(Fortune) -- Theories about yesterday's stock market swoon, where within a matter of 20 minutes, the stock market plunged by 1,000 points and then nearly completely recovered, are abounding. Fortune asked Rishi Narang, founder of the hedge fund Telesis Capital and author of Inside the Black Box, to share the theories he's heard and handicap them in terms of likelihood and plausibility.
Narang, who uses high-frequency trading techniques, explains why high-frequency traders got out of the market during the dive, and why the catalyst for the drop is far more important to understand than the drop itself:
What happened yesterday?
There are two points to understand. First, what catalyzed the activity? What was the reason for the market wanting to fall? It might be that the catalyst was of such size that it overwhelmed all other factors. There are three plausible theories:
1) The fat finger. Plausible, but unlikely. Typing in billions with a "b" versus millions with an "m" seems impossible. Trading systems don't work that way. More likely, the trading system accepts the sell/buy amount in thousands. Some trader in the heat of the moment forgets it's in thousands, types in an order for 16,000,000 instead of 16,000. That kind of thing seems far more plausible.
But even then: why on Earth would the trading entry system not have a sanity check? For almost no one in the world is a $16 billion sell order okay to send out as soon as it's entered. The trader should be fired, along with everyone in the IT department. If this happened, most likely, it was something along those lines. If it wasn't all one order, maybe it was meant to sell just $1 billion shares but was sent 3 or 5 times instead of once.
2) Software error. Plausible, likely, but doesn't fit the facts. Here, the trading software is in a recursive loop, pounding out sell orders due to a bug somewhere in the software. In a sense, this is more plausible, more likely, but doesn't seem to fit the facts well enough.
The speed of the decline in the market just doesn't seem to fit -- should be a series of small orders, not a series of large orders. In 7 minutes we saw a 580-point drop. That doesn't look like a recursive loop. But there is a lot of software, and somewhere a bug is bound to exist. You can easily imagine a software glitch happening. Things go buggy. Like the Toyota [accelerator] problem, at heart a software problem. Technology is a two-edged sword, and this is the other edge of the sword. We rely on software, but it's not always written well enough.
3) Computer hacking. Implausible without proof, but possible. This is the most interesting theory because we know terrorists are interested in cyberterrorism. We know they would target the financial markets. We know a great day to launch an attack would be one with a mild bit of panic [due to the Greek crisis and sovereign debt downgrades].
Some other really crazy things happened with stocks, like Accenture and Exelon. [Both stocks traded for one cent for short periods of time.] Two parties really transacted on these trades [at one cent], even though they were later busted and cancelled. If it was just high-frequency traders bailing out, why wouldn't [that drop] happen on every stock? It just doesn't add up. Things are too idiosyncratic and that feels uncomfortable. This also happened in the options markets, but again, only on a handful of options.
And the second point to understand?
That's the question of the enabler. What, if anything perpetuated the selloff? And did so in seconds? There's a lot of speculation about high-frequency traders vanishing from the marketplace.
The consensus is that high-frequency guys didn't provide the liquidity and that's what allowed for prices like one penny on Accenture. I do know for sure that high-frequency traders backed off, but old school market makers would've done the same thing, in a little bit different way. They just would've created super-wide market spreads. Same thing.
We shouldn't be so sanguine about taxes and impediments to high-frequency trading if we are upset when high-frequency traders leave the market. Those are incompatible ideas.
As a side point: traders have stop loss levels; one big move triggers other moves. There are systematic, discretionary, and plain-old panic trades.
But for all of those styles and programs, once they see the stock market fall 6%, a liquidation effect takes hold. That's just a function of people. Someone screams fire, and if enough people start running, everyone will. Those are the dynamics of computer software, people, animals, fires, whatever. It's how we work. That kind of stampeding effect could easily be part of the response.
But the speed of the market falling down, going back up, and partway back down again? If this was really a stampede, why not repeat the 1987 crash [which kept going]? Nothing 'stopped' this crash except that the catalyst seemed to have ended.
If it was an error or a software bug, it stopped. If it was a hack, the hackers left. In other words, the enabling side of this drop is totally irrelevant [to the catalyst]. The only interesting thing here is the catalyst. If this was a gas pedal that was stuck, it would've looked differently, kept going.
Whether this was intentional or unintentional, it happened all at once. If it was an intentional [attack], then the question is, was it a demonstration, a test, or the attack itself? Whatever it was, we didn't stop it. It stopped itself.