SoylentNews is people
SoylentNews
In meetings with clients, hedge fund representatives present flashy charts and speak equal parts oracle and mad scientist. And for technical analysts who market themselves as the most technical of analysts, the mathematical jargon—"stochastic oscillators," "Fibonacci ratios," "Elliot wave," "Golden ratio"—evinces a certain disarming beauty. "This mathematics is embedded in the structure of the universe," Cynthia Kase, who runs a firm that employs "wave analysis" to predict oil prices, told Bloomberg News in 2012. "It is the language of God."
Though much of this language is too gaudy to be embraced by sophisticated investors, there is a more subtle mathematical con that many, including editors at most of the top financial journals, overlook. The positive results that emerge from testing the performance of an investing algorithm on past market data, a process known as backtesting, can seem reliable and logical. And they sometimes are. But often, in practice, presentations of these results, though marketed as scientifically rigorous, conceal statistically insignificant methodologies.
(Score: 2, Interesting) by Anonymous Coward on Monday November 23 2015, @10:19PM
the customers' yachts? [barnesandnoble.com]
(Score: 0) by Anonymous Coward on Monday November 23 2015, @11:13PM
Well, if you can imagine that someone really had magic powers and could double capital every year, they would make far more than their clientes even if they only charge 5% of profits, cause they have over 20 customers. In that case, it would make sense.
(Score: 4, Insightful) by Snow on Monday November 23 2015, @10:21PM
When you have many, many people all looking at the same chart, and a pattern emerges sometimes it becomes a self fulfilling prophecy. If all the signals appear to suggest a rise, then people buy - creating a rise.
(Score: 5, Interesting) by JoeMerchant on Monday November 23 2015, @10:57PM
Back-testing a novel algorithm proves that the algorithm works in a marketplace which is devoid of that particular algorithm as a player.
Major deployment of the novel algorithm will alter the market behavior, often unpredictably (except that performance is more likely to be lower than shown during backtesting, rather than higher.)
🌻🌻 [google.com]
(Score: 2) by Dunbal on Monday November 23 2015, @11:08PM
Until a mutual fund or a bank decides that this rise is a great time to dump 5 million shares. And everyone is fucked. Again.
(Score: 1) by SanityCheck on Monday November 23 2015, @10:31PM
Isn't stochastic referring to essentially random or unknown? How is that a measure of financial reassurance when you base shit on things you have no idea about?
(Score: 1, Informative) by Anonymous Coward on Monday November 23 2015, @10:33PM
A stochastic oscillator is a noisy wave function. Specifically, it's a way to try to estimate the min/max of price oscillation
(Score: 3, Insightful) by VLM on Monday November 23 2015, @10:58PM
In theory nobody should ever get ahead in a game of poker other than short term fluctuations. None the less, everyone thinks they can play poker and win. Some do. Sometimes its more than survivorship bias / random change. Sometimes not even due to insider trading.
(Score: 0) by Anonymous Coward on Tuesday November 24 2015, @10:24AM
Nonsense!
This *creates* wealth.
I heard it on CNBC
(Score: 0) by Anonymous Coward on Monday November 23 2015, @11:10PM
Like this. Assume coin toss is a stochastic process where the head is as likely as the tail. You can deduce that over time, 50% of the time the head will turn up.
(Score: 3, Informative) by Non Sequor on Tuesday November 24 2015, @01:13AM
The basic theory is that arbitrage-free (think no-free-lunch) conditions say that you can construct a portfolio that earns a certain interest rate "risk-free". Treasury bonds from a stable government are generally taken as a proxy for this risk-free rate since they're generally regarded as the investment with the fewest adverse contingencies. If your country's bonds default or otherwise become substantially devalued, it's highly likely that whatever is going on is going to fry any other investments in your country.
To have an expectation of earning higher than this risk-free rate, the theory says that you have to take on some risk. One of the smallest steps up you can take is to go with a AAA rated corporate bond rather than your treasury bond. Let's use Microsoft as an example. Based on their track record and based on the underwriting of the bond issue, it's viewed as very unlikely that they would default on their bonds, so those bonds trade at an interest rate that's 1 to 2 percent higher than the Treasury. That surplus is a risk premium. It is larger than is needed to balance out the small probability of Microsoft defaulting on its bonds.
Moving to another risk category, we can look at Microsoft's stock. This is much riskier than the bond. For the bond to default, Microsoft has to do so badly that it can't pay the bond (or even issue new bonds to finance the debt over a longer period). For the stock to do badly, the valuation of Microsoft's business, less the value of its debt, has to decline. You have a bunch of different probabilities here. You have an assessment of how well Windows is going to do, how well Office is going to do, how well server software is going to do, how well Xbox is going to do, and how well their mobile offerings are going to do. People have a mixture of expectations for all of those things. There are different theories of how to relate information about a companies line of business to its stock price. The stock price can change a lot, while spread of the Microsoft bond versus the treasury bond does not fluctuate quite so much.
Generally stock market returns are modeled as following a log-normal distribution. It's hard to challenge that model because there is not overwhelming evidence for any single theory of stock value always being right. Bond prices fluctuate too as interest rates change (although if you hold the bond to term, you'll earn the yield you bought it at regardless of those fluctuations). There actually isn't even a decent model of how interest rates change. The result here is that market prices are seen as semi-coherently relating to a lot of variables that lots of people have theories about, but which no one knows for certain. Saying it's stochastic is objectively correct. Setting a deterministic prediction, and being correct in that prediction, will earn you money, although you generally should be skeptical of people who say they know what's going to happen.
There are a couple of subtle problems with this theory though. The theory implies that risks are a totally ordered set (i.e. market prices determine yields, if you sort yields in ascending order, you order the available investments from "least risky" to "most risky" under this theory). The reality is that, while we can talk about risks that are shared or not shared between some securities, it's indeterminate whether some risks are more severe than others or not. Since some of the comparisons are indeterminate, this is a partially ordered set. That arbitrage-free condition says that the market will settle on some way of ordering those indeterminate comparisons.
It's also possible for majority opinions about the riskiness of various things to be wrong, or for there to be issues that aren't even being contemplated.
Arbitrage opportunities may exist, although it's likely that they are very difficult to take advantage of. See the story of Long Term Capital Management, a hedge fund staffed with Nobel laureates, which made correct predictions about the convergence of some investments, but went insolvent when in the short term the investments went in the opposite direction, resulting in its creditors being unwilling to trust their predictions.
Write your congressman. Tell him he sucks.
(Score: 4, Informative) by mmcmonster on Tuesday November 24 2015, @12:13AM
The hedge funds have lost a lot of their luster. They've mostly done worse than the S&P 500 (with dividends reinvested) over the last five years. The problem is, as soon as a hedge fund goes under, the same people that created it make a new one and the cycle starts over again.
As I've said numerous times before, invest in low-cost S&P 500 index funds (Vanguard funds or other funds with similar expense ratios). Put money in every pay period and forget about it. Make sure the account is set to reinvest dividends, because that's where the magic happens.
Also, because you're not selling funds and shifting your money around, index funds are quite tax efficient.
See this for more info: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=88005 [bogleheads.org]
(Score: 1, Insightful) by Anonymous Coward on Tuesday November 24 2015, @12:17AM
Exactly.
Also the reason those formulas 'work' is they *only* work when the market is semi steady. You get any sort of bounce they no longer match. At all. Even the dude who came up with them will tell you that.
(Score: 3, Disagree) by Phoenix666 on Tuesday November 24 2015, @12:47AM
Hedge funds are a ponzi scheme. Those first in, win, the rest lose. It was a big scandal when Bernie Madoff was exposed, but he is representative, not aberrant.
Washington DC delenda est.
(Score: 4, Insightful) by TheRaven on Tuesday November 24 2015, @11:52AM
They've mostly done worse than the S&P 500 (with dividends reinvested) over the last five years
But that doesn't matter. The entire point of the scam is that you don't need to do better than the market on average. You create 10 funds with mostly random investments. They're likely to have a distribution somewhere over the bell curve. Some will lose all of their money, some will do a lot better. You only show the successful ones to potential investors, along with the caveat that, of course, past performance doesn't guarantee future returns (and the strong implication that it does really). Then you charge them a modest commission on the huge amounts that they're investing. If they're lucky, then your continued random investment will keep doing well. Otherwise, eventually they'll either lose their money or pull it out and put it elsewhere. By which time, you have 20 more funds building a track record, and a couple of those will look sufficiently promising that you can sell them to the next round of suckers...
sudo mod me up
(Score: 2) by JoeMerchant on Wednesday November 25 2015, @03:24AM
The thing I find disingenuous about many long term market analyses is they they discount the failures - they take a look at today's S&P 500 (or index of choice), backtrack 20 years and then say "if you had invested in the S&P 500 20 years ago, your return would have been X." What they really need to do is go back 20 years, invest in what was the S&P 500 _then_ and tell you what your return would have been - including the companies that have gone completely bankrupt.
🌻🌻 [google.com]
(Score: 1) by illness on Tuesday November 24 2015, @12:17AM
That's a general problem with empiricism. It works very good if the data comes from repeatable experiments while controlling all other variables than the one we want to measure, e.g. experiments in physics.
It does not work so good for the softer sciences where it is a lot harder to design or execute meaningful repeatable experiments, e.g. macro economics, behavioral sciences.
One way to circumvent this problem is to look at the already available data. This has been done extensively for financial markets because the data is easily available. From this perspective, the stock market trading algorithm is a hypothesis on the behavior of the stock prices. If the hypothesis is true, the algorithm will earn money, otherwise it will loose. Then the algorithm gets tested with past market data and always wins. Everything is great.
The problem is that the data set normally doesn't contain all the necessary information to define an algorithm that works for all possible situations. As a thought experiment, consider a function f(x). I give you f(0) = 0 and ask you to predict the value f(1). Without any additional information, a good guess could be "f(1) = 0".
I tell you, f(1) = 1 and you could update your hypothesis to "f(x) = x". Now suppose the function is f(x) = x for x = 200. The hypothesis will work for all past data, and even in the future. But finally, when x = 200, we're in for a big surprise.
Nassim Taleb discusses this and related arguments at length in the great swan. He also tells the story of the turkey. The turkey wakes up every morning. Everything is great, he roams around freely and has no worries. He never thinks that tomorrow could be different. He always goes to bed early because he anticipates the next day. Nobody told him that Wednesday is Thanksgiving Eve.
Anyway, they are in good company. Even (economic) nobel price winner fall into that trap. Check out Long-Term Capital Management as an example.
And now, I've read the article and realize they are talking about overfitting. Oh well, who cares about the article anyway. Right? Right???
(Score: 2, Touché) by Anonymous Coward on Tuesday November 24 2015, @12:29AM
Oh well, who cares about the article anyway. Right? Right???
There are people who actually read the article? I barely even read your comment!
(Score: 1, Interesting) by Anonymous Coward on Tuesday November 24 2015, @12:43AM
Some of the early adapters, who come in and institute statistics and data-driven analysis instead of just relying on experience and gut feel, can come out big winners. BUT as people pile in trying to do the same thing (including spinoffs from the early successes), the market becomes saturated. There's too much "smart money" chasing not enough "dumb money"; meanwhile, the house is taking the vig.
(Score: 3, Funny) by aristarchus on Tuesday November 24 2015, @06:10AM
Some of the early adapters,
Just what I was saying the other day to Sam, my barbar! "Sam", I says, "it is the early adapters that always win! I mean, just look at the adapters for the iPhone! Nobody had any choice but to use them, since they were the early adapters, and no one else at the time adapted before they did. So go with the early adapters."
I, for one, am totally amazed that people who seem to have a talent for making money are so bad at things like ethics, grammar, aesthetics, and math. Maybe it is just that they are all crooks. And to think that Wharton legitimized all this illiteracy and ignorance. (And by the way, Sam is my "Barbar", not my hair cutter, since he is an elephant and has no thumbs. Definitely not an early adopter, but he could just use an adapter, if he wanted to be a barber. )
(Score: 1, Insightful) by Anonymous Coward on Tuesday November 24 2015, @02:19AM
If a financial adviser were any good, he/she wouldn't need YOUR business.
(Score: 4, Insightful) by DIMT on Tuesday November 24 2015, @05:27AM
Getting a consistent 15% return, which would be an amazing feat, pales in comparison to the amount you can make with a 1% cut of all assets you manage for dozens or hundreds of clients each year. You don't even need to risk your own funds.
Whether most advisors are worth a damn is beside the point, there are financial incentives in place which make advising a prudent career choice regardless of effectiveness.
(Score: 3, Interesting) by jmorris on Tuesday November 24 2015, @05:28AM
You can't even be pitched by a hedge fund or really even know what a hedge fund does unless you are a 'well qualified investor' meeting pretty high requirements for 'disposable' wealth. If they are getting conned it is a good thing, economically speaking. A proper capitalist economy wants capital to end up in the most competent hands.
(Score: 0) by Anonymous Coward on Tuesday November 24 2015, @07:04AM
But if you hear a player explains how he wins with a lot of big words and math jargon you know it's bullshit :p.
Sure some players get very good at it but the rest can also improve, and in the end they're all battling for the same pool of wealth.
(Score: 1) by AlphaMan on Tuesday November 24 2015, @04:23PM
The article starts by referring to the jargon of "technical investing" - terms like stochastic oscillators and Fibonacci ratios - a discipline not taken very seriously by "quant" researchers, but the author clearly fails to understand important distinctions in the field. All these investors use numbers, yes, but in pseudo-scientific ways in the case of technical analysts versus rigorous and empirically well-supported ways by quant researchers, who are well aware of the danger of over-fitting which seems to be the gist of the article.
Nevertheless, from my own experience in the field, there are any number of clueless PhDs who are eager to apply the models they know best, regardless of how appropriate they are. Also, even egregious melt-downs that are relevant to the argument against quant-based investing, often have at their root neglect of widely-known common-sense limitations of markets versus theories about markets.
The "black swan" criticisms of Taleb are also relevant to the discussion.
(Score: 2) by aclarke on Tuesday November 24 2015, @06:26PM
I came here to say something similar. I've been messing with foreign exchange trading for a few years, and have done a bit of data modelling as what I would best describe as a hobby at this point. Even I am well aware of the dangers of curve fitting.
If I read an article that doesn't understand the concept of curve fitting, I wouldn't even really consider that it had been written by a "professional" in the field. "PSMag" isn't a site I've heard of before, and it certainly doesn't look like a financial journal of any sort. I'd chalk this one up to someone trying to write an anti-investment article to an audience of people who don't understand the basics of technical analysis.
On the other hand, history has shown us time and again that the so-called experts really DON'T know what they're doing much of the time. Maybe I'm not as far behind as I thought, which would be a pretty depressing commentary on the state of financial investing as I really don't think I'm very knowledgeable.