Closed AlphaKretin closed 3 years ago
For posterity, if we are able to pull a dataset from say YGOProDeck, we could try a k nearest neighbours, linear regression, or logistic regression classifier (those are the simplest)
The system to allow classification is largely written at this point, we are now waiting on a list of criteria from the client to bring this feature to MVP status
We've received a list of meta relevant decks and implemented checks for them. Checks for more decks will be an ongoing project, but this feature is now at MVP status.
The module needs to be able to classify a deck in terms of what archetype it is. ygopro-data provides setcodes, but this has many false positives and false negatives. MBT has offered to provide a list of relevant criteria, but I assume that will only be relevant to the current metagame. Depending on future public presence, could collaborate with community on keeping such a list of criteria updated. Some themes will presumably be identified by the presence of various key cards, while others will be defined by a certain density of setcodes - the density will likely depend on the archetype.