What good would that do? Suppose we're looking at a world with one stock. Suppose I have 100 hypotheses about that stock, each of which is relevant to some ti

What good would that do? Suppose we're looking at a world with one stock. Suppose I have 100 hypotheses about that stock, each of which is relevant to some time frame. Some are hypotheses about the next few seconds, some are hypotheses about the next few minutes, hours, days, etc. What is the harm of trying to execute trades to profit from those hypotheses? If there is no counterparty, none of the trades will go through regardless of their timeframe. If there is a counterparty who decides (for example) to participate in several transactions with me over the span of several minutes (or seconds), we should assume that counterparty is self-interested and is making trades that satisfy its goals. If on the other hand the counterparty is a firm whose algorithms and trading infrastructure was build in the 1990s and is somehow tripped up by my strategy, how is that any different from a lay investor being tripped up by a misinterpretation of (say) a low end of day closing price? All market participants are either speculative or based on fundamentals. If you're the latter, you don't care much about the ultra short-term. If you're the former, then you have some strategy for speculation, which may take as input the behavior of other market participants. In my view, since there is no reason why a rational strategy creator would prefer long term strategies to short term ones (expected risk adjusted returns being equal), there is no reason why HFT should be viewed with disdain. Much as computers replaced rotary phones on the trading floor, HFT is the next iteration of technological improvement in trading. HFT happens to be a specialty area at present, but there's no reason that any adept speculator would avoid high frequency trades if she had sufficient algorithmic and tech sophistication to make them.