Don't get me wrong, my method's are 100% stats based. Stats aren't everything, but they're enough, at least for most sports.
The danger lies in interpreting them. To make good decisions, what you're really interested in is the most likely future stats, which is basically the same as the particular skill level in question of the player/team/defence/whatever. But of course, all you have is past stats. One of the biggest mistakes you can make is to assume that past stats are an adequate indicator of true skill level. Or that a small sample of recent stats are a better indicator than a large sample of past stats (pitcher's hot last 5 games). Or that a small selective sample is a better indicator than a larger broad sample (this guy crushes the ball in April).
Sometimes those assumptions are true. Usually they are not. You really need to at least know how predictive the stat is (google "regression to the mean") and how big of a sample size you're looking at. For larger sample sizes, you also would like to know how predictive past vs recent performance is (decay functions) and how aging affects performance, but you can actually get by reasonably well without reaching that level.
Most unsophisticated bettors err way on the side of assigning too much predictive value to common stats and too much value to recent performance and other small samples.
That's why looking at common stats without a deeper understanding of what you're doing will often put you soundly on the wrong side of a game.
