"Traditional money managers face the increased likelihood of seeing orders "gamed," or deliberately gouged. High-frequency funds have myriad strategies, but many depend on sniffing out "order flow" - or what hedge funds, mutual funds and pension funds that hold stocks for fundamental reasons are buying and selling."
In other words, everyone except the insiders get screwed, but the insiders depend on a steady flow of suckers, er I mean aspiring retirees, to skim money from.
Actually, it's pretty tough for an algo fund to make money when Aunt Harriet sells the 100 shares of GE she bought in 1962. The big, dumb mutual funds that turn over their portfolios in a year do get hit by this kind of stuff.
On the other hand, algo funds can't make money without 1) making market prices reflect the intentions of traders, rather than just their most recent actions, and 2) making the market more liquid. Both of these add a lot of good to the economy in general, even if they do it at the expense of one class of investors.
It's especially hard because Aunt Harriet's order almost certainly won't actually make it into the open market. Her broker will very likely "internalize" it (trade against it themselves), or sell it to someone else who will (e.g. Citadel will happily buy their flow.. they bought Ameritrade's last I checked).
That very well may be the future of these funds - compete for the customer or pay for order flow like in US Equity Options rather than compete in the open markets. All the orders that can be traded against with any of the typical profitable strategies (most strategies at these places are fairly well known and similar).. won't even make it to the real market. Doing it this way avoids the risks in the open market, notably adverse selection and the need to over-represent your interest in multiple places to increase your execution quality.
I strongly suspect the correlation between the rise of the Tech and Housing bubbles didn't have nearly so much to do with the rules or the bankers as it did with the enormous influx of 'retirement fund' money that desperately wanted somewhere to go.
With so many small-time, long-view investors putting blind trust into the hands of big-time, short-view fund managers/brokers/CEOs, the situation was inevitable.
And the only correction that will change things, is retirement money leaving the stock market and becoming a whole lot more skeptical. Plugging regulatory holes doesn't even help; as we've seen, effective regulation just gets stripped away as soon as it's politically feasible.
The Austrian theory of the business cycle explains why these bubbles occur. It states that when there is inflation via credit expansion that the sectors that attract that increased credit are in the higher orders or production that are more capital intensive. The inflated prices end up in areas like housing or the stock market. If you stop the inflation then you won't get any more bubbles assuming of course the theory is correct.
The popularity of these strategies has spawned a cottage industry called "co-location," or "proximity hosting." Exchanges sell the funds "rack space" in the data centers where their servers process trades to gain an extra couple of milliseconds on the competition.
Serious discussion of hosting your software on their servers started a couple years ago. I don't think anyone's done it yet, but people have at least built the technology out.
Of course, it's really stupid. Nobody that actually wants to own the stock or sell an investment cares about the difference of 50us. The competition at that level is at best inane. At worst, it makes it worse because it discourages people from trading real size and floods the world with all the little nearly meaningless quote updates.
There is a problem with this article: it's equating an algorithmic trade with an informed trade, and assuming that if most trades are algorithmic, most market moves are due to algorithms and not individual traders.
But those algorithms are basically looking for what future trades will be, based on current trades. For example, if someone invests $1 million, in $100,000 increments, they'll try to predict that $1 million investment. But the $100,000 increments are not the information the market ought to react to -- the final investment is (the trader, in this case, is not $100K more optimistic about the company during each increment; she's $1000K more optimistic, but is acting on this optimism in a more measured way).
Algo trading funds are betting on secondhand opinions. They aren't creating their own. The directional effects of this algo trading should be very hard to detect -- after all, if you can make a sensible prediction about what the market is going to do, you should probably be working for a hedge fund rather than writing (or commenting on) the Journal.
Indeed; given that a basic function of the stock market is price discovery, surely automated trading causes a steepening and undesirable drop in the signal-noise ratio, amplifying signals past the point of distortion?
From my signal-processing experience, it reminds me unpleasantly of the Douglas Adams skit about the guy who gets rich by finding a way to convert stock prices to music so people can understand them without thinking. With audio, complex realtime input can result in all kinds of nonlinear distortions. This can be fun if you like crunchy sounds, but it can also spiral out of control. A lot of stock chartists (who believe in price movements rather than fundamentals as a guide to strategy) remind me of astrologers or numerologists: there's some math involved, but also a lot of mystical belief in golden ratios, fibonacci numbers etc, without any real scientific methodology behind it.
I have a hunch that a strategy of making small trades with a clearly embedded but arbitrary signal will expose a weakness in these systems for incorrectly targeting local maxima - eg if you start making meaningless trades in penny stocks where the volume Vn = Vn-1 +/- Phi, or whatever - sooner or later your arbitrary but very very consistent signal will be read as a 'strategy' and amplified by a robot.
I have a hunch that a strategy of making small trades with a clearly embedded but arbitrary signal will expose a weakness in these systems for incorrectly targeting local maxima - eg if you start making meaningless trades in penny stocks where the volume Vn = Vn-1 +/- Phi, or whatever - sooner or later your arbitrary but very very consistent signal will be read as a 'strategy' and amplified by a robot.
You'd need to do a bit better: you would have to set up a situation where your strategy triggers algorithm A, which triggers B, which at some point leads to A being triggered again. You might do something like have one options strategy, and one equity strategy, and then do equity and options trades in the same company.
Subject to the observation above that many brokers would probably just reconcile your trades with those of other customers before sending the final totals to the exchange, thus destroying your strategy, you're quite right by but you might well be able to experiment with relatively small sums on relatively worthless stocks.
Is there a way to read this article without subscribing? It sounds fascinating and I would love to read it. I'd even pay to read the individual article, but that doesn't appear to be an option.
I think we discovered a while back the the logged in / not logged in discrepancy is that the WSJ doesn't seem to prompt people to pay that are outside of the US.
Automated trading has dominated the stock market for at least 10 years. (LTC was a melt-down because auto-traders respond very quickly.)
So, if something is actually different now, it's in the word "funds". They're not talking about mutual funds, so what are they talking about that's actually new? (Hmm - I wonder which mutual funds are automated and how much.)
In other words, everyone except the insiders get screwed, but the insiders depend on a steady flow of suckers, er I mean aspiring retirees, to skim money from.