One never stops learning new things. In the last three years a new industry developed in Israel, with about fifty quant trading companies. Almost all learned the profession in Zahal's Unit 8200 which must be the largest math incubator of the world.
In the pic Nir Glikstein, 39 y.o., who lives next to Kever Benjamin City, is the founder of the most succesful of the quant partnerships (2008 profits 230 million dollars, 2009 - 170 million). When I drive to Petach Tikva, I see his name on the sign at his modest moshav house. They have 150 traders working from a Herzliya office, trading in the Seoul stock exchange and who knows where else. The whole thing is so anonymous that few know of their existence, or who are the capitalist partners and how works their FINAL algorithm.
At least they have a good answer to the eternal question:
"Dad, if you are so smart, why aren't we rich?"
"Son, we are, we are."
23 comments:
If he's so rich why does he live modestly? A rich man should have a house that is appropriate to his means. My daughter has a friend whose stepfather is a quant - their house is like a palace in a private park.
K
People here likes to pretend that they are modest moshavnikim (farmers). Maybe they are not pretending, they are.
So called quant science is to finance what lovely salesgirls are to Ferrari cars; mere window dressing to attract the moneyed men. While I admit to a twinge of jealousy over the financial success of the quants, it has to be said that there is barely a single new insight that quants have brought about. They use various theories in analogy with fluid mechanics and statistical physics to obfusticate the field of investment and line their pockets. The mathematics exist for the magical purpose of intimidating the uninitiated and to convince the awe-struck outsider to part with his money. The sage Falkenburg had a post up some time ago, pointing out that usable option prices can be derived using a simple binomial model but that is not sexy enough for most people. They would rather go with arcana like the Ito calculus, just as one can solve a minimum energy problem using soap bubbles or invoking the calculus of variations. I am sure that out- of-work physicists and mathematicians who would otherwise be driving taxis, are grateful for the opportunity to work in a field that restricts employment to themselves.
There are a few insights but they get old very very fast and become obsolete and untradeable.
I too identified a few non-too-obvious relationships but the margins are very thin and soon abandoned the wffort.
If the quants have any insights, they're not telling you about them. As J says, the margins are razor thin - in order to make them work you have to trade in millions of shares to yield anything of worth. The aforementioned palace owner was at one time responsible for a significant % of the trades on the NYSE.
K
Another difficulty is the fixed cost of maintaining a working computer-trading infrastructure. Quants need to work at certain scale to be feasible.
Looking for price patterns is as old as the hills. Another set of traders, the technical analysts use what they see as head and shoulders, pennants and trendlines to execute trades. They however have a quaint hillbilly air about them and thus can't really prise money from the major funds. But if all this is translated into what the mathematicians derisively call 'epsilontics' then the fund managers sit up and notice, for it is no secret that the economicsts suffer from an acute case of 'physics envy'.
Computerisation and the telecoms revolution had a decisive role in all this. It is the age of the nerds now. I recall some old fogeys claiming that they merely had to trade a couple of lots of cocoa a day to live in style, now the same guys may have to churn a few hundred lots with its attendant increase in risk to earn anything close.
You misunderstand the quants - they are not betting like the technical analysts, they are playing sure things. The "sure thing" might only yield a dime but if you do it a million times then you have a million dimes.
K
I dont think there are "sure things" but only probable ones. Of course, high probability means sure.
If such is the case, then I have misunderstood what the quants do.
"Statistical arbitrage" (a main quant strategy) is not 100% certain but they look for patterns with a high level of certainty. If you were to do statistical analysis on "head and shoulders" or something like that then you'd find the probabilities would be only slightly better than random. Quants laugh at "technical analysis" - you could plot random data like barometric pressure and you'd see "head and shoulders" and other patterns - humans are very good at seeing patterns even if there aren't any. A "feature" of our programming that is in some circumstances a "bug".
Also technical analysis usually implies taking a position in the market - long or short. Statistical arbitrage is typically pairs trading - buy one thing and sell another, which is agnostic as to the direction of the market. So there are no big profits when the market turns in the predicted direction but instead lots of little profits as the pair trade gets executed millions of times.
K
Another way to look at this is that the technical analysts have crude tools - an old fashioned barometer that can make rough forecasts which sometimes come true - "falling barometer = storm is coming" and the quants have the stock market equivalent of the modern weather forecasting models with thousands of datapoints run on a supercomputer. Such forecasts (though not infallible) have a much higher level of "skill" (in the technical sense, meaning accuracy) so it's not surprising that people would rather give money to the guys with a supercomputer instead of a mercury barometer on the wall (even though on rare occasions the barometer will "outperform" the supercomputer).
K
It is not the barometer but the person who interprets it.
Only up to a certain point. The most skilled barometer reader cannot produce results as good as those produced by an advanced computer model, just as the greatest chess master can no longer outplay a computer. Google now has automobiles that drive themselves and they feel that far fewer accidents would result if people were not driving - especially people who are drunk, texting on their phones, etc. Humans are no longer the top intelligence -we had better get used to it.
K
K, according to you they are essentially arbitraguers working at high speed. I guess that explains the 'sure win' part.
Some of the Google guys are downright creepy. They are in the running to replace both Microsoft and Oracle as the most hated software firm. They shouldn't give themselves too many airs; one can always pull the plug.
"Statistical arbitrage" is not quite true arbitrage...the idea is say that your computer discovers strong statistical correlations that are almost but not quite sure things.
K
A next step is to write faux orders to establish how the market really reacts and when they react (like trying to buy or sell) you go for the jugular and kill them.
I like the trap strategy but I have not finished polishing it. You dangle a small but delicious carrot and you kill them when they try to get it.
I'm not sure this works - price is sensitive to quantity - if you put 5 shares on the market the market will indicate a price of X but if you put 5 million shares on the market the price will be Y and X will not be a useful data point.
K
Yes, but some computers are born greedy and will go after the smallest bit of cheese.
Hmm. If you could figure in advance what movements triggered massive buy and sell programs there might be some way to pre-position yourself and profit. (There were inadvertent hints in the May 6 crash) . And perhaps you could find the triggers thru some kind of "teaser". But without inside information this would be very hard to do - there is so much "noise" in the market from other trading (a lot of which is competing robots - you have robots selling to other robots) that it would be hard to find the signal unless your "teaser" was very substantial, as in billions.
After May 6 they supposedly put in "circuit breakers" if any stock falls more than 10% in a short time. But suppose you could trigger a temporary 9% drop in any given stock, which would then quickly recover because the drop was artificial?
But I think this is a battle of the elephants in which the mice like us can only get trampled under foot.
K
Interesting line of reasoning! If you could figure in advance what movements triggered massive buy and sell programs there might be some way to pre-position yourself and profit. K, since computers behave reliably following blind, inflexible (and rather simple) programs, they should be predictable. Computer programs can be "exposed" and learned and therefore, gamed and exploited. You dont need inside information. As someone said: You can learn a lot by observing.
Easier said than done. A lot of the program trading is what they call "high velocity trading" - the traders position their servers physically close to the stock exchange computers so they get their trading data milliseconds earlier that the general public and in order to beat the system you might only have milliseconds of your own before the arbitrage window closes. As I said before a game for the big boys to play - one supercomputer warring against another.
K
Post a Comment