Tip: Site Searching Strategy

[JepSan]

Thanks Edi, thats very interesting. I have been wondering a bit about the concrete mechanics of magic site distribution.

Could you - or somebody else perhaps - tell me:

1) What is the difference between common/uncommon/rare sites? You mention someting about a reroll being required for a rare site to appear?

2) How often does the different site types appear? I would guess (from experience) that at least 70% of the magic sites are common sites, but perhaps I have just not looked hard enough for the rare (and often high-level) sites.

3) The graphics icons of the magic sites differ - even among sites of the same magic type (fire, death, earth, etc.). Does th different icons indicate the rarity of the site (common, uncommon, rare) or the magic path level (1-4) you need to find the site?

[Edi]

1) There are 4 possible slots for sites per each province. First slot is checked against site probability. If site appears, it is randomly determined. Then next slot is checked for site probability. As soon as a site slot comes up as empty, it and all the slots following it will remain empty. So if the first one comes up empty, no sites at all in the province.

Common sites have a greater likelihood to appear than uncommon sites, which again have a greater probability than rare sites. If the random determination results in a rare site, that result is rerolled and the second result will be the actual site that appears. In other words, a rare site appears only if the reroll also comes up as rare.


2) The exact percentages of common/uncommon/rare are not known, but I'd throw an off-the-cuff guesstimate of 60%/30%/10%, give or take something. It'd need some statistical analysis of a fairly large sample size with controlled terrains to get reliable results, and even those would be rough results.

3) The relationship of site icons to their level and rarity is not clearly established, but there seems to be some correlation, in some cases at least. The correlation seems to be more closely to the site level than to rarity.

[Kaljamaha]

Edi said:
1) There are 4 possible slots for sites per each province. First slot is checked against site probability. If site appears, it is randomly determined. Then next slot is checked for site probability. As soon as a site slot comes up as empty, it and all the slots following it will remain empty. So if the first one comes up empty, no sites at all in the province.

Is this really so? I usually play with 30-35% sites, and according to what you said, the site frequency for, let's say 30% sites would be:

0 sites: 70% provinces
1 site: 21% provinces
2 sites: 6.3% provinces
3 sites: 1.9% provinces
4 sites: 0.8% provinces

Even though I haven't done any statistical testing, I am fairly confident this is not what I'm seeing. Even for 50% site frequency (say, the abovementioned 30% plus terrain bonus) the numbers would be 50 (0), 25 (1), 12.5 (2), 6.25 (3), 6.25 (4).

As opposed to a binomial distribution, which for example for 30% sites would look like this: 24 (0), 41 (1), 26 (2), 8 (3), 1 (4).


K.

[Edi]

I'll take your word for that. I barely scraped through statistical and probability math at high school and failed every course on the subject after that.

It's not that I have a problem understanding the results of statistical analysis and probabilities, but trying to figure out the actual process from raw data, I screw that up every time.

So if the binomial model looks like it fits the observations better, then it's probably more accurate.

[LoloMo]

Is there any chance for Capitals to have magic sites other than those you start with? In other words, do you search your capital for magic sites?

[Meglobob]

LoloMo said:
Is there any chance for Capitals to have magic sites other than those you start with? In other words, do you search your capital for magic sites?

No and no.

But they can get a site via a random event.

[Jarl]

Can someone please make a suggestions for a new algorithm - that make a better default target for a site search spell? Then we can hope that Johan will implement it and include it in the next fix...

Make it prefer areas that:
- is'nt close to enemy borders (if possible)
- has'nt got any located sites yet
- start with terrain that will favor your capitals mages
- has been searched by other magic types, but not located any sites yet
- etc.

A better algorithm would save a lot of boring checking time, especially on large maps...