Comments on: Schmalpha

By: The Particle

The Particle — Fri, 22 Feb 2008 19:25:18 +0000

Pay attention to those small bubbles. Eventually, you will find one larger than the rest. As for the rest, they will settle into a state of equilibrium. Either way, you can win.

By: The Epicurean Dealmaker

The Epicurean Dealmaker — Thu, 31 Jan 2008 02:51:50 +0000

Nice post, and nice thread. I would only add one remark, which you are free to interpret in whatever way your individual philosophy permits you to:

Luck is its own reward.

By: baruch

baruch — Tue, 29 Jan 2008 21:14:31 +0000

Absolutely. But then so much is, isn’t it?

Especially in markets.

As I am sure you have noticed this IS a philosophical site, devoted to a dead dutch excommunicated jewish philosopher of political and religious freedom. If we can’t comment on this, then no-one can, I guess.

By: nodoodahs

nodoodahs — Tue, 29 Jan 2008 17:33:58 +0000

Question (2) may be unfair, but so is life.

I could have the right rate for a risk over the long term, but still lose money writing it in any particular year – or string of years. Without a “god’s eye view,” we will never know.

The wisdom of rewarding what we don’t know is or isn’t luck is a philosophical one, isn’t it?

By: nick gogerty

nick gogerty — Tue, 29 Jan 2008 13:55:40 +0000

just short the gamma in a few way out of the money 9-10 “non-correlated” options markets, cash the checks fast. The whole thing is an assymetric barrier option structure that some banks like to believe in.

By: baruch

baruch — Tue, 29 Jan 2008 08:21:26 +0000

Concurred 99.9%, nodoodahs. Your question (2) about luck may be a bit unfair, however. Some traders make money consistently and are actually not sure how they do it. Their enemies may call this luck. In the absence of any ability to separate out luck from skill, may it not be wise to reward luck as well, just in case it isn’t actually luck?

By: nodoodahs

nodoodahs — Mon, 28 Jan 2008 20:37:07 +0000

Outside of a regression equation vs an appropriately-chosen benchmark, the terms “alpha” and “beta” are meaningless. The current over-mis-use of the words has resulted in a curious form of financial jabberwocky, wherein everyone uses the words but nobody knows what anyone else means. Interestingly, when used properly (as the axis-intercept and slope of a regression line), the words have a wonderful, concrete interpretation that is filled with explanatory power about a manager’s returns. Shame nobody uses them properly …

The main issues are the same with almost any compensation package in the financial industry:

(1) can you know what the reality of the result is in a timely enough manner to reward the manager?

(2) to what degree is luck associated with these results?

Throwing buzzwords like “alpha” and “beta” at the problem doesn’t help a bit with those items. I suggest that they are insoluble problems, if one is looking for perfection. I suggest one looks for the solution that is the least vulnerable to the manager being able to “game” the system, and let the search for the “perfect” metric stop.

By: nick gogerty

nick gogerty — Mon, 28 Jan 2008 17:39:41 +0000

Gott’s theorem is fun to use for creating temporal assumptions for any phenomonae. Much more fun then extreme value approach and can almost double as a parlour game. here is a spreadsheet to play with.

http://nickgogerty.typepad.com/designing_better_futures/2007/07/predict-anythin.html

By: baruch

baruch — Mon, 28 Jan 2008 10:26:15 +0000

Well I give you airlines, and indeed cars keep getting rescued too.

However I am with you and Wolf on this one. Banking is a terrible business model, huge multi-year profits wiped out in an instant. I wouldn’t touch them with anyone’s money.

I think some of them should be allowed to fail, the depositors compensated by insurance. It would be a salutory lesson, and encourager les autres sufficiently to not be such muppets.

HOWEVER, we are talking here about traders’ pay structures, and Wolf’s and Rajan’s proposals betray a lack of understanding about how things work in the real world which make forcing them into areas where they are not appropriate* (they exist already in many areas of finance) a really bad idea.

*Especially MY pay structures.

By: interested

interested — Sun, 27 Jan 2008 22:06:48 +0000

I believe Martin Wolf put it the best when he suggested that banks have a gift for taking all the profits of their risk and socializing all the cost. I believe that no one would begrudge banks for taking huge risks if they were also allowed to fail just like any other business that does stupid things. However it appears that central banks always come to the rescue, with the serious problems of moral hazard today.
I think that the real problem is the sense that people have of unfairness, a sense that there is one set of rules for the bankers and another set for every other business (e.g. car manufacturers allowed to go under).

By: Baruch

Baruch — Sun, 27 Jan 2008 08:12:39 +0000

Don’t be surprised you have to explain this, I did history at university, not statistics. I got the wife to explain this to me and she says you are right. She knows about this stuff.

However, I stand by my observation that the standard stock beta (most commonly derived by the BETA function on Blooomberg) is not helpful in forecasting stock returns, and think I supplied sufficient support for thinking that, without needing to get on my Talebian hobby horse. My chair has many legs. If beta is not effective looking forwards, is it fair to use it looking backwards to judge people’s performance?

Revised, and resubmitted. Baruch does not persist in error!

By: wcw

wcw — Sun, 27 Jan 2008 02:19:53 +0000

No rational variance measure yet devised completely elides a distribution’s tails. To the extent beta is a variance measure (which you have half-right: that metric combines correlation with relative variance), it does not ‘exclud[e] the outcomes that lie beyond the standard deviation.’ Variance measures may not capture the full richness of real-world distributions, but that doesn’t mean they throw away data.

I am surprised I have to explain this.

There are a surfeit or real problems with the First-Semester MFE conception of financial markets modeling in general and performance analysis in particular. Do we really have to make things up in order to mock them?

I like where this one was going, but I give it a revise-and-resubmit.