does the analysis also includes HFT traders?
Also I remember hearing that Goldman Sachs didn't have a single day of trading loss. Maybe the smaller Quant funds are loosing out on talent to the bigger firms?
on an unrelated note: I monitor visitors to my online resume and I just got an email saying someone from GS.com visited my resume. I guess there is an GS developer amongst us, may be he can elaborate.
You are referring to Q1 2010, GS had 3 months without a single day of trading losses. During the same period, JPM, C, and BAC also had a perfect quarter (1).
While its true that HFT groups tend to have great win/loss records (2), those "trading profits" basically have nothing to do with the banks' HFT groups. The banks basically all had very big carry trades on -- they were borrowing at short term rates from the fed and lending at higher rates to others and collecting the spread. Since the Fed was desperate to pump money into the system, they allow banks to borrow cheaply and essentially guaranteed the carry trade for the short term.
That sort of trade tends to be highly stable in static rate environments and you get very smooth return profiles. Since they knew the Fed was essentially backing the trade, they piled onto the trade and that particular component of the firms' total daily P&L just dwarfs the rest of their "real" prop trading operation P&L.
Of course, it won't last forever. Its the sort of trade that works until it doesn't -- Long Term Capital Management basically had super levered carry trades on which gave them a really high sharpe ratio until things got choppy and they blew out.
There is kind of an odd tone to this article as if the shrinking of the quant economy is a tragedy when in fact it should be the natural course of this sector if it is not too corrupted. The efficient market hypothesis and information theory tells us that using probabilistic models to extract a competitive advantage can only give good returns when your method is better than all of your competitors'. When two firms have similarly optimized tactics, they bet against each other and the margin of money than can be skimmed off their betting market will shrink to almost nothing. I don't mean that the profits will be split between the two. I mean the bidding war will push the price towards its optimal value where no money can be extracted.
In a competitive quant market, as all the firms perfect their mathematical models, they should all move asymptotically towards the best probabilistic model possible given the available information and eventually erase each other's gains.
The boom in computing technology of the past thirty years was an immense opportunity for mathematicians to make money by racing to perfect their models before others. However, as this industry matures it should eventually take only a few very good systems to extract a thin margin that will barely cover the cost of running the systems themselves.
Moving towards that point of efficiency is a capitalist and libertarian victory that people (at least those who believe in capitalism) should celebrate. The conclusion is a market so efficient that little money can be extracted by betting in it. That means money going to the real economy instead of the wall street casino and more value for everyone. The fact that the sector is shrinking indicates the free market is actually working.
I don't believe in the free market everywhere but I think when it actually works well and create value we should be glad, not lament the loss of the casino profits.
Yes. There's been a huge hedge fund boom for years, and it'd been buoyed by the rising market. (a proper hedge fund ought to be uncorrelated to the market, else it would be cheaper to go long the market more directly. But I digress)
This is the natural evolution of an underperforming asset class.
1) There will always be non-professionals/non-quantitative people making trades that create potential profit for those who wish to arb them.
2) You can still make "good" returns even when you have the worst of the models because of 1).
3) It takes more than N>>2 firms to shrink the profit in a particular model to 0. For evidence, see the large number of high frequency funds that make great returns.
4) The market/world is always changing so the 'best' probabilistic model is always changing. As long as the professionals are moving faster than the non-professionals, they can still extract money.
Could you define what you mean by the real economy? Isn't the real economy whatever it is that solves other people's problems that they will pay you for? The fact that "the sector is shrinking" doesn't mean that its not going to grow still larger long term. I would say that we're just on the back side of a bubble. There are many many many people out there who call themselves quants that shouldn't.
You're waxing philosophic about a hyper efficient future but I don't think it matters because that future won't be here for a very long time.
See 1), 4) and the second to last setence of your allegory (Information only makes a difference if the number of people who have it is small.) for why professionals/someone will always be able to make good returns.
You make some valid points but don't forget that even the non professionals making mistakes will get less losses and a return closer to a long term index when better quant system are betting agains each other and keeping prices close to their optimum levels.
It is not certain at all that the worst professional models will make money just because there are non professionals that are even worst. If the better models are aggressive enough, they should be able to take most of what there is to take from the non professionals and also profit from betting against the lesser professional models (to a lesser extent).
I still maintain that as the models get better and better while they bet against each other, the margin for profit is reduced and the sector as a whole should shrink and that's a victory for capitalism.
"don't forget that even the non professionals making mistakes will get less losses and a return closer to a long term index when better quant system are betting agains each other and keeping prices close to their optimum levels."
Do you have an argument/explanation for this? Also, why are the quant systems necessarily betting against each other?
"It is not certain at all that the worst professional models will make money just because there are non professionals that are even worst."
I didn't mean to imply that its a certainty, only that its a possibility. I mean our whole argument is about some possible utopian future, right?
"I still maintain that as the models get better and better while they bet against each other, the margin for profit is reduced and the sector as a whole should shrink and that's a victory for capitalism."
Aren't you just saying capitalism works? Can't I say the same thing about all those darn internet startups extracting value from me because they're providing me information? Would you use the same argument to conclude that the space of internet startups will shrink?
In a completely efficient market, any stock is just as good as any other - you make or lose money solely due to chance. Nonprofessionals will only take trading losses if they are unlucky .
(Of course, a non-professional who wishes to treat eTrade as a casino may lose money on transaction costs.)
You absolutely can say the same thing about internet startups extracting value by providing you information - but you are incorrect to conclude that the space of internet startups will shrink. In fact, what will shrink is profit margins - as more startups are created, all the easy alpha is off the table. Witness what is happening to newspapers right now.
re completely efficient markets, you're right. My eyes focused on the "making mistakes" part and missed the "closer to" part. I guess I could throw something in there about gambler's ruin, but that would be a little pedantic.
re the space of internet startups : To be clear, I wasn't concluding one way or the other, just asking him to apply the same logic to something he might have positive bias towards and see if he comes to the same conclusions. Also, rather than number of companies, I think the market cap/total profit of the companies is a better metric.
I'm not sure that company margins in internet startups or quant finance will shrink. I haven't thought too much about it but I think they might even grow as information/efficiency attains greater value. Margins per action done might shrink but number of actions will grow. Isn't this the whole story of HFT?
Further, old lines of business will have their margins shrink as the service becomes commoditized. But new high(er) margin lines of business will spring up in their place.
In the case of HFT the number of actions has grown only because there was alpha sitting on the table, but it wasn't worthwhile for human market makers to chase it.
Think about a strategy which generates $100/day trading OTSS (Obscure Thinly Traded Symbol), a security with a $0.25 bid/ask spread. No human will waste their time on this strategy, so $100/day of alpha is left sitting on the table. HFT allows a computer to chase after this $100, $50 on some other stock, etc. This is simply a case of 3 programmers + computers doing work which was unprofitable for humans to do.
Similarly, twitter didn't exist before the internet because no human would be willing to earn $0.0003/day hand-delivering short messages to anyone who cared to listen, being paid to verify that "yes, this message really came from Mallika Sherawat".
Once the market becomes saturated, the bid/ask on OTSS will shrink to $0.10 and the algorithm will only generate $40/day. We are already approaching that point in HFT.
As for margins, they may grow if demand increases. I was really speaking about a situation of increased competition with all else held equal.
"Do you have an argument/explanation for this? Also, why are the quant systems necessarily betting against each other?"
The firms with the good models in the system will tend to keep prices near their optimal values by trading amongst each other so when the dumber investor makes his trades he will get near these prices. It's like in the allegory when the lines are at their optimal because there are enough individuals with knowledge of the fast line, a person that doesn't know about the fast line picking at random will get some benefits because the line lengths are already optimized.
"Would you use the same argument to conclude that the space of internet startups will shrink?"
As an example, consider the bigger gaming firms which now have difficulty competing in the mobile gaming market because the whole process has become too efficient for firms that are capital intensive. On the customer's side of things, they get to chose from a huge library of games most costing less than the price of a cappuccino. More games for less money, that's market efficiency.
I guess you could say that if the investment opportunities grow as a whole, the quant market could grow with it but it should shrink relative to it as it gets more competitive.
Actually, almost everything about Wall Street demolishes the efficient market hypothesis.
Numerous studies have said that an index fund will almost always beat a managed fund after you take out the management fees.
So who's paying all of these people's BMW leases? Their sales skills. If you can give an unsophisticated investor a fuzzy feeling that their money's safer with you than an index fund, you get your management fee, whether or not you outperform the market after it's taken out.
That's the opposite of the efficient market hypothesis, is it not?
The efficient market hypothesis states that a trader can't beat the market except by chance. Everything you just said is evidence in favor of efficient markets.
I thought he was referring to the old perfect information canard in misreadings of Adam Smith. Thanks.
EDIT: I actually don't believe in that hypothesis, either :) I think stock prices are more reflective of crowd dynamics and emergent properties of the various algos (what % are computer driven, now?) than it is by actual prospects of the business in question. Tech stocks are a great example, I see them get pummeled or inflated for terrible reasons on the few occasions I look at market prices. So I think it's eminently beatable, both in an algorithmic and in a Buffet-fundamentals sense, I just think that anyone who beats it will naturally raise their management fees by just as much.
From the article: "Still others say their models simply failed to predict how the markets would react to near-catastrophic, once-in-a-lifetime financial events like the credit crisis and the collapse of Lehman Brothers."
The fact that they still believe those events were "catastrophic," or "once-in-a-lifetime" means they still haven't grasped the fact that markets aren't Gaussian. As first Mandelbrot, then others showed, financial markets exhibit a much larger number of extreme events than a standard Gaussian distribution. Therefore, any trader that doesn't update his or her models to use a non-Gaussian model won't just be wiped out once. They'll be wiped out again and again as they get blindsided by the "once-in-a-lifetime" price swings that happen to happen every decade or so.
reply