Roughly 25% carry the rest tf, with heavier weighting on slow ewmac. It’s static through time. About 25 instruments here, although only one equity index (Nasdaq)
Not necessarily. There are those in the market who think if inflation reignites then the dollar will tank.....in which case you might worry about the opposite scenario where both move against you. I guess you can’t away from the premise that what you are doing relies on the extent of your belief in a fully automated system.
I think my main issue is that it looks diversified when in fact it isn’t. I think the whole thing is driven by these factors: * real gdp growth (indexes) * interest rates (bonds, indexes) * inflation (commodities, inflation linked bonds)
My diversification concern is the following: Stocks and bonds are negatively correlated. Great, so that means diversification. But, if stocks are trending up, then my system will be long stocks. So, that means bonds will probably be trending down, so I will be short bonds. Aren't I losing a lot of the diversification benefit by doing this? If stocks suddenly crash, that means bonds will probably rally and I'll get a double whammy loss on both instruments. What do you guys think?
Are there no non-economic factors that have an influence? For example government policies of one region/country versus other regions/countries? I don't know the answer, but see that you only mention purely economic factors.
I think this can be answered by looking a bit more in detail about what correlation actually means. If instrument A and B have a correlation of +1 (i.e. +100%) it means that if A goes up, B goes up in 100% of the cases. If A and B have a correlation of 0.5 (i.e. 50%) it means that if A goes up, B goes up in 50% of the cases. In the other 50% of cases does it not follow A. Equities and bonds do have a negative correlation, but it is not -100%.
Stocks and bonds are not negatively correlated. Correlation is closer to zero. It is negatively correlated during times of extreme stress.
Of some relevance to discussions about performance expectations, todays blog post https://qoppac.blogspot.co.uk/2018/02/cta-allocations-qe-meta-prediction-and.html GAT
A nice post, thanks for sharing it. One small thing: "these buckets are divided at the median value of the conditioning variable" -- presumably you didn't know ex-ante the distribution of the conditioning variable, so dividing at the median is a subtle hindsight bias.