validate.py

import numpy as np

validators_online = 102
consensus_size = 28
replacement_factor = 4

validators = np.arange(validators_online)
replacement_N = len(validators) // replacement_factor

N = 10000

failed_elections = {}
for nr_elections in range(1, 10):
    failed_elections[nr_elections] = 0
    for i in range(N):
        trials = []
        elected_validators = np.random.choice(validators, replace=False, size=consensus_size)
        # Choose any node that isn't already a validator
        node_id = np.setdiff1d(validators, elected_validators)[0]

        for j in range(nr_elections):
            electable = np.setdiff1d(validators, elected_validators)
            elected = np.random.choice(electable, replace=False, size=replacement_N)
            elected_validators = np.concatenate((
                np.random.permutation(elected_validators)[replacement_N:],
                elected
            ))
            assert len(elected_validators) == consensus_size
            trials.append(node_id in elected_validators)

        if np.sum(trials) == 0:
            failed_elections[nr_elections] += 1


print('Validators online:', validators_online)
print('Consensus size:', consensus_size)
print('Replacement factor:', replacement_factor)
print()
print('Probability of not being selected to consensus group N elections in a row:')
for (elections, count) in failed_elections.items():
    print(elections, '->', f'{count / N * 100:.2f}%    [1 - p = {1 - (count / N):.2f}]')