STRATEGY
1. THE STRATEGY
I believe that any large research project should begin from the premise that nothing is known for certain about the subject. All preconceptions should be laid aside so that the truth can reveal itself, sometimes in unexpected ways. Initial work for Theseus began by selecting four large, and relatively uncorrelated market types, and sourcing as much historical data for each as possible. By analysing many different markets at the same time, we can look for inherent principles that have universal validity. These were the markets chosen: The hardest data to source was the Dow Jones. While data is generally available for 100 years, I wanted to go back further, and was eventually able to obtain two other reliable sources of earlier data, one from 1885 to 1896, and the other from 1896 to 1914, and by adjusting these, they could be accurately merged with later data. The Dow was closed for over 4 months from the outbreak of World War One in July 1914, and then reopened in December that year. Having data from such different markets, across different date ranges was very valuable to the process. The data itself consisted of intraday and end of day closing prices, including open, high, low and volume where available.
OBTAINING FUNDAMENTAL DATA
I believe that any large research project should begin from the premise that nothing is known for certain about the subject. All preconceptions should be laid aside so that the truth can reveal itself, sometimes in unexpected ways. Initial work for Theseus began by selecting four large, and relatively uncorrelated market types, and sourcing as much historical data for each as possible. By analysing many different markets at the same time, we can look for inherent principles that have universal validity. These were the markets chosen: The hardest data to source was the Dow Jones. While data is generally available for 100 years, I wanted to go back further, and was eventually able to obtain two other reliable sources of earlier data, one from 1885 to 1896, and the other from 1896 to 1914, and by adjusting these, they could be accurately merged with later data. The Dow was closed for over 4 months from the outbreak of World War One in July 1914, and then reopened in December that year. Having data from such different markets, across different date ranges was very valuable to the process. The data itself consisted of intraday and end of day closing prices, including open, high, low and volume where available.
FUNDAMENTAL DATA INSIGHTS
Preparation of the fundamental data was a lengthy process, because of the necessity for validation against secondary sources. In addition, a framework had to be created that could analyse different kinds of fundamental data for each type of market.
The work was divided into three separate areas:
​
-
Seeking direct correlations between individual fundamentals and price change
-
Predicting future fundamentals from past fundamentals, independent of price
-
Identifying complex correlations between multiple fundamentals and price change
Results from all three methods were disappointing. In fact, rather than calling them disappointing, let's say they were inconsistent. You could certainly cherry pick excellent multi-year periods, but by extending the timeframe out to multiple decades, the performance became erratic.
"Most of the periods of exceptional gains in each of the markets were not well indicated by the fundamentals"
Fundamentals tended to do better during downturns or economic crises, or after major corrections in the markets.
​
These findings tally with an analysis of Warren Buffett's Berkshire Hathaway, which has always been driven by fundamentals and value-seeking. They missed out on stupendous gains in the late nineties during the tech boom.
They also failed to capitalise on much of the gains in the growth-driven, tech-heavy environment of the past 10-20 years, led by companies such as Amazon, Google and Tesla.
​
Undeniably, they have generated exceptional returns from investing in profitable 'value' companies like American Express, Geico and Coca Cola from the 1960s onwards and increasing their holdings substantially over multiple decades.
Looking at the results from the fundamental analysis, I reflected on the human need to attribute reasons to the way prices fluctuate in markets. You see it in the headlines on financial websites.
In the morning you might see something like
"DOW falls on disappointing GDP growth"
and later in the day, when the DOW is back in positive territory the headline might change to
"DOW rises on anticipation of interest rate cut"
Either might be true, or neither might be true.
After an underwhelming start with the results from fundamental analysis, the project switched to analysing price itself.