Fully automated futures trading

[Kernfusion]

Thank you for sharing this! I appreciate it as it gives me the possibility to compare with what I'm using.
If you are using four metals this means 4% per instrument. Energies two instruments, so 4% per instrument. STIR 4%. If I remember correctly are you using a larger number of instruments in the other classes, so the percentage per instrument will be lower.

I used the handicraft method which you describe in your book. However, I saw in the appendix that the correlation between STIR and bonds is 0.5. And the correlation between equities and Vol 0.6. I thought that those correlations were not extremely high so I did not cluster these into "superclasses". Which made me end up with equal weight for each of the classes you mention. Now, after running the system for two years and having gained some more insight in these instruments, I start to see that it does make sense to create such superclasses.
Later in time, when IB raised margin requirements on V2TX, I decided to reduce its weight (I don't use VIX as the contract size and associated value volatility is too large for my account). I use 5% for it currently. So that was not driven by portfolio optimization, but by margin consumption. High usage of margin by V2TX prevented other positions to be opened or expanded as IB started blocking order lines. Thus causing a deviation between the desired portfolio and the actual portfolio. By trial and error I found that this deviation became negligible when using 5% weight for V2TX.

I used Pysystemtrade for weights and everything else, for 1 instrument in each class it gave me these weights:

kinda makes sense I guess, nothing too weird..

Still paper-trading it btw, need to save some more money before I can blow it )
Planning to start in the beginning of the new year on 120k base capital and 22% target volatility (lower than that and some instruments are struggling to get 2 contracts max positions even on the maximum forecast with thresholting..), in time will add some more capital and instruments ..
Also, automatic rolls based on the pre-recorded calendar for 90 years ahead (don't plan on living longer ) seem to work well (every instrument has a soft and hard roll periods based on historical volume and exp. dates + during the soft-roll in real time the system is checking that the new contract has at least 1% volume of the previous one..)
., it's because of another dug the system sold ~25 contract of platinum around Aug 16 and lost 9k - so it's a good idea to try that "software miracle" in paper first )

Is it sort of true btw that for every additional instrument you roughly need +10k capital?

 

[HobbyTrading]

I used Pysystemtrade for weights and everything else, for 1 instrument in each class it gave me these weights:

kinda makes sense I guess, nothing too weird..

Can I make a suggestion for alternative instruments? In each asset class you would like to use the instrument with the lowest value volatility. Lower value volatility enables you to hold more contracts at the same account risk setting (how did you arrive at a risk setting of 22%?)
Equity: DJ600 correlates very strongly with ESTX50 but the contract's value volatility is about half.
Forex: M6E (Euro mini) has a value volatility which is about three times smaller than MXP.

In the discussion in the last few days was it remarked by @globalarbtrader that GE correlates strongly with bonds (such as ZT). The combined weight given to GE and ZT (24% of total) could therefore be a bit on the high side.

 

[JMW]

I don't know what anyone does currently but I can tell you what has been done in the past and / or researched.

Firstly it's not obvious what you mean. Do you mean (a) trend following, where a short option position is the underlying market? or (b) using options to do trend following rather than buying / selling delta 1 instruments like futures?

For (a) well I guess I trend follow the VIX so it sort of makes sense.

For b) It makes some sense to do trend following by buying calls or puts; then if the trend reverses you don't need to cut your position, you just watch your premium leak away, and shrug your shoulders. In many ways trend following has the same payoff profile as a long strangle, but it tends to be cheaper since you're effectively buying realised vol rather than paying for implied vol. Doing trend following by selling options then makes no sense at all.

GAT

Hi - do you have any recommendations for reading on the topic of trading options?

 

[globalarbtrader]

Hi - do you have any recommendations for reading on the topic of trading options?

Taleb dynamic hedging

GAT

 

[Kernfusion]

Can I make a suggestion for alternative instruments? In each asset class you would like to use the instrument with the lowest value volatility. Lower value volatility enables you to hold more contracts at the same account risk setting (how did you arrive at a risk setting of 22%?)
Equity: DJ600 correlates very strongly with ESTX50 but the contract's value volatility is about half.
Forex: M6E (Euro mini) has a value volatility which is about three times smaller than MXP.

In the discussion in the last few days was it remarked by @globalarbtrader that GE correlates strongly with bonds (such as ZT). The combined weight given to GE and ZT (24% of total) could therefore be a bit on the high side.

DJ600 and ESTX50 do look much smaller at lest in terms of margin requirements., but they aren't in the Rob's instrument list in Pysystemtrade, I wander why? are they less liquid or after a certain account size you actually start preferring to trade larger contracts to save on commissions?
Although, MXP isn't a problem, with this VolTgt and capital the system can hold even more that the required 4 contracts, the problematic instruments are CL, ESTX50 and PL(platinum)..

Regarding 22%, I figured Sharpe of trend following is 0.5, therefore half-Kelly gives you 25% vol target., but with only 8 instruments it will be even lower, so I wanted to use 20% and 100k but then I couldn't hold enough contracts in some instruments, so I bumped up the Vol and capital slightly. Not too scientific I guess..

Regarding instrument weights, I guess it does make sense to take 6% from ZT+GE and redistribute into MXP+CL+ZC.. But on the other hand, the optimiser gave these weights, so who am I to contradict Also, they don't seem to be jumping around much over time, although US2 has shorter history..

 

[HobbyTrading]

Although, MXP isn't a problem, with this VolTgt and capital the system can hold even more that the required 4 contracts, the problematic instruments are CL, ESTX50 and PL(platinum)..

You mentioned that your account value will start at approximately 100 k USD. I agree with your view that CL, ESTX50 and PL are then limited to only 2~3 contracts maximum. Rob's account is larger in size (he reported > 400 k GBP in this log), but he diversifies over more instruments.

Thanks for explaining how you arrived at 22% for your target volatility setting. I didn't do a "small account value correction" and left it at 25%.

In your graph I don't see the red line for ESTX50. Does it overlap with one of the other lines? DJ600 was only introduced in 2010, so you would not be able to do a long backtest with this instrument. However, as it correlates strongly with ESTX50 could you probably use ESTX50 as proxy in the backtest. This short history could also explain why Rob did not include this instrument in his list. I found it during a recent search, looking for equity index futures which have smaller value volatility than the "famous ones" such as ESTX50, ES, NQ. I ended up with DJ600 for Europe, plus XINA50 and NIFTY (and SGXNK) at the Singapore Exchange. This gives exposure to Asian equity indexes. The market data fee for the Singapore exchange is relatively cheap.

 

[sle]

Taleb dynamic hedging

GAT

Funny, I am not super-keen on that book, it's very much oriented towards old school market making and it pretty dated. My initial reaction is almost always something like Volatility Trading by filthy, which is not perfect either.

 

[Kernfusion]

You mentioned that your account value will start at approximately 100 k USD. I agree with your view that CL, ESTX50 and PL are then limited to only 2~3 contracts maximum. Rob's account is larger in size (he reported > 400 k GBP in this log), but he diversifies over more instruments.

Thanks for explaining how you arrived at 22% for your target volatility setting. I didn't do a "small account value correction" and left it at 25%.

In your graph I don't see the red line for ESTX50. Does it overlap with one of the other lines? DJ600 was only introduced in 2010, so you would not be able to do a long backtest with this instrument. However, as it correlates strongly with ESTX50 could you probably use ESTX50 as proxy in the backtest. This short history could also explain why Rob did not include this instrument in his list. I found it during a recent search, looking for equity index futures which have smaller value volatility than the "famous ones" such as ESTX50, ES, NQ. I ended up with DJ600 for Europe, plus XINA50 and NIFTY (and SGXNK) at the Singapore Exchange. This gives exposure to Asian equity indexes. The market data fee for the Singapore exchange is relatively cheap.

interesting, I should check these new instruments out..
Yes, actually the lowest curve on the graph is both V2TX and EUROSTOX (they very slightly diverge at the tip if you use a magnifying glass ), I ran the fresh backrest before producing this graph and the weights are always slightly different as it's doing bootstrapping, so it's not deterministic..

 

[HobbyTrading]

Yes, actually the lowest curve on the graph is both V2TX and EUROSTOX (they very slightly diverge at the tip if you use a magnifying glass ),

Interesting. So even though V2TX is technically the time-derivative of ESTX50 this results in a perfectly overlapping graph? I would not have expected that intuitively. However, I have no experience in running those backtests. I only used the "handicraft method" as explained by Rob in his book.
Here are two charts, showing the forecast and value volatility of V2TX (top) and ESTX50 (bottom) for the last two years. Eyeballing the blue forecast lines they don't seem to have a high correlation to me.

 

[globalarbtrader]

Funny, I am not super-keen on that book, it's very much oriented towards old school market making and it pretty dated. My initial reaction is almost always something like Volatility Trading by filthy, which is not perfect either.

Yeah, but I started life as an old school options market maker, and I'm pretty dated as well

There isn't really a equivalent to one of the many books on trading systems for options, maybe I should write one

Failing that I always direct people to my friends home page http://helderpalaro.com/#volarb

GAT