The Murdering Hitler Portfolio

Bad junk food tastes great; until the after effects kick in hours later.

Similarily, flashy financial analysis will have you happily plunging money into a shiny new scheme, only noticing the stomach churning results afterwards.

Let's take a look at some whizzy analysis that financial experts show again and again.

Typical backtests replay an investment strategy through history and show the results.

For example,

The lovely picture is summed up in even nicer, nerdier analytics.

Somewhere, a finger is now hovering over the buy button!

But then ask yourself - who invests in any product for 10 plus years? The backtest is impressive, but does it make sense?


The last 10 years will be fundamentally different from the next; and the next after that.

In fact we have multiple parallel universes ahead of us. Possible outcomes that we can never imagine

Like a historian imagining himself in 1933 and wondering what would happen if he assassinated Hitler? Or ask yourself what would happen if you could nudge Babbage into creating the first computer 100 years before Turing.

That's a guilty pleasure for historians; but financial analysts are not historians. It's our daily bread and butter.

We need to generate future parallel universes, and see how our investments fare.

A simple way of doing this is splitting up our historical data into discrete chunks.

Each chunk in this case is a yearly return of Apple stock and a future parallel universe.

Before proceeding any further, we need to ensure that each universe is more or less insulated from the others.

An easy way of doing this is plotting each return against it's previous return.

If we fit an ellipse and it looks somewhat circular we are OK.

Now we can bucket our 30 or so individual returns.

Let's compare our parallel universe analysis with the more traditional analysis.

Our 'Apple strategy' returned 16% per annum on average using the traditional analysis - 17% percent using the analysis presented here.

However, now we can see that our return will only be in that ball park less than one year in five!

This analysis underlines the huge downsides to investing in Apple stock, with losses of over two thirds possible (log losses of more than 100%) during a dot-com-like bust of 2000.

Traditional analysis, sums this possibility up with a volatility figure. Our parallel universe analysis has 30 of these! How can we use them?

Rather than summarising with a single Sharpe ratio we can now see 30.

Summing up the last 30 years of Apple stock returns in the traditional way, gives a single Sharpe ratio of around 0.3.

There's really only about a one in eight chance of being in that ball park. In fact the Sharpe ratio has a very wide range.

A single Sharpe number for this 'Apple strategy' falsely gives us a precise picture of the future.

False precision is perhaps the most common problem with financial analysis.

More data is being crunched; increasing the amount of inaccurate precision published.

Ironically more data should lead to a fuzzier more nuanced picture of the future, which makes for an honest and, unfortunately, harder sales pitch.