our forumla is:
(number of events in EDI's DB with rarity-2)*(probability of event with rarity-2)+the same for rarity(-1,1,2)= 10% for 1 check. (based on 1 , capital province, was not tested on many provinces).

Squirellord believes that there are 4 independent checks per turn (4 is a limit of events you can have per turn). Each of them is independent from each other.

So 1 check gives 10% of event, 4 checks gave him about average 36% (24 events during 64 turns).
If we knew absolute probabilities of each rarity, or just relative probabilities of rarity-1 to rarity-2, and we expect EDI's DB to be full,
we could get probability of each specific event. Which is a 99% of the way to calculate the value of each % of luck scale.

Then we could just take a sum of all output of all events which could be measured in gems or gold (including temple loss, etc), * each of them on it's probability and we will have some sum of gems and $. Which can be used as a base for farther calculations, and will let us compare luck and order.

It would be interesting to understand mechanics, mostly to simplify some decisions in certain situations, for example - is there any sense in luck scales in rich games, or O3 is mandatory, the absolute value of some heavy events which give you a lot or take a lot from you - is it worth to take special positive scales or avoid some negative to get rid of some events or to get some events.

[Squirrelloid]

that's not enough information to derive the 10%...

1 province, 65 turns:
44 no event
18 1 event
3 2 events

This fits P(event) = 10%, checked 4 times independently, not too shabbily for only 65 turns.

[Finalgenesis]

4 independent check sounds plausible, it would explain the 2-3 events in a single province I've observed.

For the 24 events during 64 turns data, were there any scale modifiers (turmoil, luck)? So it's game tested that 4x 10% base event occurence is about right?

I was thinking about grouping events into 4 pools: Rarity 1 good, R2 good, R1 bad, R2 bad and sum the output/loss for each pool, throw them into probability formula with base event occurence and we have an instant base output figure to play around with (I think that's what your saying also generally). Of course certain scales open and lock some events. Maybe set up an excel table to show probabilities under all scale scenarios...

Another thing I'm worried about is whether all rarity 1/2 event chance is equal, I've seen a lot more brigands or sharks event then any other rarity 1 events...

And we still need the probabilities for Rarity 1 : rarity 2 events to think about any calculations...