Fully automated futures trading

[globalarbtrader]

Which is just to say the yield-curve seldom inverts, no?

Not exactly. AG was very careful in his posting:

but did curves move into contango (not yield curves but like-for-like futures term structure curves)?

I will now put my bond nerd hat on.

To be nitpicky the difference between say Z16 and H17 10 year bond futures is roughly (assuming they have the same deliverable bond and let's assume it has 9 year maturity): the difference between a 3 month 8.75 year forward interest rate and a 6 month 8.5 year forward interest rate. Now it's late here and I find it hard to imagine what that looks like.

So instead and more intuitively if I decompose the returns I expect to get from holding a H17 bond future from now to expiry assuming the yield curve stays exactly the same I'll earn both:

- carry which is the difference between the 8.75 year interest rate and the GC funding rate (which will be positive unless the yield curve inverts or something crazy happens to the GC rate that's implicit in the funding cost of the future). That's 'carry' in the sense that we're earning a higher yield than it costs to fund the bond.
- and rolldown; which is effectively the difference between a 8.75 year interest rate and an 8.5 year interest rate i.e. it's a forward rate. This comes from the 'aging' of the bond; as it gets younger it's interest rate 'slides' down the curve and this pushes up the price.

Now it's possible that if the yield curve is close to flat but still upward sloping but then slopes down just before the relevant maturity that you'd end up with small positive carry plus negative rolldown: total equating to negative contango on the future. It's unusual but it does happen.

All this is irrelevant to us simple futures traders of course, only nerdy economists and people trading giant swap books (as I did as a young man) have to worry about this kind of minutiae. Nowadays I take the attitude that I can measure carry directly from futures prices, and it seems to predict futures prices, and I don't really care what it all means.

GAT

 

[AvantGarde]

Hi GAT, I did not mean to get into a discussion about exercising discretion over a systematic strategy, but rather hear about the mechanics of strategies in a rising rate environment. There's been a lot of debate about how trend will perform as rates rise, but little debate on carry by itself - it's always mentioned in conjunction with trend. I was hoping to delve into this.

Aside from that, I was wondering, do you have any information or can you point to any great sources that discuss trading strategies on the level, slope and curvature of yield curves?

 

[globalarbtrader]

Hi GAT, I did not mean to get into a discussion about exercising discretion over a systematic strategy, but rather hear about the mechanics of strategies in a rising rate environment. There's been a lot of debate about how trend will perform as rates rise, but little debate on carry by itself - it's always mentioned in conjunction with trend. I was hoping to delve into this.

Aside from that, I was wondering, do you have any information or can you point to any great sources that discuss trading strategies on the level, slope and curvature of yield curves?

I can't think of any. has some stuff. @bone might be able to help.

Perhaps this is another book I could write...

GAT

 

[FCT]

GAT, I've gone through your book and have a doubt. Weights are scaled to -20 : +20 for individual strategies, regardless of whether we are looking at time-series or cross-sectional strategies. You then combine them using weights between 0 and 100%. Doesn't this underweight cross-sectional strategies implicitly?

 

[FCT]

And another one, fishing for your ideas: do you have a few pointers on how to design non-price, macro-based futures strategies?

Thanks for your generosity & great insights this year again. Merry X'mas.

 

[asinger]

Hi GAT--love your book and journal. In the fan club.

A question about the carry/rolldown calculation for bond futures, with an example using GMB Bobl futures on Eurex.

The summary is that I'm getting wildly different carry estimates when I have to change the contracts in the carry calculation around roll dates (a strong short becomes a strong long). How do you handle this?

For carry, I use ln(near) - ln(far) and annualize (* 4), where ideally I'm trading the far contract. With bonds, usually I can trade only the nearest liquid contract and usually have to make do with the same estimate from the far, illiquid contract.

In my example, I'm getting very different values for carry/rolldown once I have to switch far contracts.

For example, let's say I'm short the GBM in the March contract during the period of time in December where both Dec and March have liquidity. Let's say I need to roll on Dec. 8, and now I'm using the June contract as far contract, but trading the near contract.

My carry/rolldown calculation changes from a strong short to a strong long, but I'm still trading the March contract. Example prices around end of roll period:

EUREX/FGBMZ2016

2016-12-08 131.11
2016-12-07 131.09
2016-12-06 131.08

EUREX/FGBMH2017

2016-12-12 132.74
2016-12-09 132.95
2016-12-08 132.7
2016-12-07 132.39
2016-12-06 132.28

EUREX/FGBMM2017

2016-12-12 130.63
2016-12-09 130.96
2016-12-08 130.7
2016-12-07 131.34
2016-12-06 131.33

Annualized log differences when trading the March contract and measuring carry from the Dec. contract:

2016-12-08 -4.82%
2016-12-07 -3.95%
2016-12-06 -3.65%

Same time period, annualized log differences trading the same March contract, but using the June contract in the carry calculation:

2016-12-08 6.07%
2016-12-07 3.19%
2016-12-06 2.88%

The traded contract is the same, but the value flips sign and now suddenly I need to be long in the same contract.

What would GAT do? Thanks!

 

[globalarbtrader]

And another one, fishing for your ideas: do you have a few pointers on how to design non-price, macro-based futures strategies?

Thanks for your generosity & great insights this year again. Merry X'mas.

Can you give me an example?

GAT

 

[globalarbtrader]

Hi GAT--love your book and journal. In the fan club.

A question about the carry/rolldown calculation for bond futures, with an example using GMB Bobl futures on Eurex.

The summary is that I'm getting wildly different carry estimates when I have to change the contracts in the carry calculation around roll dates (a strong short becomes a strong long). How do you handle this?

For carry, I use ln(near) - ln(far) and annualize (* 4), where ideally I'm trading the far contract. With bonds, usually I can trade only the nearest liquid contract and usually have to make do with the same estimate from the far, illiquid contract.

In my example, I'm getting very different values for carry/rolldown once I have to switch far contracts.

For example, let's say I'm short the GBM in the March contract during the period of time in December where both Dec and March have liquidity. Let's say I need to roll on Dec. 8, and now I'm using the June contract as far contract, but trading the near contract.

My carry/rolldown calculation changes from a strong short to a strong long, but I'm still trading the March contract. Example prices around end of roll period:

EUREX/FGBMZ2016

2016-12-08 131.11
2016-12-07 131.09
2016-12-06 131.08

EUREX/FGBMH2017

2016-12-12 132.74
2016-12-09 132.95
2016-12-08 132.7
2016-12-07 132.39
2016-12-06 132.28

EUREX/FGBMM2017

2016-12-12 130.63
2016-12-09 130.96
2016-12-08 130.7
2016-12-07 131.34
2016-12-06 131.33

Annualized log differences when trading the March contract and measuring carry from the Dec. contract:

2016-12-08 -4.82%
2016-12-07 -3.95%
2016-12-06 -3.65%

Same time period, annualized log differences trading the same March contract, but using the June contract in the carry calculation:

2016-12-08 6.07%
2016-12-07 3.19%
2016-12-06 2.88%

The traded contract is the same, but the value flips sign and now suddenly I need to be long in the same contract.

What would GAT do? Thanks!

This is a well known effect in german bonds.

Use some kind of smoothing. A moving average with a length of 180 days would remove this effect.

FYI I use an exponentially weighted moving average of carry, with various different periods. The slowest smooth is 10 business days, the slowest about 150 days.

GAT

 

[globalarbtrader]

GAT, I've gone through your book and have a doubt. Weights are scaled to -20 : +20 for individual strategies, regardless of whether we are looking at time-series or cross-sectional strategies. You then combine them using weights between 0 and 100%. Doesn't this underweight cross-sectional strategies implicitly?

I have thought about this for 5 minutes and I don't think it does. But if you disagree maybe you can provide a numerical example?

GAT

 

[FCT]

Can you give me an example?

GAT

Is there information in low-frequency, supposedly lagging, macro data that can be exploited trading futures? Do you have ideas where to begin and how to do this?