This is an excellent graphic, but the author has a strong stance against HFT (such as the pages he links to about high frequency quoting) that readers should moderate with an understanding of what HFT does in an ideal economic context--the "purpose" of HFT.<p>At the level of HFT, the decisions being made are too fast and frequent to be based on the fundamentals of the stocks being traded. Instead, algorithms are used to make trades based on "technical" analysis rather than analysis of fundamentals--technical analysis predicts how stocks will move based only on past movement, not on underlying reasons.<p>In some cases, this movement is movement that happened only seconds, or milliseconds earlier. In a market with no HFT, say that there's a stock where a large sale is made by some investor, driving price quotes down a few cents; and then a second later, another investor buys a large share, pushing prices up again to where they were before. This behavior makes stock charts jagged and gives a certain volatility to prices.<p>Now, imagine that you could algorithmically predict when a price is likely to go down and then up again in cases like that, and so you could make a profit by buying low knowing that the stock price will be back up again one second into the future. You buy stock to capitalize on this, and if you buy as much stock as you profitably can, you'll essentially drive the price back up to where it will be after the original "correction". You end up correcting the price <i>before</i> the other investor does; and if your response time to the first investor's sale is fast enough, the time when the stock is underpriced will be reduced by you quickly buying.<p>This stabilizes prices, smoothing out the curves, and stable prices are valued in the market. People are willing to pay money for stability, even at those small levels; and that amount of money is what HFT is designed to capture. HFT also makes a benefit by being "first" on finding the right price through its algorithmic clairvoyance.<p>Sometimes, obviously, the use of these algos goes wrong. You get a flash crash with a death by ten billion cuts as the market moves to smooth in some incorrect way. But HFT is not all bad, and considering how much there is, it's impressive that it works as well as it usually does. The incidents you hear about are just like plane crashes; just because planes crash sometimes and it's big news doesn't mean that air travel isn't safer per mile than car travel. Similarly, HFT has a powerful smoothing ability that it capitalizes on (among other things that it does--but smoothing is one of its largest abilities), and most people who oppose it don't realize how essential price stability is to the market at those levels given how much non-HFT trading volume goes on these days.