The hot area in trading for quite some time has been algorithmic trading.  They are in every tradable market – fixed income, equities, commodities, and FX.  The algos are some of the larger players, so you need to factor them in the negative-sum game that is futures trading.  This week, I thought I would discuss some of the more obvious trading features on the ED curve that stand out.  I’m not saying these are definitely the result of algos – it’s just that these things about the ED curve are not “natural.”

The year-end algo. 

160509 Z algoMost of you probably have noticed that the Z contracts have had a ~50% larger OI and volume than the surrounding contracts.  See the spikes on the OI chart (ED contracts on the x-axis).  If you look at EDZ7 (ED7), its OI is 438K larger than the average of the surrounding contracts.  This is 72% higher!  This effect has been around for a few years, and I’m assuming things have been going well, because it seems to get larger every year.  When I first noticed this 4 years ago, I couldn’t figure out if it was just a coincidence these were the Z contracts, and that after a contract roll, we would expect to see the H contracts with the larger volume.  But it’s definitely only been the Z contracts.  The daily trading volume also moves in proportion to the size of the outstanding OI, so it’s hard to believe these are manual orders.  I haven’t made up my mind what this is:

  • It’s natural for people to take positions on year-end rates, so it’s logical for the OI on those contracts to be higher. But 50% higher seems a little too much.  Maybe partly.
  • It could be some kind of year-end model-based trading. I suppose if the FOMC makes multi-year year-end projections off their models, some large funds could probably do at least a good a job and trade off of that.
  • It’s not clear if it is year-end turn-related.

However, the level of the Zs don’t seem to get too out-of-line on the curve, despite the massive size of the OI.  So it’s hard to pin it on anything.

The local algo.

In the old days, the locals and a lot of curve traders made a reasonable living just trading the distortions caused by large positions, in both futures and options.  I’m sure there are a bunch of locals who have automated what they had been doing, which is basically smoothing the curves.  Why do by hand, what you can have a machine do while you are at the pub?

The execution algo.

There are various algos out there that alleges to give the most mathematically efficient and discrete fill.  There’s at least one broker I know that does this, so it makes sense that some of the larger hedge funds may do this as well.  These types of algos would have the effect of reducing the number of “kinks” in the curve.  So this could also explain some of the curve smoothing.

The options algo.

I remember in the old days when the options locals would carry physical sheets of papers with them and they would flip pages when there were large price and vol moves.  Nowadays, everything is automated, including probably the hedging.

So how to make money off of this?

I think with the rise of the machines and the flatter curve, a smoother curve is to be expected.  But last week, even I was surprised at how flat the 6mo spread curve was.  This week, we got a little “movement,” but the theme is the same.  On the right is the turn-adjusted 6mo calendar spread curve.[1]  Last week, we were in a 1bp wide band and this week we are in a 1.8bp band (the light green box on the chart).

160509 algo moneyOne thing I’ve noticed is that there is a sharp drop-off in “algo” activity at the end-points.  This makes a lot of sense.  The nuance of what occurs in the next 6-12 months is probably harder to model.  And out in the purples and beyond, there is less liquidity (and fewer hedging options in other markets), so it is probably better for the machines to stay away.  So this leads to two themes of value (in blue arrows):

  • Since I lean relatively bearish (compared to the market), I like positions which are equivalent to buying calendar spreads in the back whites and front reds (starting with U6 and out). This also “happens” to roll positively.
  • In the purples, it seems like buying calendar spreads starting with U1 and out also makes sense. But you need to size small.  There’s a reason the purples and oranges are called the “roach motel” – you can get in, but you can’t get out.  All right, that’s an exaggeration, and you get paid for providing liquidity.  But size small.  This also “happens” to roll positively.

I’ll look for some specific things early next week.  I also suspect the algos aren’t very good at a large change in the shape of the curve.  This is why we keep doing the way-OTM conditional call steepeners.  But I don’t see a good bearish “story” worth pursuing right now.

[1] I made a 0.5bp adjustment to ED16 and 0.25bp adjustment to ED18, since we had an “off-settle.”

[This was published in the May 9 issue of the Curve Advisor]