Using Spot for Carry

[dr1ver1]

Leveraged Trading describes the option of using Spot for Carry, which I'm using for some instruments in my own system. I'm migrating to Pysystemtrade, and haven't seen this described in the documentation. Does the system support this?

[robcarver17]

nope

[robcarver17]

You'd have to mess with the simData objects to replace the 'raw data for carry' methods

[dr1ver1]

I'm still at the early stages of properly understanding Carry. My assumption is that if Spot is liquid it would be the most accurate thing to use:

• the nearest month would have to converge to it
• if Spot has far more volume than the futures it would be the future converging to Spot, as opposed to Spot converging to the future, which is key to how the Carry rules makes a profit

...there's clearly something fundamental I haven't understood.

[robcarver17]

Yes it's valid to use spot if you're trading the first month, the reason I don't is it involves getting an extra source of data over and above just a load of futures prices. Sometimes this is easy (eg S&P 500), often it's hard (US 10 year bond spot price anyone)?

"if Spot has far more volume than the futures it would be the future converging to Spot, as opposed to Spot converging to the future, which is key to how the Carry rules makes a profit" is probably bollocks even if we could meaningfully measure the relative volume of spot and futures....

[dr1ver1]

I'm using spot for S&P 500, Gold, GBPUSD, but using Next or Previous for other instruments I trade.

For my understanding, just sticking with futures, and using next or previous contracts for Carry, according to Leveraged Trading,
p.205:
Rule 1. Carry using Nearer contract. We're trading Dec 2018 Corn and Sep 2018 is lower: the rule assumes, if underlying prices are stable, Dec 2018 will drop to Sep 2018 level (and so go short)

p.207:
Rule 2. Carry using Next contract. We're trading Sep 2018 Euro Stoxx and Dec 2018 is lower. The rule assumes Sep 2018 will rise to Dec 2018 (and so go long)

The net expected return is the price differential annualised.

2 questions:

• For Rule 2, why is the assumption that Sep 2018 rises to current Dec 2018 price, rather than Dec 2018 falling to current Sep 2018 price? (in which case net expected return on Sep 2018 would be 0)
• Or for both Rules 1 and 2, instead assume both contracts converge and meet in the middle, in which case net expected return would be halved?

I don't understand what determines which contract will move and which will stay unchanged.

[robcarver17]

Think of it like this (I just did a lecture on this, and this explanation seems to work on my students)

Imagine we're trading some future, call it THING
Suppose it's now 15th Feb, and we have expiries on 15th March, 15th April, ....
In time space these are spot, one month, two month, three month....
And suppose the prices of THING are spot=100,one month=99, two month = 98.5, 97...

Carry basically assumes that all prices will remain the same. So in one months time, the price of spot will still be 100. The price of one month THING will still be 99. the price of two month thing will still be 98.5. And so on.

Example 1: (Current / spot) On 15th Feb we bought one month THING for 99. We wait a month. The thing we've bought is no longer one month THING, it's now spot THING (as it's expiring today, and a future will always equal spot price at expiry). The price of spot THING is still 100, since we assume all prices remain the same. So we've made 100-99 =1 price unit of carry

Example 2: (Current/nearer) On 15th Feb we bought two month THING for 98.5. We wait a month. The thing we've bought is now one month THING. The price of one month THING is still 99 as we assume nothing changes. We've made 99-98.5 = 0.5 price unit of carry.

Example 3 (further/current): On 15th Feb we bought one month THING for 99. We don't have a spot price, and we can't use current/nearer, since there is nothing nearer. We assume that the carry of spot/current is equal to the carry of current/further. We expect to get 0.5 unit of carry. In fact we'd get 1 unit, since this assumption was wrong.

[dr1ver1]

Thanks! OK, this makes a big difference how I need to think about this.
And so for Example 3, this will result in inaccurate forecasts - but we have nothing better to go on.