Fully automated futures trading

sle · Jan 20, 2018

truetype said:
Presumably all of a trendfollower's trades are (believed to be) "alpha legs."

I guess. I was just giving a general idea

For example, if I was a vol seller via put or call spreads, thats where this method would really shine - in fact, I'd recon that despite the higher costs due to split executions it would be still worth it.

itb · Jan 22, 2018

globalarbtrader said:
Ignoring the issue of whether the annual vol target is too high;

The first issue is that 80K isn't going to generate much 'income'. I guess you could put it into money market funds and earn 1.5%, i.e. $1,200 a year. I have a mortgage, five kids and a cat to feed and I need more like a 100 grand. So this works fine if you have nearly 7 million bucks spare.

Alternatively you can take capital risk on your 'income' producing asset and with an unleveraged mixture of stocks and bonds probably make 4%. I'd 'only' need 2.5 million to make this work [and in fact this is obviously pretty close to what I do].

The final problem is that your 'income' is going to be correlated to your trading performance. Let's say you lose 30% of your notional (not unlikely given a 30% vol target at some point). At that point your margin requirement is 60%*140K = 84K and you have 140K - 84K = 56K left in the bank. So now your 'income' producing asset has fallen by 30%. I don't know about you but I can't cut 30% of my income that much.

The other problem is that of the 60K we have lost we've had to cough up 40% of that from our 'income'. If we've taken capital risk with our 'income', and that's also done badly, we're screwed.

What you can't do is say assume you can draw down that 80K, perhaps over a couple of years if you need 40K a year, in the event of a drawdown.

GAT

Thanks for the reply! To be clear, my plan is to keep the 80K in the bank account and withdraw 1000 every month as long as there are funds in the account. I don`t plan on investing the 80K in anything riskier than a deposit/savings account. And of course when I make losses on the futures trading, the amount in the bank account would be used to meet margin calls.

For a strategy with a given Sharpe and skewness, it seems to me one of the biggest decisions is what risk to target. The reasons referred to so far mostly have to do with how much losses one can take mentally before losing sleep. In the Systematic Trading book you do refer to the Kelly rule, but indeed the vol target suggested by the Kelly rule is too spicy for me. Thus it becomes a trade-off between having the optimal target according to Kelly and being able to sleep well, which is very qualitative and subjective decision in the end...

pauljherrera · Jan 25, 2018

Hi Rob,

I am building an integration of pysystemtrade and IB (I know, I know, you are building it too... but I didn't have the patience).
I am making some tests, and the only data that IB provides free of charge is Forex data. So, I have made some changes to the code in order to use it with Forex. Nevertheless, I got stuck in something apparently very simple but that is making my forecasts to look not very accurate: the block value.

How would you calculate the block value for, let's say, USDJPY and EURGBP, taking a Standard Lot (100,000$) as the minimum amount you can invest?

Thanks again!

globalarbtrader · Jan 26, 2018

pauljherrera said:
Hi Rob,

I am building an integration of pysystemtrade and IB (I know, I know, you are building it too... but I didn't have the patience).
I am making some tests, and the only data that IB provides free of charge is Forex data. So, I have made some changes to the code in order to use it with Forex. Nevertheless, I got stuck in something apparently very simple but that is making my forecasts to look not very accurate: the block value.

How would you calculate the block value for, let's say, USDJPY and EURGBP, taking a Standard Lot (100,000$) as the minimum amount you can invest?

Thanks again!

The block value is the value of a 1% change in the value of a block.

If you own $100K of USDJPY or EURGBP the block value is just $1K

GAT

isotope1 · Jan 27, 2018

@globalarbtrader is there a preferred method for learning parameters in an online fashion, where a backtest isn’t possible?

globalarbtrader · Jan 27, 2018

isotope1 said:
@globalarbtrader is there a preferred method for learning parameters in an online fashion, where a backtest isn’t possible?

What do you mean by 'online'? When you say 'learning' do you mean 'finding the optimal set of parameters according to some criteria'?

GAT

isotope1 · Jan 27, 2018

globalarbtrader said:
What do you mean by 'online'? When you say 'learning' do you mean 'finding the optimal set of parameters according to some criteria'?

GAT

Yes exactly, but where you can only use live trading to inform you of the parameters. I'm conscious of using a naive method that may overfit. I can handcraft a prior, but still need a reasonable posterior.

traider · Jan 27, 2018

isotope1 said:
Yes exactly, but where you can only use live trading to inform you of the parameters. I'm conscious of using a naive method that may overfit. I can handcraft a prior, but still need a reasonable posterior.

I use kalman filters for linear stuff.

isotope1 · Jan 28, 2018

traider said:
I use kalman filters for linear stuff.

How?

I think the issue might be related to this:
https://en.wikipedia.org/wiki/Multi-armed_bandit

traider · Jan 28, 2018

isotope1 said:
How?

I think the issue might be related to this:
https://en.wikipedia.org/wiki/Multi-armed_bandit

Identify a linear relationship between variables related to your trading
Once you have that, then you need to calibrate the Kalman Filter

isotope1 said:
How?

I think the issue might be related to this:
https://en.wikipedia.org/wiki/Multi-armed_bandit

It really depends on what problem you have formulated. For me I formulated my trading into a linear equation so I could use a KF.