Fully automated futures trading

megab said:
Hi there,

First of all, just wanted to say thank you for your book, which I think is excellent, and has helped me immensely.

I am curious, why is that your system seems to outperform AHL´s own "flagship product" or other similar multi asset products from other investment companies?

Am I missing something, besides perhaps they have a lower volatility target?

For example: https://www.ahl.com/programmes/ahl-dimension
More...

AHL dimension has quite a different product profile to my stuff, I prefer to match myself against AHL Diversity which is available in a GBP share class. After accounting for vol differences, and stripping out the non futures trading (my long only stocks, plus a short futures equity hedge which has added a total of around 30% to performance since I started trading), I get these figures for the last three UK tax years:

Me 58.2%, 23.2%, -14.0%
AHL 106.9%, -10.6%, -6.2%

That actually puts AHL slightly ahead (more detail here)

Couple of other points worth making:

1) Performance judged over a period of just a few years is just luck.
2) "Small" ($500K in my case) accounts do have some advantages over large (https://qoppac.blogspot.co.uk/2015/11/david-versus-goliath.html)

GAT
 
globalarbtrader said:
... unfortunately you can't assume a normal distribution, especially if you're trading vol indices. What's the distribution of daily losses seen in your backtest?

GAT
More...

Good question, GAT. Here is my distribution of backtested daily returns over the past 7 years. That one recent brutal VIX day was actually the largest daily loss in 7 years! So, that makes me feel better that it was a major anomaly, and hopefully won't occur again for another 7 years.





 

Attachments

  • Distribution of returns.png
    Distribution of returns.png
    9.2 KB · Views: 67
Napoleon@Dynamite said:
Good question, GAT. Here is my distribution of backtested daily returns over the past 7 years. That one recent brutal VIX day was actually the largest daily loss in 7 years! So, that makes me feel better that it was a major anomaly, and hopefully won't occur again for another 7 years.





More...

Yes, and although it was brutal it doesn't stand out as an unusual outlier.

GAT
 
By the way, GAT, when I run print(system.accounts.portfolio().annual) from your code, I get the attached. Do each of these values assume the same starting capital at the start of the year? If so, you'll notice that 2004, 2005, and 2006 had major losses; if we have a repeat of something like that, that represents a cumulative drawdown of 77%. Is that to be expected when dealing with a 30% volatility target on 8 instruments? Thanks, as always, for your guidance.
 

Attachments

  • Distribution of returns.png
    Distribution of returns.png
    19.6 KB · Views: 47
Napoleon@Dynamite said:
By the way, GAT, when I run print(system.accounts.portfolio().annual) from your code, I get the attached. Do each of these values assume the same starting capital at the start of the year? If so, you'll notice that 2004, 2005, and 2006 had major losses; if we have a repeat of something like that, that represents a cumulative drawdown of 77%. Is that to be expected when dealing with a 30% volatility target on 8 instruments? Thanks, as always, for your guidance.
More...

Yes they assume the same starting capital. More discussion on that here

As a very rough guide you shouldn't be surprised by a drawdown of twice your vol target (60%) [see 'really bad drawdowns here] in a 10 year back test. 77% isn't that much more than, and this is over a longer period.

GAT
 
globalarbtrader said:
As a very rough guide you shouldn't be surprised by a drawdown of twice your vol target (60%) [see 'really bad drawdowns here] in a 10 year back test. 77% isn't that much more than, and this is over a longer period.
More...

Of course, drawdowns anywhere near that scale aren't survivable in the modern asset-management environment (though they may be in one's own account). The long-term Sharpe of trendfollowing just isn't high enough to support 20- 30 vol products that can survive predictable drawdowns -- even setting aside whether trendfollowing is "broken" post-2008.
 
1. Hi Rob, I have spent last 12 months or so reading your first book and catching up on your many detailed posts on trading. You seem to have the knack for making, what might otherwise be a dull topic, thoroughly enjoyable. So, thank you for that and congratulations on the new book, I shall certainly be lining up to order it. I am currently in the process of calibrating the amount and makeup of account equity based on the advice you have so generously provided. I am of course trying to strike a balance between broker credit risk and interest costs (i.e. how much of the account equity should be in cash and how much in ringfenced securities) and I tried a while back to get a ballpark worst case estimate using the example of the crude oil subsystem in your first book (page 159). The assumptions and calcs I have made based on your example are:

a. Target vol of £1M equates to $6million Risk Capital (based on 25% vol and FX of 0.67)

b. The ICV is for Dec15 crude oil is measured in Apr 15 (the date guessed from the FX rate you provide and the contract choice is guessed on your preference to trade Dec oil). I have then found historical margins on the CME website which shows that CL8 (i.e. Dec in April) would have had a margin requirement of $4100.

c. Contracts for a max forecast of 20 = 2 * 93.52 = 187 contracts

d. Finally using a max IDM of 2.5 I get a max margin requirement on max allowable risk on this subsystem = 187 * 4100 * 2.5 = 1.9M ~ 32% of Risk Capital. Furthermore, I assume the broker might include a safety buffer of ~ 20% on exchange margins so realistically I would need around 40%. This is conservative of course, as portfolio margin would make this number somewhat lower(perhaps closer to 30%).

I recently found your post on page 18 in this thread which suggests a 1:1 mapping of cash vol target and initial margin (which makes intuitive sense). Therefore it seems if you are trading at full tilt you would expect margin usage of 50% of capital which is clearly higher than my back of the envelope calcs. It will of course become clear from paper trading what the overall risk should be but I was wondering if my approach in calculating worst case margins is off track somewhere. Could I possibly trouble you to take a moment to point me in the right direction please?
 
Last edited:
Elder said:
1. Hi Rob, I have spent last 12 months or so reading your first book and catching up on your many detailed posts on trading. You seem to have the knack for making, what might otherwise be a dull topic, thoroughly enjoyable. So, thank you for that and congratulations on the new book, I shall certainly be lining up to order it. I am currently in the process of calibrating the amount and makeup of account equity based on the advice you have so generously provided. I am of course trying to strike a balance between broker credit risk and interest costs (i.e. how much of the account equity should be in cash and how much in ringfenced securities) and I tried a while back to get a ballpark worst case estimate using the example of the crude oil subsystem in your first book (page 159). The assumptions and calcs I have made based on your example are:

a. Target vol of £1M equates to $6million Risk Capital (based on 25% vol and FX of 0.67)

b. The ICV is for Dec15 crude oil is measured in Apr 15 (the date guessed from the FX rate you provide and the contract choice is guessed on your preference to trade Dec oil). I have then found historical margins on the CME website which shows that CL8 (i.e. Dec in April) would have had a margin requirement of $4100.

c. Contracts for a max forecast of 20 = 2 * 93.52 = 187 contracts

d. Finally using a max IDM of 2.5 I get a max margin requirement on max allowable risk on this subsystem = 187 * 4100 * 2.5 = 1.9M ~ 32% of Risk Capital. Furthermore, I assume the broker might include a safety buffer of ~ 20% on exchange margins so realistically I would need around 40%. This is conservative of course, as portfolio margin would make this number somewhat lower(perhaps closer to 30%).

I recently found your post on page 18 in this thread which suggests a 1:1 mapping of cash vol target and initial margin (which makes intuitive sense). Therefore it seems if you are trading at full tilt you would expect margin usage of 50% of capital which is clearly higher than my back of the envelope calcs. It will of course become clear from paper trading what the overall risk should be but I was wondering if my approach in calculating worst case margins is off track somewhere. Could I possibly trouble you to take a moment to point me in the right direction please?
More...

I think your maths /logic is mostly right, but you've laid out the problem slightly different than on page 162 (print edition), so to be clear:

portfolio value: £4M (based on 25% vol)
annual cash vol target: £1M
daily cash vol target = £62,500
Price vol: 1.33%
block value: $750
ICV: $997.50
fx: 0.67
IVV: £668.325
VOl scalar: 62500/668.325 = 93.52 contracts
subsystem position with a forecast of 20: 20 x 93.52 / 10 = 187 contracts

With an IDM of 2.5 this would be 187 * 2.5 = 468 contracts
Eithier (a) we're trading this crude oil with a bunch of other things, in which case we'd be multiplying by an instrument weight as well as the IDM; or (b) we're trading it by itself in which case the instrument weight is 100% and the IDM is 1.

For now let's assume it's (b) - a system with only crude oil. So the position is 187 contracts.

Now: Risk (at least in my world) shouldn't be confounded with margin

The daily risk of this position is 187 * 997.50 [IVV: expected daily risk in $ per contract) * 0.67 (fx) = £124,977 [the margin doesn't come into it]. This is almost exactly twice the daily cash vol target - by construction.

What about margin? Well the margin, as you say, is $4100*187 = $766,700 or £513,689. This is around 12.8% of the capital at risk (£4m).

Let's bring the IDM back in again. In other words I'm pretending we have loads of things trading, all exactly like crude with margins around 12.8% of capital with a forecast of 20. Now the peak margin usage will be around 2.5*12.8% = 32%. The average margin usage (forecast of 10) will be 16%. This is around half of what I use myself (around 30%, from page 18). Reason: (and this is the crux of your question) Crude oil is relatively margin friendly. Many other instruments require more margin than this.

Hope this makes more sense now.

GAT
 
You must log in or register to reply here.
Back
Top