Thoughts on pair trading system?

[Punta]

I've been trying to develop a system that is scalable, and I'm interested getting some opinions on whether my latest effort fits the bill.

It trades the HSI/SPI futures intraday (always flat at the end of the day). It's systematic and automated, so I just log in at the end of the day and make sure all looks well.

It's been going for 6-weeks now, and below is a summary of the first 50-odd trades.

The key point is that

1) It's profitable; it's turned roughly 30% profit in 6-weeks (2.5k profit, costs about 8k to hold a pair of contracts). This agrees well with backtesting over 2-years of data.

2) Expected value is pretty small - about 8-ticks on the HSI, and negligible on the SPI.

3) Clearly prone to draw downs. I am experimenting with stops to try and minimize these.

Do people have a feel for the limit to the number of contracts that a system like this can take before it is overwhelmed by slippage? I've been trading 2 pairs (numbers in figures are for one pair).

What fraction of capital would you be happy to allocate to a system like this, with good returns but pretty scary drawdowns?

 

[skc]

I've been trying to develop a system that is scalable, and I'm interested getting some opinions on whether my latest effort fits the bill.

It trades the HSI/SPI futures intraday (always flat at the end of the day). It's systematic and automated, so I just log in at the end of the day and make sure all looks well.

It's been going for 6-weeks now, and below is a summary of the first 50-odd trades.

The key point is that

1) It's profitable; it's turned roughly 30% profit in 6-weeks (2.5k profit, costs about 8k to hold a pair of contracts). This agrees well with backtesting over 2-years of data.

2) Expected value is pretty small - about 8-ticks on the HSI, and negligible on the SPI.

3) Clearly prone to draw downs. I am experimenting with stops to try and minimize these.

Do people have a feel for the limit to the number of contracts that a system like this can take before it is overwhelmed by slippage? I've been trading 2 pairs (numbers in figures are for one pair).

What fraction of capital would you be happy to allocate to a system like this, with good returns but pretty scary drawdowns?

Don't think return % compared to what it takes to hold the pair. Think what risk capital you need to run this strategy. As it's intraday and your max loss appeared to be $800 (so far), you can say perhaps you are risking 2% per trade at $800. This will put your risk capital at $800 / 2% = $40000. So your return is probably closer to 6.25%. Or 1% per week which is still very good.

All systems will experience drawdown. Your drawdown of $1k is big on your $8k margin, but not really that scary at all on your $40k theoretical risk capital.

You probably need more data running forward to see how bad your drawdown would be. With that information you can set your capital allocation with the combination of max loss as percent of capital + maximum drawdown expectation.

Re: slippage. What instruments are you using? With 8pts avg on the HSI there isn't that much room for error.

 

[CanOz]

Interesting idea punta

Got any stats?

 

[Punta]

Don't think return % compared to what it takes to hold the pair. Think what risk capital you need to run this strategy. As it's intraday and your max loss appeared to be $800 (so far), you can say perhaps you are risking 2% per trade at $800. This will put your risk capital at $800 / 2% = $40000. So your return is probably closer to 6.25%. Or 1% per week which is still very good.

Yeah I've been struggling to work out what the risk is with this system. There are no stops at the moment, so I don't really know what I'm risking. In backtesting, biggest loss for single trade is about $3k. Average return per trade is pretty well $100, after fees, giving average risk/reward of 3.

In forward testing, the average return after fees is $50; less than the backtesting mainly due to slippage, and giving R/R of more like 6.

Are you saying that as a rule of thumb, you want your Risk to be only 2% of capital allocated to the system? I'd have to think about whether that would make sense for me - it seems like there'd be a lot of capital sat around doing "nothing".

This would mean that I should allocate more like $150k to the system, which a) I don't have, and b) makes the annualised returns start to look rather pedestrian.

You probably need more data running forward to see how bad your drawdown would be. With that information you can set your capital allocation with the combination of max loss as percent of capital + maximum drawdown expectation.

Re: slippage. What instruments are you using? With 8pts avg on the HSI there isn't that much room for error.

Definitely need more data running forward.

It trades the front month for the SPI200 and HSI index futures. Same number of contracts each - I haven't thought too much about position sizing yet.

 

[Punta]

Interesting idea punta

Got any stats?

I don't really have these to hand. A quick look at the forward tested trades tells me that

no wins/no losses = 0.71
average ticks won = 0.4 on SPI; 8.7 on HSI
max loss = -760 AUD (not inc. fees)

Obviously very early days to tell whether this could run long term, but the encouraging thing is that it agrees well with backtested data.

 

[Punta]

I think that the answer to my question about scalability is

1) Not really possible to scale this vertically very far, due to slippage (maybe horizontally, looking for other pairs).

2) The stats don't justify vertical scaling anyway, without a f-load of capital.

I kind of had it in my head that as soon as I cracked a successful equity index futures strategy, it would be automatically scalable. I guess I'm realising this isn't the case ...

 

[skc]

Yeah I've been struggling to work out what the risk is with this system. There are no stops at the moment, so I don't really know what I'm risking. In backtesting, biggest loss for single trade is about $3k. Average return per trade is pretty well $100, after fees, giving average risk/reward of 3.

In forward testing, the average return after fees is $50; less than the backtesting mainly due to slippage, and giving R/R of more like 6.

Are you saying that as a rule of thumb, you want your Risk to be only 2% of capital allocated to the system? I'd have to think about whether that would make sense for me - it seems like there'd be a lot of capital sat around doing "nothing".

This would mean that I should allocate more like $150k to the system, which a) I don't have, and b) makes the annualised returns start to look rather pedestrian.

The number 1 rule in any form of trading is to reduce your risk of ruin to practically zero. If you want to focus on annualised return on margin required to support the contracts, you are simply looking for trouble imo. I can trade ASX200 index CFD on 0.5% margin. If I manage to catch a 2% move, did I just earned 200% return? Or am I just a leveraged nut? If you seriously trade this system with just $8k, 2 bad trades at close to maximum backtested loss would see you give up on the system pretty quickly imo. So your capital isn't "sitting around doing nothing" as you put it. It is your "risk" capital. It doesn't need to sit in your trading account, but it needs to be there to handle the inevitable drawdown both financially and emotionally.

The 2% is just a simple adaptation of the 2% rule. Where you lose no more than 2% of your capital in any one trade. I applied the rule backwards based on your worst losing trade to work out a capital base. Feel free to adapt a number different to 2%. Alternately/in addition, you can use a hard $ stop on your pairs.

I still feel there's something not quite right with taking just 8pts out of HSI which is like 20000 pts contract. But without knowing much more about your system I am just guessing. Someone more knowledgeable about the HSI can probably make better comments.

 

[Punta]

It is your "risk" capital. It doesn't need to sit in your trading account, but it needs to be there to handle the inevitable drawdown both financially and emotionally.

Thanks for the thoughts skc - I need to think about this. I have about 40k allocated to this strategy at the moment, but that's trading 2-pairs. Probably too risky, as you point out.

Perhaps one way to approach it is to have a pot of "risk capital", allocated to multiple systems, if you can be reasonably sure the systems are uncorrelated.

Anyway, food for thought ...

 

[skc]

Thanks for the thoughts skc - I need to think about this. I have about 40k allocated to this strategy at the moment, but that's trading 2-pairs. Probably too risky, as you point out.

Perhaps one way to approach it is to have a pot of "risk capital", allocated to multiple systems, if you can be reasonably sure the systems are uncorrelated.

Anyway, food for thought ...

I actually think that trading 2 pairs make it less risky, assuming the pairs are uncorrelated as you said (although all markets are correlated somewhat).

Find out what impact there is if you say implement a hard stop at $1k - I'd imgaine you would manage your maximum loss better, but you may have larger average loss as you basically chop off some of the loss recovery.

Without knowing much about your system, $40k isn't a bad starting point. As you trade more and gain more intimate knowledge of how well your system works and how well you cope with trading this strategy, you can refine your risk parameters down the track (hopefully with a larger account thanks to the cumulation of profits).

 

[Punta]

I actually think that trading 2 pairs make it less risky

Sorry, I meant I've actually been trading 2 contracts for this pair, so definitely not less risky!

This exposes me to a max. losing trade of at least $6k, with $40k capital, and ca. $15k minimum margin. Crazy???


I think I'll wind it back to a single contract each, and look for other pairs.

Initial indications are that FTSE/DAX works on a similar trigger ...