detect_archetype.py

import numpy as np
import matplotlib.pyplot as plt
import scipy.cluster.hierarchy as hcluster
from sklearn.manifold import TSNE

import collections
import random

import fireplace.utils
from fireplace import cards
from hearthstone.enums import CardClass

try:
    del decks
    del data
    del clusters
except NameError:
    pass

NUM_ARCHETYPES = 30
NUM_TECH_CARDS = 5
NUM_COMMON_CARDS = 30 - NUM_TECH_CARDS

try:
    decks
except NameError:
    cards.db.initialize()
    decks = []
    archetypes = []
    for _ in range(NUM_ARCHETYPES):
        deck = fireplace.utils.random_draft(CardClass.WARRIOR)
        archetypes.append(deck)

    for _ in range(1000 - NUM_ARCHETYPES):
        tech_cards = fireplace.utils.random_draft(CardClass.WARRIOR)
        base_deck = random.choice(archetypes)
        deck = base_deck[:NUM_COMMON_CARDS] + tech_cards[NUM_COMMON_CARDS:]
        decks.append(collections.Counter(deck))

# dims = { int(k.split("_")[-1]) for d in decks for k in d.keys() }

try:
    data
except NameError:
    lookup = list({card for deck in decks for card in deck})
    data = []
    for deck in decks:
        datapoint = [0] * len(lookup)
        for card in deck:
            card_dimension = lookup.index(card)
            datapoint[card_dimension] = deck[card]
        data.append(datapoint)
    data = np.array(data)

try:
    clusters
except NameError:
    clusters = range(1000)
    thresh = 0.5
    # clustering
    while len(set(clusters)) > len(data)/5: # minimum 5 clusters <--- only parameter we have to tune [threshold i mean]
        thresh += 0.5
        clusters = hcluster.fclusterdata(data, thresh, criterion="distance")
        print(thresh, len(set(clusters)))

# plotting
model = TSNE()
embed = model.fit_transform(data)
plt.axis("equal")
plt.scatter(*np.transpose(embed), c=clusters)
plt.title("thres %f; clusters: %d" % (thresh, len(set(clusters))))
plt.show()