The theory of large numbers states that the number of data points is directly correlated to the odds of the distribution equalling the equation of the curve. In simple terms, the more bets placed, the greater the odds that total return will equal the average. In the roulette example, if 1000 spins occurred, there is a greater chance that the total return would equal 94.7% than if only ten spins were made.
The lesson here is the lower number of plays or spins create a greater chance for the net results to be outside the expected result of 94.7%, both to the positive and negative sides. So, if you want to take a chance at winning big and losing big, then bet a low number of times on high odds games. For those concerned with taken a calculated gamble, spin 10 times and bet a single number. The chances of winning are 1 in 38 and the payout is $36 for every $1 bet.
In order to have a 50% chance to win on one number, 19 spins would need to be made (50%=19*1/38). Assuming that the other bets are losers, the winnings would total (36:1 odds * $10 bet - 9 losses * $10 bet) or $530. The limit on losses is $190. From a psychological perspective, most people would be willing to bet $190 to have a 50% chance to win $530.
However, if we increase the number of chances (the theory of large numbers), the loss limit changes and the payout gets closer to the average. At 1000 spins on a $10 bet, the loss limit is $10.000. On 1000 spins, the average number of winning number is 26.3, which at odds or 36:1 translates to winnings of $9473. Not many people would risk $10.000 to get back $9473, for a cumulative loss of $527. Bottom line, the more you spin, the greater the chance to lose.