Thursday’s close finally (although prematurely by at least one day;)) signalled a change from the short position on the SPY. This position was established on December 10th. It’s interesting that the preceding long position lasted only for a single day. In other words, my SPY strategy has been short on the SPY for about a month now.
This short was worth about 1%, with the 1.5 leverage, but was quite painful – it was down as much as 3%! That hurts regardless.:)
A reader’s comment on my ARMA Models for Trading post asked about different aspects of my experience with ARMA+GARCH for trading forecasting. The more I thought about it, the more it looked like a full post. So here we go.
The short position has been in place for a while, and I went through some painful moments with it.:) As of today’s close, I reversed it to a long with a leverage of 1.5.
To get a feeling of SVM performance in trading, I run different setups on the S&P 500 historical data from … the 50s. The main motif behind using this decade was to decide what parameters to vary and what to keep steady prior to running the most important tests. Treat it as an “in-sample” test to avoid (further;)) over-fitting. First the performance chart:
The long position lasted only a single day! Not typical for this strategy, curious to see what it has in mind. The leverage is 1.5.
My system finally signalled a switch from the short position on the S&P 500 (on the SPY ETF more precisely) held, in theory, since Oct 23rd, but in practice since Oct 26th (more details here). In practice, the position ended up as a loss of 1.4% accounting for leverage. The second loss in a row.
Switched from short to long at the close today, using a position size of about 1.48.
Finally all the stars have aligned and I can confidently devote some time for back-testing of new trading systems, and Support Vector Machines (SVM) are the new “toy” which is going to keep me busy for a while.
SVMs are a well-known tool from the area of supervised Machine Learning, and they are used both for classification and regression. For more details refer to the literature.
It seems to me that the most intuitive application for trading is regression, so let’s start by building an SVM regression model.
Entered this trade a day later than the arrival of the signal (the signal was in-place at the close on Friday). Using leverage of 1.48.
A week ago I read Kyle Bass’s latest newsletter (for those of you who don’t know who he is: Kyle Bass was betting against the US mortgage market back when Ben Bernanke was testifying in front of the congress that the mortgage troubles are not going to have any significant impact on the US economy) and it resonated strongly with my personal ideas and expectations, especially on the state of affairs in Japan.