Apr

14

Bloomberg news picked up this article. I am not endorsing the paper, its methodology nor its conclusions. But counters should heed the underlying message. Especially Kora. I find it surprising that he doesn't look at the multiple comparison issue nor cite Bonferroni etc, but rather prefers to ask the question, "what is chance that a backtest generates a great result by chance." He argues that if you use 10 backtests, you are very likely to find a strategy with a Sharpe Ratio of 1.6 which is over-fitting: "Pseudo-Mathematics and Financial Charlatinism: The Effects of Backtest Overfitting on out–of-sample Performance" by David H. Bailey, Jonathan M. Borwein y Marcos Lopez de Prado z Qiji Jim Zhux, April 1, 2014

What good is a hypothesis that cannot be disproven? A Cautionary Tale (In Memory of Ross Miller)

1. Kora observes: Y = Fn(X) with a significance of T.

2. Kora raises a small amount of investment capital based on the expectation of this stochastic function alone. She gives no consideration to dynamic or causal or other exogenous relationships or intellectual or information edge.

3. Kora produces excellent performance as Y= Fn(X) as predicted.

4. Kora raises a massive amount of investment preformance after establishing a track record.

5. After raising a large amount of capital and collecting substantial management and incentive fees, something happens and Y <> Fn(X), and the clients suffer horrendous drawdowns. The fund shuts down and the total net amount of loss dwarfs the net amount of gains.6. The SpecListers say, "The probability of this was extremely small. But it is an example of Bacon's Ever Changing Cycles." Rocky says, "This is a example of bad science because any utility of the observation Y = Fn(X) without a casual understanding is limited to and qualified by, the ability to anticipate the onset of a changing cycle. And if the scientist can correctly anticipate the onset of a changing cycle, then this meta-hypothesis is vastly more important than the functional hypothesis.

Unfortunately, this is a recursive paradox, because the ability to anticipate the onset of a changing requires the ability to anticipate the onset of a changing cycle of a changing cycle, and then the onset of a changing cycle of a changing cycle of a changing cycle … and this continues ad infinitum OR UNTIL spec partiers go home to bed — whichever comes first."

Jordan Neumann writes: 

I admit not to have fully read the paper — I searched for the word transaction cost but did not find it, yet it makes finding a profitable strategy much harder than it seems.

Isn't this a problem with statistics in general? How does this differ from using thousands of drug candidates to find a drug? We still don't know why Advil works, but I take it anyway based on the statistical evidence. When quants believe that earnings or margins or insider trading affect prices, I would say that the economic justification is far from random.

There is a recent series of news articles that disparage quantitative analysis, just as several quant funds suffer for a few bad years. I would think that everything moves in cycles, and this might be the bottom.

Hernan Avella writes:

Mr. Rocky offers some valid questions to the counting battalion. However, I'm afraid his argument suffers severely from the straw man problem. It assumes that one can't have an approach that incorporates: logic, an economic framework, money management rules and counting. Even more. As you move up in the frequency spectrum, the economic framework becomes optional (useless).

The real question is (for med/long term speculators). If you incorporate all the said components in your approach, can you quantify your success per component?

Ralph Vince writes: 

Kora,

Yes, in my humble opinion, more money is to be made on the assumption of EMH (the cost of being wrong in this regard is less).

Stefan Jovanovich writes:

The test of the reality of a market is whether or not there are prices for quantities exchanged in actual transactions; and the market itself is sufficiently profitable that dealers are willing to pay for the rent and other costs of keeping the lights on. Market failure happens all the time; a trade disappears because other markets have swallowed the action or the inter-mediation itself is no longer handled by bid-ask. Even now more than a century and more after they disappeared you can find the remnants of "corn exchange" buildings throughout Britain; dealing in grain continues but it is no longer handled by open outcry involving dealers and farmers within half a day's train travel of a regional hub.

Markets are efficient in the way that engines are efficient in that they work. They are inefficient in the sense that there is wasted energy, some or much of which can be the result of insider manipulation and general fraud. The debate is over numbers matter - the economics of the companies and the world of money as a whole, the prices themselves and their patterns, the numerical indices of sentiment; for that question there is no absolute answer, nor should there be. Larry Williams, the R-Man, the Watsurf, RPH and many, many others can all be right - and wrong. And, in that sense, markets are permanently inefficient because, even among people to whom Morgan would have assigned a perfect grade for their financial character, the only final word comes when the market itself disappears. 


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search