Using Software to Gauge Hedge Fund Risk

I had an interesting chat with Olivier Le Marois, the CEO of Riskdata, a couple of days ago. His company has put out a couple of interesting reports: one on blow-up-prone hedge funds generally, and one on Bernie Madoff in particular. In both cases, Le Marois claims that his software, which looks only at the data that any investor or potential investor has access too, would have raised lots of red flags, without any need for deep forensic analysis or intrusive investigations of operational risk.

Le Marois is the first to admit that red flags can be a false alarm. But if you avoided any funds with his red flags, you would pretty much have avoided all the most spectacular blowups. You might have thereby stayed out of funds which were posting what seemed to be very good risk-adjusted returns, "but risk management has a payoff," he says. "Like an insurance policy, you are very at ease if you never see it materialized. You might slightly underperform your peers. But you get rewarded for that when things go really wrong."

Le Marois says that just by looking at a fund's published returns and its stated portfolio and strategy, he can tell with surprising reliability when it's doing something dangerous. Such activity falls into one of three buckets, he says.

The first is simple manipulation of the returns, where managers smooth out their returns by "storing" outperformance from excellent months and applying it to bad months. In that case, you nearly always find a huge discontinuity at zero: There are many more small positive months than there are small negative months. Which is a huge red flag. And there are other indicators, too, which are built into Riskdata's bias ratio.

Riskdata's numbers are quite impressive: Out of a universe of 2,281 funds, 20 have had a bias ratio of more than 6 on Riskdata's scale. Those 20 included all six hedge funds -- including Fairfield Sentry, the Bayou Superfund, and Beacon Hill's Safe Harbour fund -- which proved to be outright frauds.

The second sort of dangerous behavior involves simply being lucky rather than fraudulent. If you were simply long credit between 2003 and 2006, you made lots of money -- but carried much more risk than your volatility numbers would suggest. Most relative-value fund managers, says Le Marois, are playing this game: relative-value trades are generally based on mean-reversion. But as LTCM famously found out, they can blow up spectacularly. Again, this kind of arbitrage-driven management style is quite easy to detect, and often fund managers are happy to say that's exactly what they're doing.

Finally, there's the classic Capital Decimation Partners strategy of simply selling out-of-the-money puts. This is also relatively easy to detect: You look for funds which are short gamma.

Long-gamma funds, says Le Marois, tend to be much safer places to be than short-gamma funds when a blow-up occurs. Many macro and commodity funds are long-gamma: They're long when markets are going up, and short when markets are going down. That doesn't necessarily mean they'll outperform, but it does mean they're much less prone to disaster.

Of course, there is still operational risk: Riskdata's system doesn't capture that. But "it's easy to diversify operational risk," says Le Marois -- simply invest no more than 3-4% of your funds with any single manager.

I don't for a minute think that Riskdata's system is a one-stop off-the-shelf alternative to hiring an experienced risk auditor to really take a long, hard look at any given fund. But it's a hell of a lot better than nothing. And it can be applied not only to individual funds but also to your entire portfolio, so that you can see when you're overinvested in certain strategies or are taking on more overall blowup risk than you thought. So it sounds like something helpful, which is likely to do relatively little harm so long as you don't rely on it to the exclusion of anything else.

At the end of our conversation, I asked Le Marois what he thought of the Sharpe ratio. He didn't need any prompting:

Sharpe ratio is the most stupid way to compare funds, and is a huge incentive to invest in Madoff-like funds. CAPM is absolutely outdated. It makes sense when you have very simple risk patterns, when you have Gaussian distributions, and no derivatives and no illiquid assets. But the current world has nothing to do with that world. People have highly asymmetrical payoff structures. There are plenty of illiquid securities. All this makes it very easy to get around this kind of evaluation ratio. We have to acknowledge that we are living in a complex world, and stay away from an oversimplified systematic view.

All of which makes sense to me. After all, a low high Sharpe ratio is something that both LTCM and Bernie Madoff had in common. Which hardly commends it.

Comments

[morph366]

An enjoyable piece but there is a kitchen sink quality to the criticisms from Mr Le Marois highlighting the shortcomings of the Sharpe Ratio. It is a very simple ratio and certainly suffers from all of the usual issues regarding the fallacy of using a Gaussian distribution to analyze financial returns but claiming that it cannot address the issue of how to assess the risk of holding illiquid securities seems a bit unfair.
There are other ratios that show the trade-off between maximum drawdown and financial returns that are more useful than the Sharpe Ratio but - are there any metrics that allow us to measure how much risk there is in holding assets with low liquidity especially in the limiting case, as for so much of the stuff which the banks are holding, of zero liquidity?

2009 Jan 24 08:05 AM

[Starlin]

Unfortunately, I find Mr. Le Marois comment on the Sharpe and Gaussian distributions in general a little off-setting. Not because he is not right - both of these concepts have been shown as lacking long ago (there are seminal papers on better approaches by Mandelbrot and by Fama in the 1960s! - for more resent discussions see Rachev's "Stable Paretian Models in Finance"). Rather, the comments strike as hypocritical as RiskData's system is based on these very concepts! Making slight extensions and (very) often hiding the Gaussian distribution in pretty marketing names doesn't produce a better risk picture - it just masks the problems imposed.

2009 Jan 24 12:46 PM

[ROLEX18K]

I love Mr. Bernie Madoff, he went against elite of millionaires and billionaires who respected him and he tricked them for so long.
I think he is a hero as the people he bankrupted and broke, themselves made their millions and billions stealing and lieying on the way up.
This is how the world of hedge funds operate, 99% of it is scam that delivers about the same return as SP 500 long term, only they charge for it minimum 2% a year management fee to pay their expenses like offices in best locations and between 20-50% from profits.
This crash proved that hedge funds are scam and Ponzi scheme en masse.
Never give your money to anyone, keep it only in your name in your account and already if you are fooled to invest in a hedge fund, ask them one very simple and normal question:
Guys, I want to make money just as everybody else, I will give you POA (power of attorney) on my account and you will be permitted to trade whatever you like but I will watch it, if I like the results I keep positions, if I don't like I call my broker and say - liquidate everything.
One more thing, never let no one to trade in things you don't understand, if the strategy looks to you as strange, some kind of OTC (over the counter) swaps, CDS', CDO's or anything that is not traded on the exchange- keep away from it.
Any hedge fund that will agree to your conditions, passed the test, any that doesn't keep away and be lucky that you escaped the "hedge trap".

Mark Medayski

2009 Jan 24 03:15 PM

[quaazy1]

So, I have a general question concerning the analysis of the HFRI FOF composites. The annualized standard deviation and return of the historical data for the HFOF Conservative Index is StDev 4%, Ret 7%. There is a loss of 18% over the last 12 months within that data. That's something like a 6 standard deviation event. Is this evidence that one should not use the "Index" as the model (source of return and volatility estimates) when considering whether or not to include a conservative hedge fund in your portfolio in a classic mean variance optimization? It seems to me whether or not you believe in CAPM, some how the volatility of that index was dampened. Maybe it's how it is constructed??

2009 Jan 26 04:49 PM