 # How many possible chess games are there?

Most possible chess games are totally random. For example, say in every chess position on average there are 3 good moves, on average there are around 23 possible moves!

We can come up with an estimate by writing a program to compute the average number of legal moves, and the average game length when playing randomly.

So lets do it! I'm going to be using the python-chess library, while writing this in a language like rust would be far more efficent, we really only need to run this once, so python will be good enough.

First we'll need `chess` and `random`

``````import chess
import random``````

And a function that returns the length of a random game, and the average amount of legal moves.

``````def play_random_game() -> (int, int):
b = chess.Board()
total_legal_moves = 0
while not b.is_game_over():
legal_moves = list(b.legal_moves)
total_legal_moves += len(legal_moves)
b.push(random.choice(legal_moves))

game_length = len(b.move_stack)
return game_length, total_legal_moves / game_length``````

Then we can play a few thousand random games (this takes forever, python is slow!)

``````avg_length = 0
avg_legal  = 0
n_games = 10000
for i in range(n_games):
length, moves = play_random_game()
avg_length += length
avg_legal += moves
print(f'[{i}] length: {length} moves: {moves}')

avg_length /= n_games
avg_legal /= n_games

print(f'avg_length: {avg_length} avg_legal: {avg_legal}')``````

For me this gives `avg_length: 355.9369 avg_legal: 23.09636004565067` thus, there are approximately `23 ^ 355` possible chess games.

In future I might rewrite this in rust and average a larger amount of games, but for now this should be a decent estimate.