Board Thread:Wikia Discussion/@comment-26091666-20151103015609/@comment-26486243-20151123015254

The API data, when submitted, is really good. I've also found another source as well, which was literally hiding right under our noses.

Poi-statistics. The site I was already using for updating drop information. For all drop types, including blank drops, it will record the composition of the battle node.

Annoyingly it doesn't tell me which is easy, which is medium and which is hard, but it's possible to make educated guesses based on the difficulty of the ships (as a general guide, the longer the entire composition's name is, the stronger the composition is), as well as the degree of overlap (if 3 strong types have 400, 2 middling types have 100, 3 weak types have 600, 1 type has 500 (and is between medium and hard difficulty) and 1 type has 700 (and is between easy and medium difficulty), I can easily shove the 400s in Hard, the 100s in Medium, the 600s in easy, the 500 in both Hard and Medium and the 700 in both Medium and Easy for a total of 4 compositions each. This is not at the same level of proof as a screenshot on our thread, let alone an API call, but it is about as reliable as Wikiwiki is - and MUCH FASTER.

There's an additional issue that poi-statistics doesn't recognise PT Imp Packs yet, so every time they appear, the poi version will omit them, showing fewer ships than there really is, so poi-statistics is only useful for nodes that do not contain them.

In addition I've also been collapsing identical compositions with different formations into single entries, and using  tags in order to newline the formation name, just like I already did with the node names. If I placed a comment citation in the formation name, it applies only to the one it comes immediately after. The citations placed next to the |- or }} apply to both formation types. This allows for the same level of flexibility in citing sources, while also reducing the number of boxes that show up, so people don't have to scroll so much. However, if one source is confirmed and the other isn't (say, one is poi-statistics only and the other is both wikiwiki AND poi), I still list the single-citation one with an asterisk separately from the double-citation one without, in order to indicate which can be trusted more and which may be less reliable.

Right now, my required degrees of proof go like this, in decreasing order of reliability: In addition any variant of Abyssals which share the same HP value and image that isn't proven by API call or stated explicitly by Wikiwiki will be marked as # to indicate the variants aren't confirmed.
 * 1) API calls
 * 2) Screenshots
 * 3) Wikiwiki + Poi-statistics
 * 4) Wikiwiki only (* for unreliability - in my updates, Wikiwiki has had wrong ship types, wrong formations and wrong variant types before)
 * 5) Poi-statistics only (* for unreliability - while poi can't be wrong with the ship types or formations, my assignment of difficulty may be wrong)