The so-called ''gambler's fallacy'' asserts that gamblers believe mech
anical randomizers (such as roulette) exhibit sequential tendencies. I
n this paper we introduce another false belief of gamblers, namely the
tendency to bet on a winning number under the assumption that the rou
lette wheel is biased. We discuss the proper analysis for detecting bi
ased numbers and demonstrate empirically that people notoriously under
estimate the number of observations needed for a reliable detection of
biased numbers. We label this bias as the Type II gambler's fallacy a
nd suggest that both types of the fallacy are due to misperceptions of
randomness. Implications for real life situations are briefly discuss
ed. (C) 1994 Academic Press, Inc.