Fully automated futures trading

[globalarbtrader]

What are your toughts on combining trend following strategies (and/or vol scaling risk parity strategies) with volatility trading funds?
Vol trading can mean anything but I'm referring to strategies which most of the time have a long vega exposure, which is reduced when implied volatility gets really high.

For European investors, I'm talking about something like Amundi Volatility World or Seeyond volatility strategy.

This is the excess return (vs cash) of the two funds (since they have different inception dates, I had to patch different time series).


On a stand-alone basis, they look horrible: in the long run you can assume zero to slightly negative excess returns. What I like is that they are negatively correlated with anything else. In particular, they should in theory complement nicely with trend following / volatility targeting risk parity strategies: these funds are going to shine when we have a burst of volatility after a long period of quiet (and probably trending) markets, which is just the kind of environment where trend and vol scaling suffer.

This chart represents the rolling 20-day ER of CTA (x asis) vs Amundi/Seeyond blend (y axis).
Althogh we have few observations (very few, considering observations overlap) the data seem to confirm my hypothesis


Clearly there are other considerations to be made:
- cash efficiency: in order to have a meaningful impact, vol trading should have a 5-10% allocation of the overall risk budget. Since vol trading funds usually target a volatility around 8%, if you run trend/RP strategies at reasonable volatility targets and you use mainly futures, you should have plenty of cash available to fund the new exposure. Otherwise, that could be an issue.
- fees: with IB those funds are available at 0.8%-1% fee.

What do you guys think?

I wonder how easy it would be to replicate those sorts of payoffs just by being long V2X/VIX with some TF/carry/mean reversion rules so you only go long when it's likely to be worth your while.

GAT

 

[younggotti]

I wonder how easy it would be to replicate those sorts of payoffs just by being long V2X/VIX with some TF/carry/mean reversion rules so you only go long when it's likely to be worth your while.

GAT

That's a nice question, but IMHO you'd get a very different risk profile.

Trend is probably not going to help in that kind of environments because after a long lasting quiet market you will get negative signals. Unless you use very short term lookbacks, it will take some time to adapt to new conditions and, by the time it does that, CTAs / RP will probably have already degeared.
Carry is not going to help too, as VIX curve is usually very steep when markets are quiet, so you are probably going to have negative signals.
Maybe value/mean reversion can be the only factor with positive signals in that kind of environment.

I'm not saying that those kind of signals are useless: I have a strategy which also trades vol futures on exactly those signals. But it's something completely different than those funds are doing: it's not a (although imperfect) tail hedge, but a strategy aiming at harvesting the volatility risk premium by being short volatility most of the time.

 

[newbunch]

Rob,
Do you trade the full-size currency contracts or the micros? The micros give me more granularity but the commissioners are proportionally higher (at IB, 0.33 for the micro vs. 2.47 for the full size with the micros being one-tenth the contract size). I think the spreads are comparable--either exactly the same or slightly favoring the full-size contract whenever I check.

 

[newbunch]

Rob,
Do you trade the full-size currency contracts or the micros? The micros give me more granularity but the commissioners are proportionally higher (at IB, 0.33 for the micro vs. 2.47 for the full size with the micros being one-tenth the contract size). I think the spreads are comparable--either exactly the same or slightly favoring the full-size contract whenever I check.

The cost difference between the micro and full-size currencies is not that great, but it's much greater between the mini and full-sized grain contracts. IB commissions are 2.82 for the full-size contracts vs. 1.90 for the minis (9.00 for the equivalent full size of 5 contracts).
How to think about measuring the benefit of being able to take a position (and granularity of the position) against the commission cost?

 

[globalarbtrader]

The cost difference between the micro and full-size currencies is not that great, but it's much greater between the mini and full-sized grain contracts. IB commissions are 2.82 for the full-size contracts vs. 1.90 for the minis (9.00 for the equivalent full size of 5 contracts).
How to think about measuring the benefit of being able to take a position (and granularity of the position) against the commission cost?

I run a regular report to check if I should be trading full size or micros.

rob@TradingPC2:~/pysystemtrade/sysproduction/linux/scripts$ . interactive_controls

0: Trade limits
1: Position limits
2: Trade control (override)
3: Broker client IDS
4: Process control and monitoring
5: Update configuration


Your choice? <RETURN for EXIT> 5
50: Auto update spread cost configuration based on sampling and trades
51: Suggest 'bad' markets (illiquid or costly)
52: Suggest which duplicate market to use


Your choice? <RETURN for Back> 52

However this only works if I'm collecting data for both. As it happens, for FX I haven't looked at the micros since the FX contracts aren't really that big in risk terms.

Just quickly checking the GBPUSD micro I can already see that this wouldn't meet my liquidity requirements

The rule I use is, first exclude anything that doesn't meet my cost and liquidity requirements:

Maximum SR cost? <RETURN for default 0.01>
Minimum contracts traded per day? <RETURN for default 100>
Min risk $m traded per day? <RETURN for default 1.5>

Then I basically choose the instrument with the lowest contract size (so micros, then minis, then full size).

Here are a couple of examples

Current list of included markets ['ETHEREUM'], excluded markets ['ETHER-micro']
              SR_cost  volume_contracts  volume_risk  contract_size
ETHEREUM     0.002750             176.0        16.71        95076.0
ETHER-micro  0.085866             661.0         0.13          201.0
Best market ETHEREUM, current included market(s) ['ETHEREUM']

The micros don't meet cost requirements or volume requirements, so we go full fat.

Current list of included markets ['SP500_micro'], excluded markets ['SP500']
              SR_cost  volume_contracts  volume_risk  contract_size
SP500        0.000546         1035654.0     24946.50        24088.0
SP500_micro  0.000575          728398.0      1755.54         2410.0
Best market SP500_micro, current included market(s) ['SP500_micro']

Both markets meet the requirements but the micro obviously is preferred. As it happens, in this case the costs are almost identical.
(note SP500 here is actually the e-mini - the full fat contract is delisted)

Now of course there is a more sophisticated trade off to be had between costs and contract size. The sort of heuristics I use in this post would be useful. But this is very complex, since unless you're trading only a single instrument you really ought to make the decision jointly across all instrument selection decisions.

However now I'm using the dynamic optimisation, this decision has become less critical, so I'm less interested in getting that decision right - I'm happy doing it in this relatively crude way.

Rob

 

[younggotti]

Just quickly checking the GBPUSD micro I can already see that this wouldn't meet my liquidity requirements

Which requirement is not met by GBPUSD micro?
I don't have extensive price data, but right now the spread on the micro (March) is the same as the full contract.

 

[globalarbtrader]

Which requirement is not met by GBPUSD micro?
I don't have extensive price data, but right now the spread on the micro (March) is the same as the full contract.

No costs are fine (around 0.2 SR units, actually slightly cheaper than the full size contract)

I don't think it meets liquidity requirements, although it's pretty close (average daily volume in risk terms is about $1.4 million per day).


Rob

 

[younggotti]

No costs are fine (around 0.2 SR units, actually slightly cheaper than the full size contract)

I don't think it meets liquidity requirements, although it's pretty close (average daily volume in risk terms is about $1.4 million per day).

If spreads are tight, why do you think volumes are important? Are you worried that, in martet turmoils, liquidity could dry up?

 

[globalarbtrader]

If spreads are tight, why do you think volumes are important? Are you worried that, in martet turmoils, liquidity could dry up?

Yes, but even in normal times spreads only show you what you are going to pay to trade if you aren't an excessively large part of the market. Consider an extreme situation where you spreads are tight as a gnats wotsit, but the volume is very low such that your trading would consitute 50% of the daily volume. Once you started to put that kind of order flow through the volume at the top of the order book would evaporate and the effective spread you are paying would widen to the point you could drive a whole bus full of gnats through it.

Rob

 

[younggotti]

Yes, but even in normal times spreads only show you what you are going to pay to trade if you aren't an excessively large part of the market. Consider an extreme situation where you spreads are tight as a gnats wotsit, but the volume is very low such that your trading would consitute 50% of the daily volume. Once you started to put that kind of order flow through the volume at the top of the order book would evaporate and the effective spread you are paying would widen to the point you could drive a whole bus full of gnats through it.

Rob

(Un)fortunately my portfolio is small enough that I will never be a large part of any market