Fully automated futures trading

[globalarbtrader]

@globalarbtrader I just wanted to know, what were the consequences of open sourcing pysystemtrade? I'm thinking about open sourcing my system, which goes all the way from quandl through to IB. My hope is that others will use it, improve it and help iron out the bugs. I've never done an open source project before.

Advantages-
Yes bugs are found more quickly
Pressure to do things properly - documentation, tests, regular updates to cope with pandas or IB api changes etc
Some people have come along and contributed, but you shouldn't assume this will happen in any great numbers - open sourcing isn't an easy way of getting lots of people to help you for free

Disadvantages-
People bothering you with technical support questions
Because you have to do things properly, more time consuming

GAT

 

[globalarbtrader]

If it’s a cash settled future, my intuition would be to let it expire. You get guaranteed savings on the trading costs of one leg. The gap is zero expectation, unless you have a prior about it’s bias (in which case you should make a strategy out of it). In fact, I actually do this if the listed rolls are liquid but trading at an adverse level.

Thinking about this more deeply: yes the expectation of the gap for a longer term trader will be close to zero, but not exactly zero. If for example I take S&P 500 then I might have an average expectation of a 20% annual move (roughly 1 SR) which is 0.08% per day. Half the bid/ask is 0.125 price points and the commision works out to another 0.02; in % terms thats roughly 0.005% (I can usually trade more cheaply because of my execution algo, but lets be conservative); if I can't trade the spread cheaply then it would be twice that 0.01%. Although the costs are of course more certain than the return this is a big enough gap that I'd probably stick to rolling this particular contract. However the maths might work differently on another contract so I accept it's not a universal truth.

In case of a physical settlement with an active EFP market (assuming you can take delivery and the deliverable is the same across months), it’s still to the traders advantage to let it expire. In fact, it’s better situation than a cash settled futures since you will need to do a single EFP transaction to get back into a futures position and there is no gap risk. So the only time I see legging a roll as a better deal if it’s a physical settlement and there is no EFP market.

Of course, this is a purely theoretical discussion since I cannot, for the life of me, think of a contract that is liquid but does not have an active roll market (at least an implied one).

I find Korean bonds are very hard to roll using a spread, but that is also down to time pressure (the roll happens over one day, and is it's the middle of the night I don't want to risk the roll failing so I normally do the roll as two separate legs).

PS. Not trying to be adversarial or calling you out, simply want to understand your thought process (especially when we arrive at different conclusions).

All taken in the spirit of healthy discussion, and I love having my assumptions questioned.

GAT

 

[maciejz]

If it’s a cash settled future, my intuition would be to let it expire. You get guaranteed savings on the trading costs of one leg. The gap is zero expectation, unless you have a prior about it’s bias (in which case you should make a strategy out of it).

My experience with letting VIX futures expire was not very positive My system missed some rolls earlier this year (due to a combination of technical issues). One of the months early in the year, the gap between the last trade price (I believe it was 11.5) and the settlement price was over 1 point (I believe settlement was at 12.65). Since I was short, that lesson cost over $1K per contract Just to ensure that I had properly learned the lesson, I "audited" the course again a few more months (although the spread was never as large again).

(Yes, I do trade the front VIX contract despite the fact that GAT advises against it. My excuse is that when I initially "calibrated" my system I haven't noticed at that time that GAT recommends the 2nd VIX contract and I've been too lazy to recalibrate. Actually, I've been working on an improved approach to trading Volatility -- and since this "improved" approach has been "only a few days away from being complete" for the last 6 months, I didn't want to take the time to recalibrate the existing system

A bigger issue is that it seems to me that the VIX settlement is being manipulated. There was a paper listed on Quantopian (March timeframe I believe) about the manipulation of the VIX settlement -- apparently its pretty doable due to the "cheap" deep out of money options that go into the VIX calculation and the fact that the "settlement" takes prices from only about 1 minute of time -- so the manipulation does not have to be long lasting. To your point about having an expectation about this and creating a strategy, I have thought about it. I haven't looked into it though mostly because I'm scared

I have not noticed a similar spread between last trade and settlement in other markets where my system missed rolls, however any delay in closing positions that result from expired futures contract can pose significant risk. On my end, while my Futures trades are automated, any "cleanup" of currency positions due to expired contracts has to be done by hand. Since I am lazy, it usually takes me a while (several days) to get around to this. Murphy's law usually kicks in the meantime -- I was in the middle of manually entering orders to close the FX positions that resulted from expired short GBP and EUR contracts just as the FED announced a rate increase (May or June timeframe). Surprisingly, the dollar actually plummeted due to the FED commentary released along with the rate increase. The result was another "cattle prod" moment.

Since the question of the "best" method has been posed, I figured that sharing some personal experiences with one of the approaches mentioned would potentially be helpful.

 

[sle]

The belief about VIX expiry being manipulated is rather outdated - the new settlement process (circa 2015) makes it expensive to do so. What you have seen is an actual demand for the strip due to the arbitrage trades. Anyway, that’s an aside here.

 

[AvantGarde]

Hi all,
Let's say you were using back-adjusted futures prices that had experienced so much cumulative backwardation that historical values were negative. Let's say you were computing a simple momentum system using the percentage change in price over say the last day and you were working with values such as -1, -0.5, 0.01, 0.05. If you calculated the percentage change on these daily values, this would yield -50%, -102% and + 4900%, which are not useful. How can you deal with this problem? I guess you just use the normalized price delta instead of percentage change. I am wondering how the academics work with regressions on historical returns with back-adjusted data...

 

[globalarbtrader]

Hi all,
Let's say you were using back-adjusted futures prices that had experienced so much cumulative backwardation that historical values were negative. Let's say you were computing a simple momentum system using the percentage change in price over say the last day and you were working with values such as -1, -0.5, 0.01, 0.05. If you calculated the percentage change on these daily values, this would yield -50%, -102% and + 4900%, which are not useful. How can you deal with this problem? I guess you just use the normalized price delta instead of percentage change. I am wondering how the academics work with regressions on historical returns with back-adjusted data...

You're exactly right - you shouldn't use % changes, but use delta prices.

GAT

 

[FCT]

I’m a bit confused about the roll discussion. Apart from Korean bonds (which are apparently different), why would you ever decide to do anything else but to trade the calendar spread? As in, suppose you’re long 5 lots DEC17. Before expiry, you sell 5 lots of the DEC17-JAN18 spread and you are now 5 lots long JAN18. Calendar spreads are (at least) 50% cheaper than 2 outrights and you have continuous exposure. I really see no reasonable alternative, or...?

 

[Gambit]

I’m a bit confused about the roll discussion. Apart from Korean bonds (which are apparently different), why would you ever decide to do anything else but to trade the calendar spread? As in, suppose you’re long 5 lots DEC17. Before expiry, you sell 5 lots of the DEC17-JAN18 spread and you are now 5 lots long JAN18. Calendar spreads are (at least) 50% cheaper than 2 outrights and you have continuous exposure. I really see no reasonable alternative, or...?

Reduces variance and comes with a ready made hedge. Earns the carry in some cases too. But maybe that is a little too clever. There's got to be a catch. Commissions? Lack of an implied market in many cases which means execution risk... Just spitballin.

 

[HobbyTrading]

Calendar spreads are (at least) 50% cheaper than 2 outrights and you have continuous exposure.

Can you explain why it is at least 50% cheaper? I tried this with my IB account and noticed that the commission for a spread was exactly the same as for two outrights.

 

[sle]

I’m a bit confused about the roll discussion. Apart from Korean bonds (which are apparently different), why would you ever decide to do anything else but to trade the calendar spread?

The most common case is when a roll is very rich or cheap to fair (which happens all the time for some futures). Instead of locking in negative value by trading a calendar spread it might be worth waiting for the expiration and re-entering the position after the roll cycle is over and situation normalizes (if you can take EFP or the futures is cash settled, obviously). We were specifically taking a case were rolls are either mispriced or can not be traded at all.

Can you explain why it is at least 50% cheaper? I tried this with my IB account and noticed that the commission for a spread was exactly the same as for two outrights.

It's the same commission costs, but rolls are usually quoted tighter and you can frequently roll at mid (via voice or electronically).