Although it is possible for a few experts to successfully arbitrage the market for an unusually large return, this does not mean that a larger number could earn the same returns even given the same tools, 
techniques and data sources.
This is because the arbitrages are essentially drawn from a pool of finite size; although information about how to capture arbitrages is a nonrival good, the arbritrages themselves are a rival good. (To draw an analogy, the total amount of buried treasure on an island is the same, regardless of how many treasure hunters have bought copies of the treasure map.)
Retail traders are - almost by definition - undercapitalized. Thus they are subject to the problem of gambler's ruin. In a fair game (one with no information advantages) between two players that continues until one trader goes bankrupt, the player with the lower amount of capital has a higher probability of going bankrupt first.
Since the retail speculator is effectively playing against the market as a whole - which has nearly infinite capital - he will almost certainly go bankrupt.
The retail trader always pays the bid/ask spread which makes his odds of winning less than those of a fair game. Additional costs may include margin interest, or if a spot position is kept open for more than one day the trade may be "resettled" each day, each time costing the full bid/ask spread.
According to the Wall Street Journal (Currency Markets Draw Speculation, Fraud July 26, 2005) "Even people running the trading shops warn clients against trying to time the market. 'If 15% of day traders are profitable,' says Drew Niv, chief executive of FXCM, 'I'd be surprised.'
But I say to customers: if this is money you have worked hard for – that you cannot afford to lose – never, never invest in foreign exchange.