User blog:Fouryuu/Drop Mechanics

Thanks to data mining, we now know the exact mechanics behind drops for the Vita version. Here we will see if they apply also to the browser version.

KanColle Kai Drop Mechanics
KanColle Kai stores drop data in drop tables of exactly 100 rows each. Each map gets two tables, one for normal nodes and one for boss node. This means that there are 100 possible results shared by all normal nodes in a map, and 100 possible results for the boss node.

Rows may be empty. Rows may have duplicate ships.

Example drop table:

To determine a drop:
1. See if you get a chance at any drop at all.
 * Normal Node:
 * S: 70%
 * A: 60%
 * B: 70%
 * Boss Node:
 * 100% all ranks

2. Get a RANGE. RANGE determines a subset of the table to draw from.
 * RANGE is determined by RANK, DIFFICULTY, and TURN NUMBER.

3. Get an OFFSET. OFFSET determines where the RANGE starts. 4. Get the result.
 * Every ENEMY COMPOSITION has its own OFFSET value. As rarer ships tend to be closer to the top of the table, harder compositions tend to have lower OFFSET.
 * This is how different normal nodes may have different drops.
 * This is how different enemy compositions in a single node may have exclusive drops (think Prinz in Fall 2015 E4-M).
 * Roll a random number n from RANGE, inclusive.
 * Add OFFSET to n.
 * The ship you get is the nth row of the drop table.

Example:
On Kou difficulty, you score an S rank on 1-1's boss node, against the Ho-class/I-class/I-class composition, which has OFFSET=3.

RANGE = 0 - 49

OFFSET = 3

EFFECTIVE RANGE = 3 - 52

Using the drop table above, you can get anything from 3:Tatsuta to 52:Blank.

Applying Mechanics to Browser
To show evidence that browser uses the same rules, we will need to produce:
 * A 100 row drop table
 * RANGE values for each battle rank (turn count and difficulty are Kai only mechanics)
 * OFFSET values for each enemy composition for each node, and no contradictions by recorded drops with the drop table

Using data from poi-statistics, we may try to reverse-engineer a drop table.

Starting with 2-3 because it has plenty of records for all ranks and a good number of nodes to compare.

2-3 Boss
Data S Data A Data B

The undetailed rates aren't useful. Instead we will focus on per-composition drop rates (by clicking the +). You will quickly see that most values are about the same (~1.75%), or a multiple of that amount. We can count each ~1.75% as one row on the table, and each ~3.5% as two distinct rows, etc. We can ignore the "(no drop)" rates because those are caused by full ship slots (it's already known empirically that boss nodes always drop on S and A). In fact, adjusting for no drop, ~1.75% becomes ~2%. Then we must rearrange the rows such that there are no breaks in between for each rank+composition combo, and not go over 100 rows.

Note that some row placements are arbitrary, but key ones which are thresholds that drop in one composition but not another are significant.

Other Evidence

 * The rank ranges match up with Vita's Historical and Kou ranges.
 * The numbers are "nice". As these mechanics are human-designed, we can reasonably expect patterns and clean numbers.
 * Range values are clean: 0-49, 20-69, 50-99
 * Offsets per enemy composition are all in intervals of 3.
 * Sister ships/classes tend to be adjacent to each other (across thresholds, see 2-3 Boss 52 and 53).
 * Offsets correspond roughly with composition difficulty.