Problem 121: Disc game prize fund

(see projecteuler.net/problem=121)

A bag contains one red disc and one blue disc. In a game of chance a player takes a disc at random and its colour is noted.
After each turn the disc is returned to the bag, an extra red disc is added, and another disc is taken at random.

The player pays £1 to play and wins if they have taken more blue discs than red discs at the end of the game.

If the game is played for four turns, the probability of a player winning is exactly 11/120, and so the maximum prize fund the banker
should allocate for winning in this game would be £10 before they would expect to incur a loss.
Note that any payout will be a whole number of pounds and also includes the original £1 paid to play the game, so in the example given the player actually wins £9.

Find the maximum prize fund that should be allocated to a single game in which fifteen turns are played.

Algorithm

I compute all possible states after each round.
Before the first round, the player had obviously no red and no blue disc. This State was the only possibility.
In each round, the number of possible combinations are possibilities_n = possibilities_{n-1} * (red + blue) because the player picks any disc at random.
Several choices lead to the same next state, e.g. it doesn't matter which red disc he/she picks.
Therefore the total number of states increases by one per round but the possibilities explode in a factorial way.

The final prize is \lfloor dfrac{sum{possibilities}}{sum{possibilities(blue>red)}} \rfloor, e.g. \lfloor dfrac{120}{11} \rfloor = 10 → £10 payout.

Modifications by HackerRank

Solving the last test case requires more than 40 rounds. That's beyond the reach of unsigned long long and/or double.
19 rounds are fine with my current implemenation. Switching to double works surprisingly well up to 30 rounds.

Note

My states don't need to store red and blue because they can be derived from the index of the State object in the open container.
I kept those member variables for clarity, though.

My code

… was written in C++11 and can be compiled with G++, Clang++, Visual C++. You can download it, too.

#include <iostream>
#include <vector>
 
typedef unsigned long long Possibilities; // replace unsigned long long by double to compute up to 30 rounds
 
struct State
{
// picked red and blue disc
unsigned int red;
unsigned int blue;
// number of different choices leading to this state
Possibilities possibilities;
};
 
int main()
{
unsigned int tests = 1;
std::cin >> tests;
 
while (tests--)
{
unsigned int maxRounds = 15;
std::cin >> maxRounds;
 
// number of red and blue discs in the bag
unsigned int availableRed = 1;
unsigned int availableBlue = 1;
Possibilities possibilities = 1;
 
// initially no disc taken yet
std::vector<State> open = { { 0, 0, 1 } };
 
// evaluateall 15 rounds
for (unsigned int round = 1; round <= maxRounds; round++)
{
// total number of possible choices
possibilities *= availableRed + availableBlue;
 
// set number of blue and red discs per state
std::vector<State> next;
for (unsigned int blue = 0; blue <= round; blue++)
{
State state;
state.red = round - blue;
state.blue = blue;
state.possibilities = 0;
next.push_back(state);
}
 
// update possibilities of each state
for (auto current : open)
{
// picking red
next[current.blue ].possibilities += current.possibilities * availableRed;
// picking blue
next[current.blue + 1].possibilities += current.possibilities * availableBlue;
}
 
// prepare next iteration
open = next;
availableRed++;
}
 
// compute chance that blue > red
Possibilities moreBlue = 0;
for (auto current : open)
if (current.blue > current.red)
moreBlue += current.possibilities;
 
// chance of winning is moreBlue / possibilities, therefore maximum prize is the inverse
unsigned int prize = possibilities / moreBlue; // integer division !
std::cout << prize << std::endl;
}
 
return 0;
}

This solution contains 13 empty lines, 13 comments and 2 preprocessor commands.

Interactive test

You can submit your own input to my program and it will be instantly processed at my server:

Number of test cases (1-5):

Input data (separated by spaces or newlines):
Note: number of rounds

This is equivalent to
echo "1 4" | ./121

Output:

(please click 'Go !')

Note: the original problem's input 15 cannot be entered
because just copying results is a soft skill reserved for idiots.

(this interactive test is still under development, computations will be aborted after one second)

Benchmark

The correct solution to the original Project Euler problem was found in less than 0.01 seconds on a Intel® Core™ i7-2600K CPU @ 3.40GHz.
(compiled for x86_64 / Linux, GCC flags: -O3 -march=native -fno-exceptions -fno-rtti -std=c++11 -DORIGINAL)

See here for a comparison of all solutions.

Note: interactive tests run on a weaker (=slower) computer. Some interactive tests are compiled without -DORIGINAL.

Changelog

May 19, 2017 submitted solution
May 19, 2017 added comments

Hackerrank

see https://www.hackerrank.com/contests/projecteuler/challenges/euler121

My code solved 2 out of 3 test cases (score: 50%)

I failed 1 test cases due to wrong answers and 0 because of timeouts

Difficulty

Project Euler ranks this problem at 35% (out of 100%).

Hackerrank describes this problem as easy.

Note:
Hackerrank has strict execution time limits (typically 2 seconds for C++ code) and often a much wider input range than the original problem.
In my opinion, Hackerrank's modified problems are usually a lot harder to solve. As a rule thumb: brute-force is never an option.

Links

projecteuler.net/thread=121 - the best forum on the subject (note: you have to submit the correct solution first)

Code in various languages:

Python: www.mathblog.dk/project-euler-121-coloured-discs/ (written by Kristian Edlund)
Java: github.com/nayuki/Project-Euler-solutions/blob/master/java/p121.java (written by Nayuki)
Scala: github.com/samskivert/euler-scala/blob/master/Euler121.scala (written by Michael Bayne)

Heatmap

green problems solve the original Project Euler problem and have a perfect score of 100% at Hackerrank, too.
yellow problems score less than 100% at Hackerrank (but still solve the original problem).
gray problems are already solved but I haven't published my solution yet.
blue problems are already solved and there wasn't a Hackerrank version of it (at the time I solved it) or I didn't care about it because it differed too much.

Please click on a problem's number to open my solution to that problem:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
The 133 solved problems had an average difficulty of 16.9% at Project Euler and I scored 11,174 points (out of 12300) at Hackerrank's Project Euler+.
more about me can be found on my homepage.
some names mentioned on this site may be trademarks of their respective owners.
thanks to the KaTeX team for their great typesetting library !