I estimate off the top of my head, without getting to nitty-gritty, I played at least, no less than 25,500 shoes of baccarat since I started playing. And, I would have to say, without 100% written records--no more than 10% in a haphazard accumulation of shoes within that 25,500 even came close to any statistics in most ways.
I think any casino would be seriously worried if after so many shoes only 10% "came close" to the predicted statistics. Of course it depends what you mean by "close", but as a counter-example, here are the results of the Wizard of Odds 1000 6 and 8 deck shoes in terms of B, P, and T percentages:
6 Deck Shoes:
Banker Wins = 28708 (45.8418)
Player Wins = 28038 (44.772%)
Tie Wins = 5878 (9.38618%)
8 Deck Shoes:
Total player wins = 36193, ratio = 0.447496
Total banker wins = 36887, ratio = 0.456076
Total tie wins = 7799, ratio = 0.096428
By any standard, these stats are close to the theoretical prediction, and after "only" 1000 shoes. Sure, you can point to this or that shoe and say it's far off the theoretical stats -- say 35% B and 65% P -- but these by definition can't be typical shoes, otherwise the averages would not be what they are. By the Normal distribution approximately 68% of shoes will be "close" to the predicted averages; a long way from 10%.