Be aware that full-sized and mini/micro contracts do not have 100% correlation. Their correlation is high, but not 100%. Those products are traded separately and their prices deviate slightly from each other. I am not familiar with Fisher transformations. Would it cause computational problems if you have instruments with a very high (but not 100%) correlation?
FYI I'm running the same system as rob (different currency, and a higher notional value) I have gold_micro. Have been in and out of silver the last couple of weeks.
I shrink the correlation matrix for DO (discussed elsewhere), but there are other places where I use it and I'd be worried about the ramifications elsewhere as you say. Also, since I'd be calculating forecasts for two instruments, that would create double counting in the cross market trading rules I use. At the risk of repeating myself, I use a simple decision matrix here. First of all, drop any instruments that don't pass my cost or volume filters. Second of all pick the instrument with the smallest multiplier, even if it's more expensive. I'll then choose forecast weights depending on the cost of the instrument. I have an automated report that does the selection for me. This also helps where an instrument is traded across multiple venues, eg JGB, or for futures that do very similar things (competing EU and US stock sector futures). As a rule, even with DO, I have more markets than I know what to do with. There are also costs to interpretability in having very large numbers of markets. Getting rid of some can only help. Rob
I think this is the key point. It's almost impossible to work out with DO precisely why any single instrument has a particular position, or not. If you aren't comfortable with this approach, then don't do it. What we're looking for is a portfolio that has the same 'character' and risk exposures as the unrounded portfolio. To take an extreme example, if I run DO with $100k and 102 instruments with a trend/carry strategy it has a 0.91 correlation with the unrounded portfolio and the SR is still over 1 versus 1.27 for the unrounded portfolio (figures from chapter 25 of my forthcoming book). But it holds just 7 instruments on average and never more than 15! It's quite possible that there are over 90 instruments without positions on a given day, and some of those may have very high forecasts, but do not have a position. Rob
That's something new from IB - got my order filled at avgFillPrice "0", messed up my PnL reports obviously (it was a short position so looked pretty bad ). Upon further inspection, all logs that the system recorded for this order have avgFillPrice:0., even "Fill Px" column in TWS for this order displays "?" (the other 3 zeroes are fine, it's from cancelled orders). Although, one of the log records has LastFillPrice populated with seemingly correct price. I'm thinking to add logic to maybe use LastFillPrice as a fallback when avgFillPrice is 0.. The order type was "Snap MID", but it worked fine for me before on other products..
Hard last few days, virtually every move in the market went against my positions. But I guess I have only myself to blame, as just prior to the move I was guilty of having the idea to post about the great September I just thad. I am curious how everyone else has fared, including the dynamic optimisation crowd, during the recent brutal reversal in risk sentiment in the markets.
Rather similar experience. Last month went great, and I could withdraw a substantial amount of cash near the end of the month (moved it to my lower-volatility investment account). This week however is the opposite. The account value is dropping fast, and the volatility is high. However, as I had ensured that the account value was near the upper threshold, did the system respond immediately when things started to go south. Had I not made the withdrawal then the drop in account value would have been much larger as the system would have started to respond with a much longer delay. By the way: I am not using the dynamic optimization, just the basic system as described in Rob's book "Systematic Trading".
Same. Hit a HWM on the 28th September, withdrew my excess capital, then watched it crater. Current d/d is 7.3%, basically I'm back to where I was around the 20th September. Rob
Note that some of my volatility was due to the pound dropping, and then rising, resulting in a revaluation of my USD and EUR cash in my account. Rob