Monday 29 March 2010

Trading plans - the next work

As the New York trip drew to a close, I actually began to feel quite excited about getting back and seeing the results of the blind test set up while we were away. One of the great things about using very short term data is that you quickly accumulate a whole stack of trades. Afterall, with 10 minute bars, there are around 75 bars in a full days trading, equivalent to 3 months of daily data. Three or four weeks is the equivalent of 5 or 6 years of daily data. So you start to get some statistically significant information quite quickly - relying on the principle of self-similarity in market data.

Some of my thinking over the past week has been aimed at various trade-by-trade statistical measures that I need to keep a record of. Things like "how bad does a profitable trade get", "how good does a bad trade get", etc, which generate in turn their own statistical distributions and interesting conclusions. But it is such a pain to have to derive these manually - it takes about 6 hours to record a month's trading stats with all the associated measures - no doubt I could convert all the data into an excel format and work it out from that, but that would also take quite some time - but might be worth looking into, especially if I end up with some "family" type systems.

Many recent thoughts of my own time as a hedge fund manager (finished in 2006). On the one hand, reading books on statistical arbitrage and other quant strategies has indicated just how large the task was for us and how the odds really were stacked against us. But on the other hand, we did once have close to $20m under management and it seems strange to think we got to that point. But also, I am frustrated by thinking about some of the trading ideas we had but never found time to implement. Many of those did involve intra-day and very short-term holding periods and that seems to have been the right way to have gone - at least if one thinks about the success of Renaissance or SAC.

So I will have finished the manual extraction of trading performance data in the next few days and will then start the detailed statistical analysis of this data with a view to making a preliminary system selection in a few week's time. Odd to feel so excited about this again, after so long out of the markets.

No comments: