I think the best way to find out about the mechanics and odds about Pinkerton is by putting up a theory and comparing it to the actual gathered data to verify if it is true or not. Most of the theories will end up being false. Some theories are simple facts for your knowledge.
So let's play Mythbusters!
The structure will be as followed:
Text written in purple are the theories.
This theory is: TRUE/
FALSE/
Inconclusive
Text why the theory is true/false/inconclusive so far.
Pinkerton has a limited pool.
This theory is: TRUE
This is a simple fact. He never offers an item above the rarity of 2*. Usually he offers all 1* and 2* items of most item types but there are exceptions.
Materials don't include the Transmutation 1* and 2* items.
Battle has a fixed pool of exactly 4 items although there could be more.
Other only covers the Trinket items.
Specialty and Skins are not covered by Pinkerton at all.
Pinkerton offers all items Crim, his sister, asks for as the lore would suggest.
This theory is: FALSE
Crim asks for items outside of the 1* & 2* category which Pinkerton does not provide. So the lore of him giving away any of Crims items is actually wrong as well.
You are able to obtain holiday exclusive items.
This theory is: TRUE
So far only the 3 food point items from the micro holidays are obtainable like the
Heavy Grainbasket as they are 2* items.
For the following, if I speak about all items I mean all items that Pinkerton can provide.
All items have the same chance of being pulled.
This theory is: FALSE
Similar to gathering it seems that more than one step of RNG is at work here.
The first RNG step is which type of item you will get.
Second step is which item. If there is a step for rarity as well will be discussed later.
This can be concluded by looking at the All-time pulls-graph compared with how big the pools for these types are. Battle, the smallest pool with 4 items, appears just as often as Other items with a pool of nearly 200 items.
All item types (Food, Material, Apparel etc.) have the same chance of being selected.
This theory is: FALSE
When looking at the cake-graph for all-time pulls it is obvious that the chances are more in favor for Food and Material.
Food and Material have a frequency of 3/10 each.
Apparel, Familiar, Battle and Other have frequency of 1/10 each.
This is unfortunate since the pool for apparel items is just as big as Foods but has a much smaller chance of getting pulled.
2* items are less likely to be pulled than 1*
This theory is: FALSE
If we only look at pulled items percentages it might seem so. But if we compare these numbers to how the pool is divided into 1* and 2* items it is clearly visible that both percentages are the same. Usually there are more 1* items in a pool then 2* so the eye is easily deceived.
The All-time pulls are as follows in percentag:
Type
Food
Material
Apparel
Familiar
Battle
Other
|
--
|
1*
69%
65%
42%
50%
74%
57%
|
--
|
2*
31%
35%
58%
50%
26%
43%
|
--
|
|
|
|
|
|
|
|
|
--
|
# in pool 1*
519 (66%)
244 (64%)
350 (45%)
132 (55%)
3 (75%)
97 (53%)
|
--
|
# in pool 2*
264 (34%)
139 (36%)
430 (55%)
110 (45%)
1 (25%)
86 (47%)
|
There is a high chance to get a lot of Battle items from Pinkerton
This theory is: TRUE
Although the Battle pool is at a disadvantage vs Food and Materials and only has a chance to be picket 1/10 of the time. The pool itself is very small with only 4 items in it. This grants each of these 4 items the chance of around 2.5%. This may sound very low, but is actually incredibly high compared to the chance of 0.04% per Food item which are "boosted". So if you're feeling Pinkerton gives out Battle items most of the time it is technically correct and a pool of 4 items is easier memorable vs picking 3 different Foods in a row.
The former error of Material items being labeled as Trinket in the Tooltip at Pinkertons page had an influence on the odds
This theory is: Inconclusive with high tendency to
FALSE
This is something that is hard if not impossible to verify without being able look at how the item data is stored and handled by Pinkerton-RNG.
If the theory is true then the odds for items that are not Materials/Other is not affected.
The total frequency of Material+Other would be 4/10. Both together cover nearly 600 items with Material covering nearly 400 items (2/3) and Other covering about 200 (1/3).
The cake-graph shows that if only looking at those two Material shows up 70% (~2/3) of the times versus Other 30% (~1/3). Following this logic we can assume that this theory is false.
By now this error was corrected without a change visible here.