I hope everyone can agree that if there is a real statisitically significant pattern in the data and if someone can recognize it when others do not, then they can beat the market in a statistically significant way, not just in the sense of "well there're lots of participants so some must win."

So, are there real patterns in financial markets?  To me the answer is clearly yes for these reasons:

• Historically there have been many patterns. They existed due to real-world factors like financiers liking to sell off some of their assets before going on holiday, like certain international events making people think more optimistically about life in general, like corporation taxes being levied at particular times of the year, like breweries being anti-correlated with GDP and so on and so on.  Shortly after they are identified publicly they disappear, because the market starts to predict them ahead of time, hence smoothing them out.  So how come you can look back and see all those patterns historically and yet you look now and there doesn't seem to be any patterns?  Well it could be because there are in fact no more patterns because they've all been discovered, but much more likely it's because there are patterns that exist now but they have either not yet been discovered or have not yet been made public.  In ten years time, today's patterns will have been revealed and made public...I wonder if in ten years time people will still be saying "but now is different, now I can quite surely say that there are no more patterns because they have all been discovered".
• With the (possible) exception of modern algo trading, trades are done by humans.  Humans have emotions, they trade partly on sentiment, they make flawed analyses. They buy because they want to impress the hot sales girl.  They sell because the risk-manager is breathing down their neck.  They double up when they get drunk.  Humans cannot even generate random numbers if they try...how can one plausibly claim that a price series generated by humans is perfectly random when humans don't seem capable of generating a random series of numbers at all?
• Just because a trader cannot stay ahead of the pack indefinitely, that does not mean that he never spotted a real (statistically significant) pattern.  It would be like saying there's no evidence Pele was a good footballer because he currently cannot get into the Brazilian team. Maybe one day I'll spot something that nobody noticed about company X.  I'll trade it, make some money and that would be it.  I wouldn't become famous, I wouldn't get a nobel prize, maybe I would never spot any pattern again, maybe I would never trade again. But that doesn't mean that my trade was random, it just means that it would never be picked up and given as evidence for the existence of patterns.  Remember, there's a world of difference between on the one hand patterns actually existing and on the other being able to statistically prove that patterns exist given public data.

So I hope I've given pretty strong arguments that patterns exist, and that some people can use them to make money, even if you can't necessarily prove it using statistical means and even if it gives you no indication of what those patterns actually are.