A belief that independent samples somehow relate to one-another. “We haven’t rolled a 7 in forever, so this one is more likely to be a 7.”
Gambler’s Fallacy is when people think the likelihood of a given result for an independent trial is affected by the results of previous trials. That is to say, you expect the probability curve of describing the distribution of expected results is dynamic in accordance with previous results. By definition, independent events are independent. If you draw a marble from a bag containing 3 blue marbles and 1 red marble, your chance of getting the red marble is always 25% (obviously, this assumes you replace whatever marble you drew back in the bag after having drawn it out).
Note - for dependent events, such as drawing colored marbles from a bag one by one without replacement, the probability of each successive draw really is a function of all the previous draws.