You've created a trading system with the characteristics you want, given some amount of capital. You shouldn't change the trading system just because you have a different amount of capital. What if the reverse were true? What if you were reducing your capital, but you had increasing forecasts in one part of your system? Would you act to increase your positions locally there; flying in the face of Kelly? I think the answer is no. Of course there is a minor issue here - you might incur extra trading costs as a result of this conflicting behaviour, but the use of buffering will reduce these - and unless your vol target is insane trading due to capital changes will be dwarfed by vol and forecast changes. GAT
GAT, Apologies if this question has been asked and answered already. I'm going through another iteration on my system -- this time incorporating costs into the calculation of trading rule weights. My understanding is that you suggest that trading rule weights should be calculated based on pre-cost data of all instruments in an asset class; this is to increase the amount of data used for the calibration and thus improve the statistical significance. But, the different instruments in the asset class could have different trading costs (volatility standardized costs). Your book mentions the speed limit in terms of SR (0.13), which should lead to different weights for each instrument once the costs are taken into account (a weight of zero for any rule that exceeds the speed limit for the instrument). I'm trying to reconcile the two thoughts -- one seems to suggest that all instruments in an asset should have the same rule weights, and the other thought suggests that each instrument should potentially have different rule weights based on costs. Would you use the bootstrapped weights for the entire asset class and just exclude (for an instrument) any rules that exceed the speed limit for that instrument? On another note, I've been experimenting with different trading rules in the Volatility asset class (I split these instruments out of the Equities class into their own class). Anyhow, it seems that calculating the spread between implied volatility and historical volatility is a very good indicator for trading VIX & VSTOXX. The other one that I've been trying is the carry curvature -- taking the difference between the carry at the front of the term structure and the carry further (5, 6, or 7 months) back. When I performed a bootstrapped optimization of rule weights with the carry curvature rule, carry based on front month and VIX index, and carry based on front and back month futures, pretty much all the weight is getting assigned to the carry curvature rule and not either of the other carry rules. Interesting. Also, it seems that trading costs suck up almost the entire SR of a cross sectional approach to VIX and VSTOXX. Anyhow, I wanted to run the indicators (carry curvature and IV-HV) and the cross sectional observation by you. Thanks, --Maciej
Yes, you will sometimes get different sets of rules for different instruments. Using the same weights but then doing exclusion makes things difficult with the code, and also leads to potential problems (eg if all your TF rules but one get dropped you'd end up being underweight TF and overweight carry). Its probably worth (re)reading this post: https://qoppac.blogspot.co.uk/2016/05/optimising-weights-with-costs.html Some random thoughts: Where do you get your implied / historic vol data from? (Presumably it is the implied/ historic of the underlying index). I assume you're using the carry difference 6 months / front but still trading the 2nd contract? It would be interesting to see if this rule works on other asset classes. And finally, yes, fast cross sectional rules often cost too much. Anyway, these seem like sensible ideas. But personally I wouldn't use them exclusively. GAT
Breaking the brand Have CTAs outgrown their label? Chris Hawes 5 September, 2017 The quant futures sector has come a long way since traders first climbed out of the Chicago pits and converted their knowledge into rudimentary systematic models... Man AHL is a systematic trading “pioneer” driven by “science-based research”.... “I think that ‘CTA’ is probably the most misleading label that is out there,” says AHL’s former head of fixed income Robert Carver, now an independent systematic trader, writer and research consultant. “Formally, the acronym refers to a specific type of US CFTC-regulated fund, but the term is now used to describe a very broad spectrum of products, many of which aren’t even offered through a CTA.”...
The funny thing is I couldn't read this article as I'm not subscribed. I asked Chris if I used my quote, and if so to send me a copy, but he didn't reply so I assumed he didn't! GAT (AKA Robert Carver, to those who didn't realise)
I appreciate this is off topic, but I learnt pretty much all I know from this thread, so thought I'd ask here. I trying to come up with a systematic argument to answer this question. I own a London flat * Market value £480k * £250k mortgage. * The net income is 7.2k after taxes, costs and interest. * I have capital gains relief on the flat until 2022. I also have £370k in a trend following futures account, running at 25% vol with after tax sharpe of 0.65. If I sell the flat, I will allocate the cash (230k) to trend following. Arguments to sell flat: * I hate dealing with problems. Both my neighbours hate me because of leaks which I cannot fix. * Law of active management doesn't apply * Diversification doesn't apply * Unknown correlation to trend following (I would guess highly correlated with ftse) * Illiquid * Positive skew means potential hidden big loss in future Arguments to keep flat: * Positive skew investment, makes money most years * Diversified from trend following, depending on unknown correlation * If I screw up trend following, at least I have a backup * No tax drag, cgt is deferred until flat is sold rather than paid out of capital annually * Fundamentals that made this investment work (low interest rates) have not changed. I just have no idea how to systematically evaluate this. How does one factor in property into an investment portfolio?
I'll give you my opinion but I should warn you I've always hated the idea of rental property as an investment so I'm biased. Given that your net yield on your housing equity is just 3.1% I think I'm justified in this opinion here. In fact the most important issue here for me is whether this will increase or decrease your diversification. You've put down two alternatives, when in fact you should be considering your entire portfolio holistically. In fact a fairer comparison would be between the flat and a property ETF that would give you similar asset class exposure, is more diversified, less risk and a similar yield. Important questions: Do you have any other investments? If not 100% invested in a TF strategy will mean less diversification than you currently have. A better comparision would be owning the flat, or owning an real estate ETF that gave you exposure to the same asset class with minimal hassle. Do you own significant other property as well (your own home)? If so you're massively exposed to property and you should definitely lose the flat. Do you have a job or do you live off your wealth? If not then you need investments that will provide an income, and no trading strategy, including TF can do that. Personally... I'd sell the flat. I'd then use the proceeds in such a way to give me an asset allocation that looks something like this: Trend following: I'd suggest no more than 20% of your wealth should be in this if you're not working; if you're working you could probably go a bit higher, but more than a third would make me uncomfortable. Property: I'd suggest a 10% to 30% allocation is reasonable here; again no more than 50%. But that should factor in your other property (eg your own house or flat if you have one). If necessary buy ETFs to make up the difference. Other investments (bonds and equities): the rest GAT
The greatest speed up came from switching from storing price data in CSV to using HDF5 in pytables mode. I calculate the position for a notional capital amount (500k) for each instrument separately and then combine them, and finally rescale the entire portfolio based on a moving average of the whole portfolio return volatility, so that it hits the target. Each instrument is calculated on a separate cpu core using processpoolexecutor.map(). By vectorise everything, I mean use vectors for calculation, so that you would do something like return = prices.diff() * position (in pandas syntax here), as opposed to looping over every row. This works well for this kind of system because you assume have the same capital every day.