Slashdot Mirror
Contemplating Financial Trading At Picosecond Resolution
pbahra writes "One complaint made of the modern stock market is that it is concerned too much on the short term. A second is a long time in cash-equities trading. Four or five years ago, trading firms started to talk of trading speeds in terms of milliseconds. But in recent weeks trading geeks have started to talk about picoseconds, in what is a truly mind-boggling concept: a picosecond is one trillionth of a second. Put another way, a picosecond is to one second what one second is to 31,700 years."
In Kurt Vonnegut's 1997 novel Hocus Pocus, the United States is brought to its knees financially by a company called Microsecond Arbitrage. Everyone invests through them and makes lots of money until a glitch happens and someone else's computer is faster that day. Then the entire country loses its shirt.
Word to the wise.....
Well, this isn't a case of time-lag so much as it's a recognition that interactions with stock trading is actually a control system. The traders are trying to use feed-forward control to predict when and where the market is going to move and apply the right 'control action' (ie. buy or sell) at the appropriate time. Unlike most dynamic systems, however, you are not the only controller - you are trying to predict and exploit the behaviour of other controllers in the system.
With multiple players, the aggregate dynamics are something akin to a dog-fight as each trading algorithm circles and dodges, trying to exploit weaknesses and failures in its adversaries so as to make fractional gains in the time available. If you can control on a tighter time-scale than your opponents, then you can achieve more finely-grained dynamic buy-sell strategies that maximise your profit.
The fixed time-lag between you and the actual market is actually largely irrelevant because of the way the fast dynamic control strategy is being employed. To extend the metaphor, think of them like a attack formation flying to the enemy. If your aircraft can make their attack and withdraw before the enemy can reposition defenses, you will be more successful. The fact that it took an hour for your formation to reach the enemy in the first place is irrelevant.
Scientists point out problems, engineers fix them
altslashdot.org: The future of slashdot.
We still haven't solved the problem that was first noticed in the Industrial Revolution: How to occupy workers replaced by technology, and share the financial benefits of technology equitably.
Luddites and Communists attempted to supply answers early on.
Both answers have obvious flaws.
Later in the 20th century, it looked like Keynsian economics and moderate socialism might be the answer; but that's debatable because WW2 caused massive re-employment and reconstruction which occupied a generation.
Things seemed to be humming along in the late 20th century, the problem was forgotten--then the dot-com crash initiated what will most likely, in retrospect, be regarded as the true beginnning of a new economic and social crisis.
Consider the postal service--essentially frozen in hiring, trying to cut service, and headed for bankruptcy without government support. The replacement of mail with email is cited as a major reason. This is just one small example of technological unemployment.
What does this have to do with HFT? Well, HFT is one example of something the market creates in this situation. There's a general consensus that it isn't productive, and perhaps even harmful. Yet at the same time, it absorbes some of the otherwise unemployed.
When discussing this issue with a friend, he actually labeled me as a neo-luddite. I think that's unfair. I'm not saying we should perpetuate something like the postal service, just to maintain the status quo. Plainly, a policy like that could have negative long-term consequences, since the economy as a whole would be discouraged from innovating.
At the same time however, we still need to come up with an orderly way of compensating displaced workers, and preventing harmful "innovations" from arising in the wake of technological progress. The problem is, determining what is "harmful", who is "displaced", and what, if any "compensation" should be dispensed is fraught with political peril.
The problem remains unsolved and, IMHO, inadequately acknowledged by policy makers.
I don't think it should be illegal, I just think one dimension of the assessed taxes should be length of time the asset is held. Set thresholds at, say
1 second -- 99.99% marginal tax
1 minute -- 99% marginal tax
1 hour -- 95% marginal tax
1 day -- 90% marginal tax
3 months -- 50% marginal tax
1 year -- 15% marginal tax (today's capital gains rate)
10 years -- 10% marginal tax
I actually know a local guy who implements trading algorithms in programmable-gate-array hardware for the purposes of instant trading.
Nothing that crazy is needed just add a 10c per trade tax, the only problem is all the major trading countries would need to do it simultaneously or else the market would just move. I thought we were close a few years ago when the central banks and regulators started moving in lock step on policy but that only lasted until the first European debt crisis hit and China and the US got in the spat about currency valuations.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
They're silently inserting themselves as a middle man, making your trades just a little bit worse off for you. That behavior is incompatible with a free market. I believe there's a saying: no single snowflake is responsible for the avalanche. Or, in terms you could understand, no single sperm is responsible for the stinging in your eyes after a half dozen niggers make you the center of a bukkake session. And yet, we're still feeling the effects of a financial crisis caused by individual actions which don't usually have much effect in the the long term.
Although the picosecond thing is silly, the New York Stock Exchange now operates a co-location facility in which each trading platform gets a uniform 35 microsecond latency for the incoming trade data. Some systems can turn around that data and do a trade within 12 microseconds.
Computers aren't fast enough for this. The latest thing is writing trading algorithms in Verilog and compiling them into an FPGA.
This worries me.