taw / magic-search-engine

Search engine for Magic cards
MIT License
47 stars 19 forks source link

Decklists with ambiguous cards #141

Closed taw closed 3 years ago

taw commented 4 years ago

Data for decks (including JSON export) is here.

It generally only lists card name, and not its set or number. Indexer has extremely accurate heuristics to figure out which set a card comes from.

But it pretty much gives up on the question which printing within a set it is.

This is mostly because vast majority of such conflicts are basics, and we don't have information which mix of basics a deck includes, and honestly we mostly don't care.

Unfortunately between recent sets now containing multiple printings of each card, and mtgjson including a lot more card variants in its database, such conflicts proliferated.

There are 1222 basics-only conflicts (and indexer only cared for sets with full art land, as none of the decks have those).

There are also 199 other conflicts, full list below. First, a few heuristics need to be added to the indexer (mostly about mtgjson variants, and matching by foiling) to reduce this list. After that the rest need to be annotated manually.

taw commented 4 years ago

Here's the initial list of 199 conflict cards:

taw commented 4 years ago

Indexer changed so it ignores any card with special characters if card with base number exists. I think this is correct every time.

For other cards, it now always picks one with lowest number, which should fix most of the issues. There are the remaining conflicts. If anything is wrong, it will need manual annotation (which is super easy, barely an inconvenience, but someone needs to give me a list):

taw commented 4 years ago

After adding rules that consider full art / frame effects, the list is reduced to just 31:

taw commented 4 years ago

What's left is just 13 cards. I don't think these are possible to figure out without watching unboxing videos. If anyone wants to do that, let me know:

taw commented 3 years ago

New issues with JMP and CMR.

Figured out a lot of them, remaining:

taw commented 3 years ago

And CM2 is also resolved. The whole thing can now be closed.