[ngould]

I think your argument is logically correct, but you are using numerical assumptions that are off by one or two orders of magnitude. Your typical successful algorithmic trader is probably flipping their metaphorical coin 1,000,000 times, and getting 520,000 heads. Each individual trade may only be slightly profitable, but there is often no statistical ambiguity about the effectiveness of the strategy.

Individual trading strategies often become less effective over time, though. Whether this kind of success can be sustained at the level of a trading firm over many years is an entirely different question. Whether they can beat the market after fees is a third, also entirely different question.