Data Science for Algorithmic Trading

It is likely some other underlying causal factor like oil and metal prices that creates this dynamic, but it was still good to have a look inside some of the data to verify that it makes sense.It turns out that there were some very strong correlations in the 2010s between some IPPI factors and USD_CAD..We continue to see car stuff like SUVs, and also war stuff playing a big role in USD strength..I think there is an underlying message about 2001's September 11th attack and the resulting wars boosting the USD both as a safe haven and as a military power..I think that had an overhang into the 2010s..These correlations pushed the lumber story out of the top results..Although the dispute is still not resolved, many pulp and paper mills in Canada simply shut down in the 2010s.We see in the chart below that oil and related products as well as metals continued to be a part of the negative correlations for USD_CAD.During the 2010s there were a lot more factors available in the dataset..The following chart shows the distribution of USD_CAD correlations among these factors.Now let’s have a look at how these many factors can be used to make predictions.Unlike the 1990s, there appears to be a correspondence between the correlation and prediction distributions..Keep in mind that the order of the factors in these charts is not the same, and so this data does not tell us if factors that correlate with USD_CAD also predict changes in USD_CAD.A quick summary, and on to model generationWe saw that the overall dataset has more stuff from this decade than in past decades, and that data from decades ago is probably not relevant to us today..There are some key points in time where the underlying story in the data changes, and we can’t use data from before that time to make predictions about the period following it.We also saw that the data is rational with respect to our expectations about the world..A model based on this data should be relatively interpretable..For example, if a few factors point to a USD_CAD rise, we should be able to see what those factors are, and validate that this long or short position makes sense.At this point I want to refer back to my article on lots of things that you should not do when setting up a trading strategy..Keep this stuff in mind as we proceed to build a model.. More details

Leave a Reply