BlackRock's Robot Stock-Pickers Post Record Losses
bloomberg.com
bloomberg.com
I'm particularly interested as I was recruited for this group multiple times, so a bit relieved I didn't pursue it. I wonder if there are any lessons (about what to avoid) to be learned here about the way the group was structured...or if it was simple randomness gone wrong.
Hard to blame the robots then.
They've got some pretty good arguments in favor of this, too: https://vanguardblog.com/2016/10/13/when-nothing-is-somethin...
They do a great job with VTI and their other funds to attempt to minimize the requirement of actually selling assets when liquidation requests come in, but that's not the same as saying that they, Vanguard (and more accurately the fund managers), practice a "Buy and Hold" strategy. Tracking an index well actually requires lots of buying and selling!
Venture captialists' strategy involves mostly taking sizable losses, but with enough breakout successes to still make money in the aggregate.
"The market can stay irrational longer than you can stay solvent" -JMK
Also, they tend to mess with your mind (loss-aversion psychology, deep feeling of pain at losses) and make you exit and enter the trade multiple times, adding up to transaction costs, too.
All in all, a) Buy and Hold may actually be quite sensible and b) stop-losses are actually much more complicated than you'd think and seem to be a mathematical rabbit hole.
Stop-losses may make sense in, say, an intraday pair-trading strategy where an analysis of past return evolution clearly shows that winners keep winning and losers are hopeless beyond a point.
I'm especially surprised it's used in a model like this. Most quant funds use some type of portfolio risk management to allocate capital toward bets that with the highest expected return while controlling overall risk. That losing stock may be offsetting other risks in the portfolio.
As I mentioned in my other comment, a stop-loss is a very crude approach to risk management. It will help control the middle of your left tail, but the far left tail extreme events cannot be protected against by such a rule.
Remember it's trading, not longterm investing.
This is exactly what not should happen when investing longterm: Being driven by emotions rather than rational facts.
Putting a limit on one's potential losses is pretty rational, not emotional. Models are not perfect, and you do not want to find yourself with high paper losses when they fail.
Of course, if your models don't work well together with stop-loss limits, then that's a problem.
At least here they say that it is not a good idea with a quant approach: http://en.swissquote.com/epb/support/faq#node-301
The BlackRock example plus the fact that many ETFs beat managed Funds in terms of performance seem to support the position that human interference is in most cases for the worse.
You'll have a bad time if you think your model is infallible.
By bugs and adverse selection I'm assuming you're talking about something like automated market making. For traders like that, a loss limit makes sense because of technology risk. If your order router has a bug like Knight's did, or your data feed gets stuck, you can lose a lot of money very quickly.
My track record is much much better!!! Honest!!!
Its an interesting question about machine learning and market returns. But there is a remarkable number of people who seem to just step over the line of legality and pick up the extra profits they need which taints the whole industry.
[1] https://www.theguardian.com/business/2016/jul/04/libor-riggi...
Not betting on it, though.
P.S. Looking for something simple, e.g. "If you have derivative trading, your GDP will be 5% larger, and the gain is spread across the population in proportion to income." The kind of things you have for engine efficiency, or compiler optimization ...
With short selling, derivatives and high frequency trading, I can at least see the argument for increased liquidity and so forth... But how does society benefit when leading AI research minds spend their time building gambling models that try to predict the sentiment of other gamblers?
At the same time, I recognize that forcing any such social logic onto research is a slippery slope towards Soviet-style government-stunted research. So I guess I'll rather have machines gambling trillions of dollars.
The idea is that letting algorithms decide will result in better/more productive allocation of resources, which will result in more and better everything.
Of course, is that reality ? I would point out, however, that stock markets resource allocations are far better than royalty/governments allocating resources.
Is it? I thought the idea behind HFT was to make a shedload of money for the firms with the most effective algorithms. To be clear, I have no problem with that, and I'm not close to the industry. But it would surprise me to hear that those in it conceive of themselves as optimizing social resource allocation, rather than, say, setting themselves up to retire at thirty with all the money they'll ever need.
The algorithms talked about here are not HFT algorithms, but investment algorithms. Things like "if it's down for 3 days in a row, allocate 5% in the stock" type programs, written using machine learning. They generally will not change orders rapidly. They are about more efficient/effective investment, and the argument I gave is more focused on what the function of a stock market and investing is.
Because these algorithms are much more like traditional investors than HFT.
GM's stock price has very much to do with GM's actions. GM can do countless things which will impact its equity value.
Also, changing the effective|marginal tax rate by 50% has material implications on your business model and capital structure, e.g. debt (and tax shields).
10% to 20% is a 100% increase not a 50% increase. Now sure, paperwork changes, but the price of your widget or the layout of the factory etc don't really care about the taxes outside of extreme cases.
Anyway, profitable companies like GM can issue stock to raise capital or buy back stock. Buybacks don't really depend on the stock price as it's not an investment it's another form of dividend. Issuing stock is an inefficient way to raise capital better to issue bonds or not issue dividends.
Now sure, there are second order effects of stock price such as stock options. But again +/- 10% to stock price on a given day does not do much.
PS: Consider Microsoft if the tax rate where to increase to say 40% what would they change?
For your widget company, it does matter! ;-) What widgets you make, your price, your price relative to competition, market share protection, pricing power, where you build your factory, PP&E decisions, and more all depend on your tax rates. I promise, and want to compete against firms that overlook these parameters.
For GM, you're overlooking other (mis)management decisions that will show up in the share price.
Right now, MSFT is facing just such an issue re: re-patriating money from overseas. If there were a one time foreign tax holiday as floated by Obama, MSFT would choose very different decisions than the status quo in terms of buy backs and R&D spend. In terms of operations changing, I guarantee they'd re-think their debt and their product mix within their 3 reporting segments. Some products wouldn't be profitable enough to sell if the profit margin shifted 2-3%.
I have had several successful CEO's all tell me to ignore taxes. That does not mean you install or don't install solar panels based on tax breaks. It can be very important, but it's generally premature optimization. Further, it's not the top tax rate that is important it's differential tax rates aka A @X or B @ less than X.
MSFT's re-patriating money is a good problem to have. They may see a tax holiday in the next administration or they may not. But again, it's getting to that point is the issue.
John Bogle, founder of Vanguard and renowned investor, seems to disagree with you: "The stock market has nothing—n-o-t-h-i-n-g—to do with the allocation of capital. All it means is that if you’re buying General Motors stock, say, someone else is selling it to you. Capital isn’t allocated—the ownership just changes. I may be an investor, you may be a speculator. But no capital goes anywhere. This is basically a closed system. You have new IPOs and whatnot, but they’re very small compared to this vast thing we call a market, which is now around $24 trillion. The allocation of capital? That’s just nonsense."
I think you are mixing up the concept of a market economy with a specific kind of market, the "stock" market.
Why would people sell equity in their company except for cash, and why would you pay cash for ownership in a company except for the stream of income it represents and the secondary market for that ownership?
How come people decided it is illegal to trade ivory from an elephant killed a couple hundred years ago? Was it made illegal to possess child pornography that was already created only because it is morally toxic, or does said consumption also induce demand for additional victims?
Even though debt markets have a lot more to do with day to day financing of corporations, equity markets have a great influence on terms. Furthermore, the unavoidable importance of secondary markets for debt and their derivatives can be understood by considering the importance of secondary markets with respect to prices of homes, automobiles, or any other large consumer purchases.
My main point really was that the examples parent gave seemed much closer to examples of market economics in general and would hold true, with or without stock markets.
It's a distinctively American quality to feel dirty for having these doubts and associating them with the Soviet system. Free markets often create perverse incentives, races to the bottom etc. and it's the genuine role of governments to counteract these: Outlaw it, tax it, or – what may be enough in these cases – make absolutely sure those standing to profit from these activities also bear the full brunt of their failures.
But in the broader context of the entire real world, a well-functioning market is valuable in itself. It enables people to transfer risk inexpensively, instantly, automatically, and at fair prices. This competitive pricing mechanism also provides valuable signals to the real economy so people can invest in areas that need it.
Imagine a market with 4 participants:
-A farmer who grows corn and wants to lock-in a price for next year's harvest so he can invest in a new tractor and refurbishing his grain elevator.
-A baker who wants to lock-in his price for his corn purchases next year so he can invest in new ovens for his plant.
-A speculator who monitors price trends and takes risk intermediating between buyers and sellers.
-An exchange that provides a meeting place for these parties.
On Monday, the farmer places his orders with the exchange to sell his 2018 corn harvest. The baker is on vacation until the next day, so the farmer can't transact. However, the speculator knows bakers tend to buy corn around this time of year, so he buys the contracts from the farmer at a slight discount. The exchange also takes a fee from the farmer, and from the speculator.
The farmer goes off to the tractor dealer, and gets to work refurbishing his grain elevator.
On Tuesday, the baker comes to the exchange. No farmers are around, just the speculators who bought from them yesterday. Today is a lucky day for the speculators who can now sell to the bakers at a slight premium. Sometimes they misread the market and take losses instead. Again, the exchange takes a fee from the baker, and from the speculators.
The baker is comfortable funding purchases of his new ovens, knowing that he isn't at risk of corn prices rising next year, which would cut into his profit margins and make his plant unprofitable.
So far, the exchange has made money for providing a meeting place, and the speculator has made money for matching up buyers and sellers who arrive at different times. The farmer and baker each lost a bit of money by trading against the speculator--his profits are their losses, and vice-versa when he's wrong.
2018 rolls around. The cost of corn has risen due to import tariffs. The baker made money on the contracts and the farmer lost an equal amount. The exchange and speculator each made some money. Transactionally, this is actually a negative-sum game, because the exchange makes money no matter what, but it let everyone involved focus on their particular business and insulated them from risks they didn't want to bear.
The criticism was directed at the ever-expanding high end of financial markets where it seems to me the benefits it produces in the "real world" are marginal at best and in no way proportional to the money earned in the sector and the brain capacity it utilises. Some deep learning outfit with 200 quants basically putting pressure on the "speculator" in the scenario you sketched out, that may or may not create some small marginal improvement in the market, but could also just serve to better capture any residual utility the farmer and baker may have had from participating in the first place.
Like an engine, generalizations about quantities of reagents and thermodynamics aren't going to be violated by the real world, yet the engineers designing engines will understand the dynamics of low level interactions that are unexpected by someone with only a top level view. That is, in general more funding and more instruments that fine tune allocations of resources and risk really do accelerate growth. And yet, the details within the financial industry that can pervert incentives or corrupt signals do undermine some of the gains in funding efficiencies that are to be had. Given a fixed amount of design work a simpler engine might be more reliable than a complex one, but it won't be as efficient as one with more features and sufficient engineering to make them reliable.
You can have rent extraction through complexity as well as through friction. It's not obvious to me that a more complex, deregulated system is more efficient than a regulated, simpler system. I know that is the dichotomy that is usually talked about - a balance between efficiency and morality, but the sort of marginal increases in GDP we're talking about could easily be wiped out by other factors, for instance complexity that the clients can't understand, and greater instability in the general economy.
What I was saying though, is that a blunt approach of declaring that financial instruments can only have some specific measure of complexity would eventually limit the resources available to new projects and businesses. The tangent about rent extraction was a distracting and motivated by my suspicion that a lot of people simply resent incomes in the finance industry, and want to see them make less money regardless of net impact on the economy. However, just as a baker counter intuitively makes more when the price of flour increases, less fine-tuned and efficient financial markets could ultimately make more money for the people working in finance.
I think that there is a belief that complex financial instruments inefficiently insert extra steps into the flow of finance and extract rent from those manufactured inefficiencies. Yet, a better reading is that they circumvent some inefficiency elsewhere in the financial system, and the people who implement that gain in efficiency are able to extract a little of the decreased waste until they too are circumvented.
A healthy derivatives market helps businesses focus on producing useful things for society. Every human should buy health insurance. Every big business should buy commodities and currency insurance (they don't call it that).
An unhealthy derivatives market encourages business to gamble rather than produce. GE Capital, before it got shut down, was at one point a bigger business than the rest of GE.
I mean, imagine there is a machine that is capable of predicting the markets with 100% reliability. Then soon enough, the machine will acquire a reputation and everyone will follow its advice. The machine says "the price of X will go up", everybody would buy X, so the price of X would indeed go up, but this price surge would have only been initiated by the machine statement and we'd have a self-realizing prophecy. Other situations would be more complicated I suppose, but in every case the machine would have to take into account its own prediction, which would be some kind of a mise en abyme.
In the end just like antic oracles, the machine would have to make cryptic predictions that would make sense only with hindsight.
If someone had a machine that could predict the stock markets with 100% reliability, or even 55% reliability, do you think they would make those predictions known to the public? Or would they start a hedge fund and become one of the wealthiest individuals of all time?
Quant shops have to continually research to detect new patterns to get ahead of the competition for this reason.
It isn't a whole lot different than an FFT which pickts apart frequency components, HFT algorithms amplify correlation and dampen noise. Then they trade on the correlated signal against the affected securities in an anticipation of the correlated signal's effect.
If correlation was equal to causation they would be 100% effective at making lots of money, but that isn't. So to the extent that the algorithm identifies a correlation that is close enough, it makes money. But it can be crazy like 'sunny days in Paris' is correlated with higher shopper volume in at 'The GAP' and as the sunny days begin to pile up the algorithm buys GAP shares and as the sunny days decline it can sell off GAP shares.
That example we can imagine that people think about buying new clothes when it is sunny but the algorithm doesn't care what the correlation is, just that it has a strong probabilistic link to the changing value of a security. Nothing magical about it.
Nowadays, though, you see this sort of behavior in legal and more respectable ways. However, I think the purported results are highly disingenuous. What do I mean? Let's use a well known investment advisory service known as AAII.
AAII has a model shadow stock portfolio with fairly well known and transparent stock selection criteria. However, they often announce stock picks and stock sells to their followers within a day, supposedly, of their actual transaction. I've been studying the behavior of their stock pick/sells announcements, and on average, stocks are trading +5% to -5% (depending on whether it was a buy or sell) within the day following the announcement. This is HUGE for many reasons:
1) They often have multiple transactions on the portfolio in any given year.
2) The performance on the portfolio is boosted by, on average, 5% for any buy or sell. Compounded, this makes a big statistically significant difference on their purported shadow stock portfolio's CAGR and strategy.
3) They do not disclaim this behavior as a caveat emptor.
4) Though their strategy is transparent, their data feed / source is not. No one will be able to replicate their stock picks/sells on their own with their own live data source.
Was tickled by the allegory that could be made, upon reading this from https://en.wikipedia.org/wiki/Jinn:
> The Quran says that the jinn were created from a smokeless and "scorching fire", but are also physical in nature, being able to interact in a tactile manner with people and objects and likewise be acted upon. The jinn, humans, and angels make up the three known sapient creations of God. Like human beings, the jinn can be good, evil, or neutrally benevolent and hence have free will like humans.
I guess you're attracted by the part "can be good, evil, or neutrally benevolent" but it's too little for my taste, the rest is more problematic to be related. And can't most things "be good, evil, or neutrally benevolent" anyway?