If you know the enemy and know yourself then you will not be periled in hundred battles – Sun Tzu
This book was recommended to me by a friend as a quintessential “quant” book. Curiosity got the better of me and I decided to read it.
There are two very strong argument for using computers to automate trading. The first is that the computers have no emotions and hence they are not inclined to make emotional/behavioral errors. The second is that once they have a set of rules to follow they never veer outside their limits. Humans or discretionary investors have famously committed both types of errors.
I was surprised at the design of trading models. Overall they are very commonsensical and the aura created around quants is largely misplaced.
A quant trading strategy can be divided into 5 parts i) alpha model, which is the starry eyed optimist trying to find ways to make money, ii) risk model, the pessimist who wants to limit the downside, iii) transaction cost model, which figures out how much the transactions will cost, iv) portfolio construction model, which takes input from all three models and decides what changes to make in the current portfolio, and v) execution model, which executes the trade in the most cheapest way.
The book then treats each part of the trading in some detail. Going into the problems one faces and ways to solve them.
What was surprising to me was the amount of stuff we can already use as a value investor. There are “value” oriented quant strategies which use fundamental data to decide about the quality of a stock. Something similar is done by Vuru. The best use of such a strategy for us is a stock screener.
Many of the strategies that we know are already used by quant trading strategies. In fact, all the observations we have made in our investing life can be tested on data and then implemented. For example Steve Romick has gone long Renault and shorted Nissan. This is called “pair trade” in quant terms and is used to limit the risk while gain from the underlying “fundamental observation”. If you think that XOM is cheap then going long XOM will still make you lose money if the whole stock market goes down. But if you think that XOM is cheap while CVX is expensive then going long XOM and short CVX will remove the risk of market going down. It does still make you lose money if the relative premium continues or increases.
All of this and similar strategies were already in place at Long Term Capital Management and they went bankrupt. Rishi addresses this issue at a few places in the book. He does not think that LTCM was a pure quant based strategy – which is arguably true. In Rishi’s words
They were engaged in a very broad, cross-border and cross-asset class yield game in which they constantly sought to own risky assets and sell safer ones against them. It was, in most respects, a highly leveraged, one-way bet on ongoing stability and improvement in emerging markets and the markets in general.
I would not say that this is a sufficient reason to discount LTCM as a quant firm. Any trading strategy has an underlying philosophy. The “qaunt” part comes because of quantitative ways to measure risk and the amount of leverage one should take. It is quite clear that LTCM’s leverage was based on “quantitative” ways to measure risk.
He then moves on to address the 2007 fiasco, which in his opinion is closer to home. He claims that this was caused by i) the “crowded trade” effect ii) poor year-to-date performance, iii) cross holding of illiquid credit based strategies, and iv) use of VaR based volatility targeting and leverage adjustments.
In my opinion, this is using a lot of words to say that their model was faulty. I don’t see how they have fixed the problem. Is there a way to do so ? In my opinion – No.
The book is excellent. If you want to get a good introduction on quants and what they do – this is a very good book to start with.