Passive Risk Management Tools

Almost a month ago, I wrote this (second) review of Wall Street Journal reporter Scott Patterson's recently-published book, The Quants.

The book's early chapters, on which I did not dwell in my review, discuss Ed Thorp's early exposure to Claude Shannon at MIT and, through Shannon, the work of John Larry Kelly.

Thorp was the prototypical quant hedge fund manager, but, in a sense, he simply followed, then outstripped, Shannon's and Kelly's own instincts of applying information theory to gambling opportunities, or any other situation in which imperfect information allowed for profitable employment of risk capital.

As I sped through the early chapters of Patterson's tale, I noted the mention of Kelly's name and general idea, but, since Patterson never focused on it, neither did I.

Fortunately, one of my friends and former colleagues, once a member of Bell Labs' EMT staff, reads this (and its companion political) blog on a daily basis. Thinking I'd find Kelly's work useful, he lent me William Poundstone's Fortune's Formula, published in 2005. The linked page on Kelly is from Poundstone's website.

Here is a paper Ed Thorp wrote concerning Kelly's Criterion over a decade ago.

My friend has been considering applications of Kelly's Criterion to his own work involving defense project management, so passing on the references and book to me were topical for him.

When I began reading Fortune's Formula, I found it engaging, but not particularly compelling. Until I came to Poundstone's painstaking explanation of how elegant Kelly's application of Shannon's theory was in the area of risk capital investment, be it at the track, casino, or securities markets.

Unlike the unsuccessful active risk management methods applied by traders before and during the financial market meltdowns of 1987, 1998 and 2008, Kelly's approach is passive in nature. The difference is important.

As Patterson portrayed in the latter chapters of his book, the major quant hedge funds employed pretty much similar VAR-based risk management systems. These systems naively assumed that, in the event of sharp losses in portfolio positions, there would exist a relatively unfettered ability to sell, or buy, the offsetting leg of existing portfolio positions, in order to hedge exposure.

Patterson didn't really dwell on the details of Ed Thorp's original employment of what Poundstone correctly identified as 'delta hedging.' It's an active risk management approach for adjusting hedges between equities and the companion warrant or option position.

Thorp originally did this in relative isolation and obscurity, with little in the way of market pressure from similar strategies swinging tens or hundreds of millions of dollars of such activity.

In the 1980s, the variant of active risk management was known as portfolio insurance. And it failed in 1987. When major trading operations all attempted to hedge the plummeting values of their positions in major indices, the values of which they were arbitraging, values plunged even further and orders went unfilled.

In 1998, similar results occurred in the last gasp of pre-swaps markets amidst a panic. In that era, there were quite a few statistical arbitrage clone funds, but only LTCM was actually brought down by the market plunge.

The 2007-08 financial market meltdown featured newer instruments, e.g., CDOs and swaps. But, in reality, these were simply variants of older financial products, albeit inherently possessing greater risks.

And, thus, when market values reversed their ascent, and active hedges became unbalanced, the usual stampede of many managers employing similar strategies to sell additional exposure in order to re-align their hedges once again failed.

Among my more experienced securities markets friends, and me, we agree on the notion that risk management solutions need to be passive and in place before a market event, because active risk management approaches never actually work as planned. A sort of 'fallacy of composition' defeats them, as abstruse investment strategies, employed in parallel by many smart quants, simultaneously react identically, pushing market values even further in the direction that is already causing them such large losses.

Kelly's Criterion is the ultimate passive risk management tool because it addresses the size of the investment, and is not affected, per se, by the actions of others. In fact, one of the valuable extensions of Kelly's work by Thorp was his apparent refusal to leverage his funds.

This was something ignored by the next generation of quants, including his own acolyte, Ken Griffin, founder of Citadel.

Reading Poundstone, on Kelly, provides a very clear, refreshing portrait of a seriously intelligent, rigorous mind grappling with the realities of risk capital deployment. And Thorp, with his casino beginnings, understood this perspective intuitively.

Sadly, quant kids two generations and more on, do not.

And that is why active risk management has been so ineffective. It becomes not a matter of discretion, but of survival.

That is, in Thorp's world, trimming the exposure to a Kelly bet is advantageous, but not essential. Because the bet was scaled according to rigorous theoretical principles, losing it all isn't crippling.

When that bet is with 3x, or more, levered money, you only have a 25-33% loss cushion before you've burned through equity and are out borrowed money.

The gambler's ruin. Busted.

Needless to say, we've already begun working on applications of the Kelly Criterion to our proprietary options investment strategies. As a passive, rigorously theoretical and internally-focused approach, it meets our criteria.