I haven't got the foggiest data

The first full week of the year has passed. Christmas decorations have come down. Lazy morning starts are fading from memory. In their place, has come the hectic tempo of early morning commutes to work, the flip of 2015 to 2016 in the calendar, that feeling of starting all over again.

In markets, the volatility which had been absent over the holiday period, has returned. The first week has seen stocks sell off, in particular in China. Whether the Shanghai composite is of key importance for world markets is another question (after all, international markets mostly ignored their bubbly rise last year and Chinese stock market is dominated mostly by local retail investors). Geopolitical tension has increased in the MidEast, which only failed to stop crude oil's continuing decline for a couple of an hours.

Does the first week of the year in markets have any significance on the rest of the year? I recently ran a simple test, plotting the returns from S&P500 during the first week against the rest of the year. Whilst in the past decade there was at least a passable relationship, in the decades before that, it's very difficult to spot any relationship. The difficulty is that we have a comparatively small number of points to test this idea upon, given a trading rule would involve only one trade a year.

We can come up with other examples of this limited data problem in markets. For example, if we are trying to create a model to estimate when (or if) a managed currency might experience a sudden regime change. Rather than attempting to precisely time such a difficult binary event (which is near impossible!), we can instead try to build a probability distribution for that event. In our managed currency instance, we could look at central bank reserves data and other critical economic variables. We can then compare the market pricing for such an eventuality and compare to our model. Indeed, this approach looking at market pricing and also modelling other market variables is the approach I took in recent research on the peg of USD/SAR (see my interview here on the subject).

Should we never do analysis when we have very small amounts of data, given the problems? I would argue not. Once we have a probability assessment of our model, we can then overlay our own judgement on top of that and compare to market expectations. Analysing very small datasets might help us see a bit further into the fog of the future: after all it is likely better than doing nothing! At the same time we need to cast a critical eye on the output of our analysis.

Like my writing? Have a look at my book Trading Thalesians - What the ancient world can teach us about trading today is on Palgrave Macmillan. You can order the book on Amazon. Drop me a message if you're interested in me writing something for you or creating a systematic trading strategy for you! Please also come to our regular finance talks in London, New York, Budapest, Prague, Frankfurt, Zurich & San Francisco - join our Meetup.com group for more details here (Thalesians calendar below)

20 Jan - London - Nick Firoozye - Managing Uncertainty, Mitigating Risk
29 Jan - Budapest - Robin Hanson - The Age of Em: Robots
08 Feb - London - Saeed Amen/Delaney Granizo-Mackenzie - CTA/Pairs trading (joint Thalesians/Quantopian event)
29 Feb - London - Jessica James - FX option performance (TBC)
21 Mar - London - Robin Hanson - The Age of Em: Robots