Hacker News Reader

Comment by lmm | Hacker News Reader

I've long thought there should be an efficient-markets solution to flash crashes. If the majority of sudden price drops are flash crashes and not fundamentals trading, then someone ought to be able to print money by simply resting a big bid 19% (or however low you can go without tripping the circuit breaker) below the current price, for everything, every day, and cashing in when the price crashes and rebounds. That would smooth out the crashes without any need for external intervention.

So why is no-one doing this?

View on HN

One problem with this strategy is that when things go crazy, traders with unsophisticated execution technology start to get terrible fills on their market orders; they complain to their brokers who complain to the exchange, who busts the trade. If you buy at the flash crash nadir, and sell halfway back to "normal" you're likely to have the buy cancelled, and be stuck with the sale.

>, and cashing in _when_ the price crashes and rebounds.

[my emphasis on "when"]

I think the "when" is a huge part of the puzzle. How can an algorithm know the difference between a flash crash that will get remedied within 24 hours vs a longer period of irrational pricing that takes months to sort out?

For example, at the time of the famous LTCM downfall[1], it held several arbitrage positions that were eventually proven right... but they didn't have sufficient capital reserves to survive the short term irrational spreads (Russian default, flight to US Treasuries, etc). I'm not saying that LTCM didn't have many other flaws of risk analysis that would have also caused their downfall but in that one case, it was an illustration of "the markets can remain irrational longer than you can remain solvent."

[1]http://www.amazon.com/When-Genius-Failed-Long-Term-Managemen...

GP is talking about a long-only flash-crash arb strategy -- so assuming you're not buying on margin you don't have the need to have capital reserves to "survive" until the recovery.

Of course, especially these days in the markets, whatever can be leveraged, will be leveraged...

They are doing this, but with some hesitation. Many "flash crashes" are caused deliberately to do just this, and the risk/reward of trying to game them is not worth it. In many cases, exchanges will roll back or invalidate any orders placed within 5 minutes of the flash crash, so the potential reward is not guaranteed. Any large orders placed immediately after the crash would be looked upon suspiciously.

Any company capable of exploiting a flash crash in the way you describe would need to be a timing arbitrage player -- and that's not a cheap business to be in. But exploiting short term dips in pricing is exactly what that type of HFT is designed for. It's incredibly profitable, but also zero-sum because there are only so many lucrative opportunities. So it's an arms race to have the fastest systems, the fastest connection to the exchanges so you can preempt orders, etc.

It's not really zero-sum in the sense that HFTs are usually criticized for; if it stabilizes the market for longer-term investors then that's a positive effect.

Market stabilization/liquidity is the effect that HFTs are supposed to have. If they didn't have a benefit they would have been banned long ago.

And arbitrage is always zero-sum: when your purchase of an asset drives the price of that asset up, there is a finite maximum on the number of shares you can buy without driving the price up so high as to not be profitable. There is also a finite number of shares out for bid at any given time; if you buy all of them then there aren't any for anyone else to buy.

I would expect it to be very hard to distinguish between a temporary crash and a long term price movement. Prices are moving all the time and often for good reason. This strategy would have bought up large amounts of Nokia stock shortly before they became an also-ran.

Id be more worried by a real bond price crash when the bank rate returns to normal as that what a lot of pensions are invested in.

A "flash crash" in equities is not the same as a "flash crash" in bonds.

"then someone ought to be able to print money"

Putting aside whether that'd be a good idea or not, government money is generated via the bond market. All new money in the US is generated via the bond market. Therefore, you can't use printing money as a way to stabilise the market that generates it.

I didn't mean literally print money. I meant make a lot of profit.

In any case I'm more thinking about equity markets - are there even flash crashes in treasuries yet?

Yes. http://www.bloomberg.com/news/articles/2015-03-30/bond-trade...

I won't say these are flash crashes, but it's very normal to see 6 or 7 sigma events in Treasuries. Such large deviations are very rarely seen in stocks

Reply on news.ycombinator.com