Financial markets: Are we playing chess or poker?

As Warren Buffett and a million other fellows have said, “If you can't find the sucker at the table, you're it.”

?

Interestingly, Mr Buffet refers to poker rather than chess.

?

In 2004, I moved to Paris to become a Credit Trader at Fortis Investments. Within six months, I was spending the vast majority of my time trading CDS contracts to form bespoke CDO tranches. These tranches were constructed based on an underlying portfolio that usually consisted of 125 companies, and thus 125 individual CDS contracts. The size of each contract was calculated using a Gaussian copula in the form of a correlation model. Thankfully, I worked with a team of gifted mathematicians including Sébastien De Kort, Ryan Lemand, PhD and Julien Houdain, PhD. They painfully guided me to a sufficient level of knowledge to make me dangerous, but by no means fully understand all of the concepts and certainly only a limited amount of the mathematics. As a team, we went on to build CDO^2 (a CDO containing several sub-CDOs) and CPPIs (Constant Proportion Portfolio Insurance). Ramping up the CDO^2, which involved four sub-portfolios, two maturities, and trading circa $2.5 billion in single-name CDS contracts (my recollection may not be accurate on that number) was among the most difficult experiences in my career.

?

Fortunately, some of my team became close friends and also poker companions.

?

In my role at Fortis, I was at the coal face of trading each of these CDS contracts. I expected market makers to seek to maximise their bid/offer spread. However, with greater experience, I came to understand that every bank had a different delta, which their model generated and thus defined the notional amount of CDS that was required. In essence, the banks attempted to increase their profit by moving their delta. In time, I came to recognise that the role of banks is simple to maximise their short-term profits, and thus one should always expect them to use all possible tactics. All is fair in love and war!

?

However, a more important seed had been planned in my mind. I came to question the efficacy of the models. To go down the rabbit hole, following the GFC, the recovery assumption in the correlation models was changed from fixed to stochastic so that they could continue to function under the new paradigm.

?

Concurrently, I also become aware of many of the structural flaws in fixed income and credit strategies, in particular:

1.???? Negatively convex bonds, derivatives, and portfolios

2.???? Tail risks

3.???? Credit assets are short liquidity (liquidity decreases as the price of bond falls)

4.???? Gap risk (prices can fall in a vacuum where potentially no price is tradable. This is particularly acute for large positions as was witnessed with the London Whale).

?

For the next decade, the frailties of financial models lingered in my mind. I am not nearly bright or talented enough to prove the thesis. However, I am curious and stubborn enough to continue the investigation. Furthermore, as more of the greatest minds in finance have published books, research, and given interviews, they have revealed the vast number of inefficiencies in financial markets and the optimal strategies and approaches to consistently delivering superior returns while mainly avoiding large losses. Thus delivering the mythical alpha.

?

As I have read further including following Howard Marks' public advice to read Annie Duke's seminal book 'Thinking in Bets', it became apparent to me that not only are financial models mainly driven by subjective inputs, but that many of the people applying these models believe they are playing chess. As Annie Duke stated, "Chess contains no hidden information and very little luck...Poker, in contrast, is a game of incomplete information". She goes on to write, "Incomplete information poses a challenge not just for split-second decision-making, but also for learning from past decisions". While the market leaders have accumulated vast pools of performance data and thus gained extraordinary insight into the PnL and risk of individuals, teams and their entire business, this level of analysis is far from the industry norm.

?

While CDOs have all but disappeared, financial models are ubiquitous in pricing a range of securities. Our focus has been on the pricing of convertible bonds and single-stock options. In Jack D. Schwager's book Hedge Fund Market Wizards, Jamie Mai identified five generally accepted assumptions that are sometimes invalid:

1.???? Prices are normally distributed

2.???? The forward price is a perfect predictor of the future mean

3.???? Volatility scales with the square root of time

4.???? The trend can be ignored in the volatility calculation

5.???? Current correlations are good predictors of future correlations

Jamie Mai also said,"Option math works a lot better over short intervals. Once you extend the time horizon, all sorts of exogenous variables are introduced that can throw a wrench into the option-pricing model". Considering that many convertible bonds are longer than two years in duration and that many of the options are deep ITM or OTM, it is fair to expect that a large number of 'exogenous variables' are introduced. He also states that "often, the longer the duration of the option, the lower the implied volatility, which makes absolutely no sense". This too is important when pricing the call option that is embedded in a convertible bond.

Since 2019, Fairwater Capital LLP has been employing PhDs from STEM disciplines to firstly aggregate and clean data, and then to interrogate the data to understand the relationships between listed securities. It is only through their extensive work that we can now identify with clearer evidence that not only do financial models have flaws, but that often market participants use subjective inputs into their models. This results in high dispersion in returns. Some investors make large returns, while others make large losses, and the majority of investors are in the middle. However, over time, this dispersion leads to a higher correlation with traditional assets, inconsistent returns, and larger drawdowns.

?

For a wide range of psychological reasons, it is easier for many investors to convince themselves that there is no hidden information, that the markets are efficient, and that the assumptions that the models are reliant on are acceptable. Thus, they often believe they are playing chess when in reality they are playing poker.

?

I love playing poker, whether it is with my oldest friends in Sydney, my colleagues and friends in Paris during the years at Fortis IM, or with all sorts in Las Vegas.

?

However, my aspiration as an investor is to deliver repeatable alpha. By recognising the flaws both in the data and the models, we have sought to complete the information where possible, allowing us to use objective inputs, and thus start to play chess.