Combinatorial Game Theory | Set 4 (Sprague – Grundy Theorem)

Prerequisites : Grundy Numbers/Numbers and Mex
We have already seen in Set 2 (https://www.geeksforgeeks.org/combinatorial-game-theory-set-2-game-nim/), that we can find who wins in a game of Nim without actually playing the game.
Suppose we change the classic Nim game a bit. This time each player can only remove 1, 2 or 3 stones only (and not any number of stones as in the classic game of Nim). Can we predict who will win?
Yes, we can predict the winner using Sprague-Grundy Theorem.

What is Sprague-Grundy Theorem?
Suppose there is a composite game (more than one sub-game) made up of N sub-games and two players, A and B. Then Sprague-Grundy Theorem says that if both A and B play optimally (i.e., they don’t make any mistakes), then the player starting first is guaranteed to win if the XOR of the grundy numbers of position in each sub-games at the beginning of the game is non-zero. Otherwise, if the XOR evaluates to zero, then player A will lose definitely, no matter what.

How to apply Sprague Grundy Theorem ?
We can apply Sprague-Grundy Theorem in any impartial game and solve it. The basic steps are listed as follows: 

1. Break the composite game into sub-games.
2. Then for each sub-game, calculate the Grundy Number at that position.
3. Then calculate the XOR of all the calculated Grundy Numbers.
4. If the XOR value is non-zero, then the player who is going to make the turn (First Player) will win else he is destined to lose, no matter what.

Example Game : The game starts with 3 piles having 3, 4 and 5 stones, and the player to move may take any positive number of stones upto 3 only from any of the piles [Provided that the pile has that much amount of stones]. The last player to move wins. Which player wins the game assuming that both players play optimally?

How to tell who will win by applying Sprague-Grundy Theorem?
As, we can see that this game is itself composed of several sub-games. 
First Step : The sub-games can be considered as each piles. 
Second Step : We see from the below table that 

Grundy(3) = 3
Grundy(4) = 0 
Grundy(5) = 1 

We have already seen how to calculate the Grundy Numbers of this game in the previous article.
Third Step : The XOR of 3, 0, 1 = 2
Fourth Step : Since XOR is a non-zero number, so we can say that the first player will win.

Below is the program that implements above 4 steps. 

C++

/*  Game Description-
 "A game is played between two players and there are N piles
 of stones such that each pile has certain number of stones.
 On his/her turn, a player selects a pile and can take any
 non-zero number of stones upto 3 (i.e- 1,2,3)
 The player who cannot move is considered to lose the game
 (i.e., one who take the last stone is the winner).
 Can you find which player wins the game if both players play
 optimally (they don't make any mistake)? "

 A Dynamic Programming approach to calculate Grundy Number
 and Mex and find the Winner using Sprague - Grundy Theorem. */

#include<bits/stdc++.h>
using namespace std;

/* piles[] -> Array having the initial count of stones/coins
            in each piles before the game has started.
   n       -> Number of piles

   Grundy[] -> Array having the Grundy Number corresponding to
             the initial position of each piles in the game

   The piles[] and Grundy[] are having 0-based indexing*/

#define PLAYER1 1
#define PLAYER2 2

// A Function to calculate Mex of all the values in that set
int calculateMex(unordered_set<int> Set)
{
    int Mex = 0;

    while (Set.find(Mex) != Set.end())
        Mex++;

    return (Mex);
}

// A function to Compute Grundy Number of 'n'
int calculateGrundy(int n, int Grundy[])
{
    Grundy[0] = 0;
    Grundy[1] = 1;
    Grundy[2] = 2;
    Grundy[3] = 3;

    if (Grundy[n] != -1)
        return (Grundy[n]);

    unordered_set<int> Set; // A Hash Table

    for (int i=1; i<=3; i++)
            Set.insert (calculateGrundy (n-i, Grundy));

    // Store the result
    Grundy[n] = calculateMex (Set);

    return (Grundy[n]);
}

// A function to declare the winner of the game
void declareWinner(int whoseTurn, int piles[],
                    int Grundy[], int n)
{
    int xorValue = Grundy[piles[0]];

    for (int i=1; i<=n-1; i++)
        xorValue = xorValue ^ Grundy[piles[i]];

    if (xorValue != 0)
    {
        if (whoseTurn == PLAYER1)
            printf("Player 1 will win\n");
        else
            printf("Player 2 will win\n");
    }
    else
    {
        if (whoseTurn == PLAYER1)
            printf("Player 2 will win\n");
        else
            printf("Player 1 will win\n");
    }

    return;
}


// Driver program to test above functions
int main()
{
    // Test Case 1
    int piles[] = {3, 4, 5};
    int n = sizeof(piles)/sizeof(piles[0]);

    // Find the maximum element
    int maximum = *max_element(piles, piles + n);

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int Grundy[maximum + 1];
    memset(Grundy, -1, sizeof (Grundy));

    // Calculate Grundy Value of piles[i] and store it
    for (int i=0; i<=n-1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER1, piles, Grundy, n);

    /* Test Case 2
    int piles[] = {3, 8, 2};
    int n = sizeof(piles)/sizeof(piles[0]);


    int maximum = *max_element (piles, piles + n);

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int Grundy [maximum + 1];
    memset(Grundy, -1, sizeof (Grundy));

    // Calculate Grundy Value of piles[i] and store it
    for (int i=0; i<=n-1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER2, piles, Grundy, n);   */

    return (0);
}

Java

import java.util.*;

/* Game Description-
"A game is played between two players and there are N piles
of stones such that each pile has certain number of stones.
On his/her turn, a player selects a pile and can take any
non-zero number of stones upto 3 (i.e- 1,2,3)
The player who cannot move is considered to lose the game
(i.e., one who take the last stone is the winner).
Can you find which player wins the game if both players play
optimally (they don't make any mistake)? "

A Dynamic Programming approach to calculate Grundy Number
and Mex and find the Winner using Sprague - Grundy Theorem. */

class GFG {


/* piles[] -> Array having the initial count of stones/coins
            in each piles before the game has started.
n     -> Number of piles

Grundy[] -> Array having the Grundy Number corresponding to
            the initial position of each piles in the game

The piles[] and Grundy[] are having 0-based indexing*/

static int PLAYER1 = 1;
static int PLAYER2 = 2;

// A Function to calculate Mex of all the values in that set
static int calculateMex(HashSet<Integer> Set)
{
    int Mex = 0;

    while (Set.contains(Mex))
        Mex++;

    return (Mex);
}

// A function to Compute Grundy Number of 'n'
static int calculateGrundy(int n, int Grundy[])
{
    Grundy[0] = 0;
    Grundy[1] = 1;
    Grundy[2] = 2;
    Grundy[3] = 3;

    if (Grundy[n] != -1)
        return (Grundy[n]);

    // A Hash Table
    HashSet<Integer> Set = new HashSet<Integer>();

    for (int i = 1; i <= 3; i++)
            Set.add(calculateGrundy (n - i, Grundy));

    // Store the result
    Grundy[n] = calculateMex (Set);

    return (Grundy[n]);
}

// A function to declare the winner of the game
static void declareWinner(int whoseTurn, int piles[],
                    int Grundy[], int n)
{
    int xorValue = Grundy[piles[0]];

    for (int i = 1; i <= n - 1; i++)
        xorValue = xorValue ^ Grundy[piles[i]];

    if (xorValue != 0)
    {
        if (whoseTurn == PLAYER1)
            System.out.printf("Player 1 will win\n");
        else
            System.out.printf("Player 2 will win\n");
    }
    else
    {
        if (whoseTurn == PLAYER1)
            System.out.printf("Player 2 will win\n");
        else
            System.out.printf("Player 1 will win\n");
    }

    return;
}


// Driver code
public static void main(String[] args)
{

    // Test Case 1
    int piles[] = {3, 4, 5};
    int n = piles.length;

    // Find the maximum element
    int maximum = Arrays.stream(piles).max().getAsInt();

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int Grundy[] = new int[maximum + 1];
    Arrays.fill(Grundy, -1);

    // Calculate Grundy Value of piles[i] and store it
    for (int i = 0; i <= n - 1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER1, piles, Grundy, n);

    /* Test Case 2
    int piles[] = {3, 8, 2};
    int n = sizeof(piles)/sizeof(piles[0]);


    int maximum = *max_element (piles, piles + n);

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int Grundy [maximum + 1];
    memset(Grundy, -1, sizeof (Grundy));

    // Calculate Grundy Value of piles[i] and store it
    for (int i=0; i<=n-1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER2, piles, Grundy, n); */

    }
}

// This code is contributed by PrinciRaj1992

Python3

'''  Game Description-
 "A game is played between two players and there are N piles
 of stones such that each pile has certain number of stones.
 On his/her turn, a player selects a pile and can take any
 non-zero number of stones upto 3 (i.e- 1,2,3)
 The player who cannot move is considered to lose the game
 (i.e., one who take the last stone is the winner).
 Can you find which player wins the game if both players play
 optimally (they don't make any mistake)? "

 A Dynamic Programming approach to calculate Grundy Number
 and Mex and find the Winner using Sprague - Grundy Theorem.

     piles[] -> Array having the initial count of stones/coins
            in each piles before the game has started.
   n       -> Number of piles

   Grundy[] -> Array having the Grundy Number corresponding to
             the initial position of each piles in the game

   The piles[] and Grundy[] are having 0-based indexing'''

PLAYER1 = 1
PLAYER2 = 2

# A Function to calculate Mex of all
# the values in that set
def calculateMex(Set):

    Mex = 0;

    while (Mex in Set):
        Mex += 1

    return (Mex)

# A function to Compute Grundy Number of 'n'
def calculateGrundy(n, Grundy):

    Grundy[0] = 0
    Grundy[1] = 1
    Grundy[2] = 2
    Grundy[3] = 3

    if (Grundy[n] != -1):
        return (Grundy[n])

    # A Hash Table
    Set = set()

    for i in range(1, 4):
        Set.add(calculateGrundy(n - i,
                                Grundy))

    # Store the result
    Grundy[n] = calculateMex(Set)

    return (Grundy[n])

# A function to declare the winner of the game
def declareWinner(whoseTurn, piles, Grundy, n):

    xorValue = Grundy[piles[0]];

    for i in range(1, n):
        xorValue = (xorValue ^
                    Grundy[piles[i]])

    if (xorValue != 0):

        if (whoseTurn == PLAYER1):
            print("Player 1 will win\n");
        else:
            print("Player 2 will win\n");
    else:

        if (whoseTurn == PLAYER1):
            print("Player 2 will win\n");
        else:
            print("Player 1 will win\n");

# Driver code
if __name__=="__main__":

    # Test Case 1
    piles = [ 3, 4, 5 ]
    n = len(piles)

    # Find the maximum element
    maximum = max(piles)

    # An array to cache the sub-problems so that
    # re-computation of same sub-problems is avoided
    Grundy = [-1 for i in range(maximum + 1)];

    # Calculate Grundy Value of piles[i] and store it
    for i in range(n):
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER1, piles, Grundy, n);

    ''' Test Case 2
    int piles[] = {3, 8, 2};
    int n = sizeof(piles)/sizeof(piles[0]);


    int maximum = *max_element (piles, piles + n);

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int Grundy [maximum + 1];
    memset(Grundy, -1, sizeof (Grundy));

    // Calculate Grundy Value of piles[i] and store it
    for (int i=0; i<=n-1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER2, piles, Grundy, n);   '''

# This code is contributed by rutvik_56

C#

using System;
using System.Linq;
using System.Collections.Generic;

/* Game Description-
"A game is played between two players and there are N piles
of stones such that each pile has certain number of stones.
On his/her turn, a player selects a pile and can take any
non-zero number of stones upto 3 (i.e- 1,2,3)
The player who cannot move is considered to lose the game
(i.e., one who take the last stone is the winner).
Can you find which player wins the game if both players play
optimally (they don't make any mistake)? "

A Dynamic Programming approach to calculate Grundy Number
and Mex and find the Winner using Sprague - Grundy Theorem. */

class GFG
{


/* piles[] -> Array having the initial count of stones/coins
            in each piles before the game has started.
n -> Number of piles

Grundy[] -> Array having the Grundy Number corresponding to
            the initial position of each piles in the game

The piles[] and Grundy[] are having 0-based indexing*/

static int PLAYER1 = 1;
//static int PLAYER2 = 2;

// A Function to calculate Mex of all the values in that set
static int calculateMex(HashSet<int> Set)
{
    int Mex = 0;

    while (Set.Contains(Mex))
        Mex++;

    return (Mex);
}

// A function to Compute Grundy Number of 'n'
static int calculateGrundy(int n, int []Grundy)
{
    Grundy[0] = 0;
    Grundy[1] = 1;
    Grundy[2] = 2;
    Grundy[3] = 3;

    if (Grundy[n] != -1)
        return (Grundy[n]);

    // A Hash Table
    HashSet<int> Set = new HashSet<int>();

    for (int i = 1; i <= 3; i++)
            Set.Add(calculateGrundy (n - i, Grundy));

    // Store the result
    Grundy[n] = calculateMex (Set);

    return (Grundy[n]);
}

// A function to declare the winner of the game
static void declareWinner(int whoseTurn, int []piles,
                    int []Grundy, int n)
{
    int xorValue = Grundy[piles[0]];

    for (int i = 1; i <= n - 1; i++)
        xorValue = xorValue ^ Grundy[piles[i]];

    if (xorValue != 0)
    {
        if (whoseTurn == PLAYER1)
            Console.Write("Player 1 will win\n");
        else
            Console.Write("Player 2 will win\n");
    }
    else
    {
        if (whoseTurn == PLAYER1)
            Console.Write("Player 2 will win\n");
        else
            Console.Write("Player 1 will win\n");
    }

    return;
}


// Driver code
static void Main()
{

    // Test Case 1
    int []piles = {3, 4, 5};
    int n = piles.Length;

    // Find the maximum element
    int maximum = piles.Max();

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int []Grundy = new int[maximum + 1];
    Array.Fill(Grundy, -1);

    // Calculate Grundy Value of piles[i] and store it
    for (int i = 0; i <= n - 1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER1, piles, Grundy, n);

    /* Test Case 2
    int piles[] = {3, 8, 2};
    int n = sizeof(piles)/sizeof(piles[0]);


    int maximum = *max_element (piles, piles + n);

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int Grundy [maximum + 1];
    memset(Grundy, -1, sizeof (Grundy));

    // Calculate Grundy Value of piles[i] and store it
    for (int i=0; i<=n-1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER2, piles, Grundy, n); */

    }
}

// This code is contributed by mits

Javascript

<script>

/* Game Description-
"A game is played between two players and there are N piles
of stones such that each pile has certain number of stones.
On his/her turn, a player selects a pile and can take any
non-zero number of stones upto 3 (i.e- 1,2,3)
The player who cannot move is considered to lose the game
(i.e., one who take the last stone is the winner).
Can you find which player wins the game if both players play
optimally (they don't make any mistake)? "

A Dynamic Programming approach to calculate Grundy Number
and Mex and find the Winner using Sprague - Grundy Theorem. */


/* piles[] -> Array having the initial count of stones/coins
            in each piles before the game has started.
n     -> Number of piles

Grundy[] -> Array having the Grundy Number corresponding to
            the initial position of each piles in the game

The piles[] and Grundy[] are having 0-based indexing*/
let PLAYER1 = 1;
let PLAYER2 = 2;

// A Function to calculate Mex of all the values in that set
function calculateMex(Set)
{
    let Mex = 0;

    while (Set.has(Mex))
        Mex++;

    return (Mex);
}

// A function to Compute Grundy Number of 'n'
function calculateGrundy(n,Grundy)
{
    Grundy[0] = 0;
    Grundy[1] = 1;
    Grundy[2] = 2;
    Grundy[3] = 3;

    if (Grundy[n] != -1)
        return (Grundy[n]);

    // A Hash Table
    let Set = new Set();

    for (let i = 1; i <= 3; i++)
            Set.add(calculateGrundy (n - i, Grundy));

    // Store the result
    Grundy[n] = calculateMex (Set);

    return (Grundy[n]);
}

// A function to declare the winner of the game
function declareWinner(whoseTurn,piles,Grundy,n)
{
    let xorValue = Grundy[piles[0]];

    for (let i = 1; i <= n - 1; i++)
        xorValue = xorValue ^ Grundy[piles[i]];

    if (xorValue != 0)
    {
        if (whoseTurn == PLAYER1)
            document.write("Player 1 will win<br>");
        else
            document.write("Player 2 will win<br>");
    }
    else
    {
        if (whoseTurn == PLAYER1)
            document.write("Player 2 will win<br>");
        else
            document.write("Player 1 will win<br>");
    }

    return;
}

// Driver code

// Test Case 1
    let piles = [3, 4, 5];
    let n = piles.length;


    // Find the maximum element
    let maximum = Math.max(...piles)

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    let Grundy = new Array(maximum + 1);
    for(let i=0;i<maximum+1;i++)
        Grundy[i]=0;

    // Calculate Grundy Value of piles[i] and store it
    for (let i = 0; i <= n - 1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER1, piles, Grundy, n);

    /* Test Case 2
    int piles[] = {3, 8, 2};
    int n = sizeof(piles)/sizeof(piles[0]);


    int maximum = *max_element (piles, piles + n);

    // An array to cache the sub-problems so that
    // re-computation of same sub-problems is avoided
    int Grundy [maximum + 1];
    memset(Grundy, -1, sizeof (Grundy));

    // Calculate Grundy Value of piles[i] and store it
    for (int i=0; i<=n-1; i++)
        calculateGrundy(piles[i], Grundy);

    declareWinner(PLAYER2, piles, Grundy, n); */

// This code is contributed by avanitrachhadiya2155
</script>

Output :

Player 1 will win

Time complexity : O(n^2), where n is the maximum number of stones in a pile. 

Space complexity :O(n), as the Grundy array is used to store the results of subproblems to avoid redundant computations and it takes O(n) space.

References :
https://en.wikipedia.org/wiki/Sprague%E2%80%93Grundy_theorem

Exercise to the Readers: Consider the below game. 
“A game is played by two players with N integers A1, A2, .., AN. On his/her turn, a player selects an integer, divides it by 2, 3, or 6, and then takes the floor. If the integer becomes 0, it is removed. The last player to move wins. Which player wins the game if both players play optimally?”
Hint : See the example 3 of previous article.

If you like GeeksforGeeks and would like to contribute, you can also write an article and mail your article to review-team@geeksforgeeks.org. See your article appearing on the GeeksforGeeks main page and help other Geeks.
Please write comments if you find anything incorrect, or you want to share more information about the topic discussed above