I am Eric Hunsader from Nanex and created this animation using our own custom software tools and our NxCore data feed.

Our position on HFT can be summed up in the first lines of text below the image: "It's not high frequency trading (HFT) that concerns us. It's high frequency quoting". Links are included for details. Our latest paper on HFT can be found here: http://www.nanex.net/aqck2/3532.html

Anyone who has taken the time to really look at the contents of the tsunami of data that some HFT creates will come to the conclusion that something very wrong and harmful going on. We've found this is nearly impossible to convey in text or graphics (with the possible exception of this animation), and sooner or later as more and more academics get up to speed, they will generally come to the same conclusions we have.

Our primary business is providing a real-time (and historic ) data service for U.S. Stocks, Options and Futures. In the course of monitoring our feed for our subscribers, we run across anomalies that we think need to be published for the public good or long term health of our markets. We have not received a dime from any of our analysis and not a week goes by that we don't regret opening that Pandora's box.

I will check back later today to answer any questions. If you don't receive a reply for a few hours, it's because we are working at our "day job".

I'm wondering why there tends to be so much value judgment and moralizing in this area coming from various sources (such as Themis).

For example: HF traders are taking value from "real" traders ("real" defined by whom?), the systems are "too fast" (compared to what?) or quoting "too much" (compared to what?), and so forth.

Exchanges already seem to be penalizing participants with very high quote-to-trade ratios, suggesting that the existing regulatory and commercial system is responding to the needs of its stakeholders.

Maybe this kind of reaction is inevitable. I'm sure that there was moralizing in ancient Greece when some entrepreneur bought a load of olive oil at a low price in Athens and ran it on a fast chariot to Thessaloniki, unfairly undercutting the honest merchants of Thessaloniki and pocketing a tidy profit through their "high-speed trading" … same thing with those who used an undersea New York - London cable to gain advance knowledge of events.

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.

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.

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 before 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.

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.

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.

I learned recently that there have been many similar crashes throughout history, well before even the invention of computers eg

http://en.wikipedia.org/wiki/Wall_Street_Crash_of_1929

http://en.wikipedia.org/wiki/Black_Monday_%281987%29

http://en.wikipedia.org/wiki/Friday_the_13th_mini-crash

http://en.wikipedia.org/wiki/October_27,_1997_mini-crash

It seems to be more a matter of market dynamics rather than crazy algos. If some market shock drops prices too quickly you pass from a stable equilibrium into panic selling. Market makers will only accept so much risk before they start trying to offload stocks too, leading to a feedback loop. Lots of exchanges now have measures in place to break this feedback loop by halting trading on a stock if the price moves too fast and running an auction instead.

The 2010 flash crash was really only notable in that the move to electronic trading made both the crash and the subsequent recovery much faster. Panic selling itself is a problem with the market structure, not with the machines on the other end.