random hero selection generatot

Heyall,
here i post some pseudo-code on the random generator.
Current problems:

1. some heroes appear twice as often in the game than others.
   
2. often, there is an imbalance with regard to stuns per team
   
3. often, there is an imbalance with regard to healers per team
   

proposed fix:

1. adjust random chances of double-heroes (kubi, tc/batman, High Elf/Grim Reaper...)
   
2. make the random generator increase/decrease the likelyhood of selecting a stun hero, based on the heroes that already spawned
   
3. make the random generator increase/decrease the likelyhood of selecting a healer, based on the heroes that already spawned
   

plz make commments on the heal/ stun weights below here is a editable list

Code:


prepare table of heroe properties (double hero, stun weight, heal weight)
prepare number of desired heals & stuns in the game

for each -random spawn
    get the combined stun and heal weights of all heroes that spawned already in the same team
    determine the number of stuns & heals the next spawning hero should have based on the desired and actual (already spawned) numbers
    convert the desired number of heals & stuns into a probability weight for each hero to be spawned
    throw the dice
    spawn winning hero
end for
   



here is a python example:

import numpy as np

# number of stuns and heals that we desire ingame
targetNrOfStuns = 2.5
targetNrOfHeals = 1.5
NumberOfPlayers = 5  # players per team, possibly 6 later
WeightWeights=[1,1,1] # weight the impact of double hero, stun and heal modifier


# Alliance Hero stats matrix
# hero number # double-hero-weight # stun weight # heal weight # name

statsAlliance=np.array([
[ 1 , 1 , 0 , 0 ] , # potm
[ 2 , 1 , 0 , 0 ] , # ilidan
[ 3 , 0.8, 0.2, 0 ] , # warden
[ 4 , 1 , 0 , 0 ] , # pandaren wardancer
[ 5 , 1 , 0.8, 0.8] , # high elf
[ 6 , 1 , 0 , 0 ] , # samurai
[ 7 , 1 , 0 , 1 ] , # pala
[ 8 , 1 , 1.2, 0 ] , # mountain king
[ 9 , 1 , 0.3, 0 ] , # panda
[ 10 , 1 , 1 , 0.2] , # arthas
[ 11 , 0.5, 1 , 0 ] , # batman
[ 12 , 1 , 0 , 0 ] , # naga viper
[ 13 , 1 , 1.2, 0 ] , # barbarian
[ 14 , 1 , 1 , 0 ] , # tinker
[ 15 , 1 , 0.1, 0 ] , # bloodmage
[ 16 , 1 , 1 , 0 ] , # wizard
[ 17 , 1 , 1 , 0.3] , # kotg
[ 18 , 1 , 1 , 0 ] , # firelord
[ 19 , 1 , 1 , 0 ] , # ranger
[ 20 , 0.5, 0 , 1 ] , # kubi
[ 21 , 1 , 0.5, 0 ] , # naga queen
[ 22 , 0.8, 0.9, 1 ]]) # druid


# Horde Hero stats matrix
# hero number # double-hero-weight # stun weight # heal weight # name

statsHorde=np.array([
[ 1 , 0.5, 0 , 0 ] , # blademaster
[ 2 , 1 , 0 , 0 ] , # dark ranger
[ 3 , 0.8, 0 , 0 ] , # skeleton warden
[ 4 , 1 , 0 , 0 ] , # axemaster
[ 5 , 0.5, 0.8, 0.8] , # grim reaper
[ 6 , 1 , 0 , 0 ] , # naga???
[ 7 , 1 , 0 , 0.8] , # death knight
[ 8 , 1 , 1 , 0 ] , # tauren warrior
[ 9 , 1 , 0.5, 0 ] , # pitlord
[ 10 , 1 , 1 , 0 ] , # beasty
[ 11 , 0.5, 1 , 0 ] , # taure
[ 12 , 1 , 0 , 0 ] , # alchi
[ 13 , 1 , 1 , 0 ] , # crypt lord
[ 14 , 1 , 1 , 0 ] , # clown
[ 15 , 1 , 1.3, 0 ] , # dreadlord
[ 16 , 1 , 0 , 0 ] , # farseer
[ 17 , 0.8, 0.9, 1.2] , # shadowhunter
[ 18 , 1 , 0 , 0 ] , # zuul
[ 19 , 1 , 0.3, 0 ] , # lich
[ 20 , 1 , 0.5, 0 ] , # archimonde
[ 21 , 0.5, 0 , 1 ] , # kubi
[ 22 , 1 , 1 , 0 ]]) # nether mage

# alliance player now types "-random",



# the numbers of current heroes must be fetched (here hardcoded):
numbersOfSpawnedHeroes=np.array([7,22,1])
# = pala, druid and potm already on the field (hardcoded)

#assign alliance stats, less the already spawned heroes
stats=np.delete(statsAlliance,numbersOfSpawnedHeroes-1, axis=0)

SumStats=sum(stats[numbersOfSpawnedHeroes-1,2:])
# would be [2.8 stuns ,0.9 heals] now from pala, druid and potm

#now check how much is "left" to the desired number of stuns & heals
#taking into account the number of players (make heals & stuns not preferrably
# chosen for the first guy who makes -random)
SumDiff = (len(numbersOfSpawnedHeroes)/NumberOfPlayers) * \
(np.array([targetNrOfStuns,targetNrOfHeals])-SumStats)

#finally calculate the probability bias for each hero
chanceDouble = WeightWeights[0]*stats[:,1]
chanceStun   = WeightWeights[1]*np.exp(stats[:,2]*SumDiff[0])
chanceHeal   = WeightWeights[2]*np.exp(stats[:,3]*SumDiff[1])
bias=chanceDouble*chanceStun*chanceHeal
chance=bias*np.random.random(len(bias))

#aaaaand the winner is.....
WinnerIndex=chance.argmax()

#now spawn the winning hero