Dicewars

A simplified version of game DiceWars between two players goes such that each player is assigned a stack of 6-sided dice which equate to relative power. On a player’s turn, the player challenges other player. The challenge is resolved when each player rolls their dice, with the higher total winning.

Each player roll all of their dices and the value at the top of each dice is added to find the players score. Whoever has the higher score wins and the equal score is a tie. The goal is to find the probability that, in a game between two players with 𝑚 and 𝑘 dice respectively, any player wins over the other or there is a tie.

Before we calculate the probability of win or a draw, lets calculate the probability that a player with 𝑘 dices rolls a total of 𝑛. Since for a standard dice the number on each face is one of 1,2,3,4,5,6. When the player rolls 𝑘 dices the minimum sum is 𝑘 when each dice rolls to be 1 and a maximum of 6⁢𝑘 when each one rolls to be 6 thus 𝑘 ≤𝑛 ≤6⁢𝑘.

The total exhaustive cases of rolling 𝑘 dice is 6𝑘. The total possible ways we can get a sum of 𝑛 (subject to above constraint) when we roll 𝑘 dice is the same as the total number of ways we can write 𝑛 as a sum of 𝑘 integers such that each integer𝑛𝑖 is between 1 ≤𝑛𝑖 ≤6. The order of integers that we break the number 𝑛 is important too.

Lets us define a function 𝑓𝑘⁡(𝑛) such that it gives the total number of ways we can break 𝑛 into 𝑘 integers where each integer is one of 1,2,3,4,5,6

Because of the constraint 𝑘 ≤𝑛 ≤6⁢𝑘 we can immediately write

It is not hard to see the constraint

These are the edge cases we can identify because of constraints in hand. Before trying to find the function for general 𝑘 and 𝑛 lets try a special case.

For a special case of 𝑓3⁡(7), i.e., the number of ways we can express 7 as a sum of 3 integers between 1 and 6

We can write 7 as a sum of 2 integers as

1 +6 or 2 +5 or 3 +4 or 4 +3 or 5 +2 or 6 +1 There are a total of 6 ways we can do it. But if we want to express it as sum of 3 integers then we can write the second integer as a sum of 2 integer. Then the number of ways we can express 7 as a sum of three integers becomes the sum of the number of ways we can express 6 as sum of 2 integers plus the number of ways we can write 5 as a sum of 2 integers and so on.

This leads us to write

This suggests us a general expression for 𝑘 ≥2

This is a recursive function the general expression for all 𝑘 then is

Below is a python function that calculates this function

def F(k,n,minno=1,maxno=6):
    if n < 1 or n < k:
        return 0
    if k == 1:
        return 0 if n > maxno else 1
    else:
        return sum([F(k-1,n-i) for i in np.arange(minno,maxno+1,1)])

The probability 𝑝𝑘⁡(𝑚) that we get a total of 𝑛 when we roll 𝑘 dice can then simply becomes

Where the denominator is the number of all possible ways of rolling 𝑘 dice.

Below is a python function that calculates this probability

def p(k,n,dice_op=np.arange(1,6+1,1)):
    return F(k,n)/len(dice_op)** k #len (dice_op) is just 6

Before we go to the game of win an draw lets see graphically how do the number look like. The plot below shows the count (left) and probability(right) of the sum (x-axis) when 𝑚 dice (inset) are rolled

def plot_counts(ms,dice_op=np.arange(1,6+1,1),plpr=False,ax=None):
    for m in ms:
        total_op = len(dice_op)**int(m)
        ns = np.arange(m,6*m+1,1)
        cnts = [F(m,n) for n in ns]
            
        pvl = np.array(cnts)/total_op if plpr else cnts
        ax.plot(pvl,'-o',label=f'm={m}')

    ax.legend()

fig,ax = plt.subplots(1,2,figsize=(18,6),sharex=True)
plot_counts([1,2,3,4],plpr=False,ax=ax[0])
plot_counts([1,2,3,4],plpr=True,ax=ax[1])
ax[0].set_xlabel('Sum'); ax[0].set_ylabel('Count')
ax[1].set_xlabel('Sum'); ax[1].set_ylabel('pmf');
plt.tight_layout()
plt.savefig('Prob_count',dpi=120,bbox_inches='tight')

Probability of tie

The pobability of tie in a game of two players throwing 𝑘 and 𝑚 die respectively is simply the probability that they roll the same sum. The set of possible sum of player with k die is

Similarly for other player

Which can be easily written as

The python function for this written as

def get_tie_prob(m,k):
    n1 = np.arange(m,6*m+1,1)
    n2 = np.arange(k,6*k+1,1)
    return sum([p(m,n)*p(k,n) for n in n1 if n in n2])

I have this helper function that calculates generates the table of values.

def generate_table(ms,ks,func=get_tie_prob):
    prob_table = np.zeros(shape=(len(ms),len(ks)))
    for m in ms:
        for k in ks:
            prob_table[m-1][k-1] = func(m,k)
    df = pd.DataFrame(prob_table,columns=ms,index=ks)
    return df

Lets say 𝑚⁢𝑠 is the set of number of dice player1 has and 𝑘⁢𝑠 be the number of dice playe 2 has then

maxn=6
ms = ks = np.arange(1,maxn+1,1)

%time tie_prob_table = generate_table(ms,ks,get_tie_prob)

CPU times: user 2.29 s, sys: 18.3 ms, total: 2.31 s
Wall time: 2.29 s

tie_prob_table

The probability of tie when players have (𝑚,𝑘) dice respectively corresponds to the cell (𝑚,𝑘) in the table above.

The probability for win

Similarly the probability of win for player with 𝑚 dice is the probability that the player 𝑘 throws a sum 𝑖 times the sum of probabilities of ways player 𝑚 throws 𝑗 dice such 𝑗 >𝑖

The python function implementing this function is

def get_win_prob(m,k):
    if m > k:
        wkm = get_win_prob(k,m)
        tkm = get_tie_prob(k,m)
        return  1 - wkm - tkm
    n1s = np.arange(m,6*m+1,1)
    n2s = np.arange(k,6*k+1,1)
    
    win_prob = sum([p(k,i) * sum([p(m,j) for j in n1s if j > i]) for i in n2s])
    
    return win_prob

Generating the table

%time win_prob_table = generate_table(ms,ks,get_win_prob)

CPU times: user 18.9 s, sys: 152 ms, total: 19 s
Wall time: 18.9 s

win_prob_table

The probability of win for player 𝑚 when players have (𝑚,𝑘) dice respectively corresponds to the cell (𝑚,𝑘) in the table above.