Previous studies have illustrated human misperceptions of randomness and resultant suboptimal decision making with reference to the "hot hand" or momentum effect in sport, the notion of serial dependency between outcomes. However, issues of omitted variables bias have plagued many due to a historical reliance on nonparametric techniques or basic regression models. This paper examines across-game and within-game momentum in the English Premier League football competition using fixed effects regressions to control for time-invariant heterogeneity in conjunction with traditional nonparametric techniques. Although the results show evidence of performance reversal following winning streaks, no such evidence is found for streaks of draws or losses or in goal scoring performance within games. This suggests momentum is better suited as a post hoc label of performance than a robust causal phenomenon.