Forums  > Pricing & Modelling  > Market Making Illiquid Parts of a Futures Curve  
     
Page 1 of 1
Display using:  

EspressoLover


Total Posts: 245
Joined: Jan 2015
 
Posted: 2016-04-01 06:38
Say you're looking at a futures curve like Nat Gas. (Nothing particularly special about NG, just thought it's a nice archetype.) Volume is heavily concentrated in one or two expiries at the front of the curve. But there's still liquidity and regular volume for a dozen or so expiries further out in the curve. They might trade at 1-10% of the volume of the front-month contract, but they still have pretty good liquidity. Usually no more than a few ticks wide spread, with pretty good size on the touch throughout the day.

It would seem that market makers are pretty active in these contracts and quote pretty competitively, at least given the relatively low volume. What are the common approaches that liquidity providers here are taking? Based on this thread, I would guess it's not altogether too different than quoting front month options.

That is make most of your edge when the front contract moves and you pick off resting orders in back contracts. Lean your quotes in the direction of keeping higher order curve factor and expiry-specific exposure to a minimum. Maybe collect some spread edge when the front month is looking stable. Does this reasonably describe how most market participants are operating?

The main component I'm unsure about is how exchange supported calendar spreads change this equation. Obviously a cal-spread has much lower net exposure to curve level/PC1, at least relative to higher order curve factors. I'm not sure if the calculus of spread quotes would significantly changes the outright markets.

Lebowski


Total Posts: 66
Joined: Jun 2015
 
Posted: 2016-04-01 15:10
[double post]

Lebowski


Total Posts: 66
Joined: Jun 2015
 
Posted: 2016-04-01 15:10
I've wondered about this myself, and I have no idea what *explicitly* they're doing. Picking off resting orders in the less active contract is a big part of it I'm sure. What I've always also considered is they may be trying to leg into synthetic spread positions at an edge. Does that theory hold any water?

Luciender


Total Posts: 74
Joined: Aug 2008
 
Posted: 2016-04-01 16:46
I know thats what many locals did back in the day in stir futures, staring at their tt/autospreader screen all day until they get a favourable fill on one leg. I suppose they don't exist anymore tho..

Patrik
Founding Member

Total Posts: 1338
Joined: Mar 2004
 
Posted: 2016-04-01 18:46
For commods spread contracts can often be more liquid than the outright further out contracts. That's quite natural as e.g. a producer of NG comes to hedge next winter, so market maker gets given next winter gas, he turns around and sells the most reasonable liquid point to offset flat price risk (i.e. could be this winter gas), then he bids the spread (in this case this winter vs next winter spread) and over time manage to even up his spread position. So you have natural triangles of spr vs legs, but also spr vs spr vs legs etc etc. I'd assume there's at least some of that going on from the market making machines. Not being in that game I'd assume they live more in the statistical world in terms of how they make decisions. The time horizons are short and liquidity high so they can ignore a lot of the bigger picture.

But it's worth noting that for e.g. NG the curve does move, and move a lot. Winter gas vs summer gas can be almost like 2 independent commodities (depending on storage levels etc). At higher frequencies that may not be so important, i.e. you're literally executing legs pretty much at same time, but for the market maker example above it's a big risk. If we instead of next winter made it a few years out and it was a big clip the spread position could take a long time to work out of. Then the human market maker will probably not be living in the statistical world but much more in the world of "art" and factor in fundamental view of the spread, knowledge of other hedging going on, knowledge of existing positioning, was this clip only shown to him or a bunch of other market makers and hence others know his risk, etc etc. The list goes on and is highly subjective.

Capital Structure Demolition LLC Radiation

HockeyPlayer


Total Posts: 125
Joined: Nov 2005
 
Posted: 2016-04-01 22:11
If you looking at the price ladder to gauge market maker presence, make sure you are differentiating between implied bids/offers and outright bids & offers.

Looking at NGU6 right now, the implied market is 2 ticks wide, but the outright market is 23 ticks wide.

EspressoLover


Total Posts: 245
Joined: Jan 2015
 
Posted: 2016-04-05 08:17
@ Patrik

> I'd assume [machines] live more in the statistical world in terms of how they make decisions... Winter gas vs summer gas can be almost like 2 independent commodities (depending on storage levels etc). At higher frequencies that may not be so important, i.e. you're literally executing legs pretty much at same time, but for the market maker example above it's a big risk.

I think this an insightful point. In the "statistical world" when you're trading on tight electronic quotes most of your counter-parties are other bots. Getting picked off is probably much more of a risk than dealing with "natural flow" like producers looking to hedge. I think the principals are generally the same: get paid to take risk, keep exposures contained, be cognizant of hidden factors accumulating in the portfolio. But in the electronic world you're dealing with substantially more toxicity, particularly with regards to price discovery in the liquid contracts.

W.r.t to the NG winter vs summer, that's an interesting observation. This may be going off on a tangent from the original point, but playing around with the data I've found that the NG curve behaves different around different times of the year. Front-year summer vs winter contracts seem to move more tightly together during the summer, with maximum disconnection peaking around November. But even during November the longer-dated summer contracts are still trading with betas of 0.3 or higher to the front winter months. If you're market making in the former that gives you a little more freedom, but you can still be picked off quickly if the front contract moves a few ticks.

@ HockeyPlayer

Ahh, that seems to to explain a lot. Great point, thanks for pointing that out. I was looking at consolidated quotes and assuming that the contribution of implied calendar liquidity was negligible. Bad assumption. I'm going to have revisit the data...

I'm guessing the matching engine updates implied quotes across contracts atomically, which changes the nature of the game. If most of the liquidity is coming from spreads, it's probably directly or indirectly chained to the liquid front month. If that's the case and front ticks up, that probably propagates a change in spread-implied quotes down the curve. In which case picking off stale quotes, or even hoping to reprice an outright quote after a tick is hopeless. Though you could still try to pre-anticipate front month ticks with some sort of alpha.

Just thinking out loud, it may also be the case that spread implied quotes might cause back-curve prices to overreact to front month price changes. On the sizable majority of futures curves the "true" beta of price changes from the back to the front month is well less than 1.0. There may still be opportunity to pick off quotes in the illiquid back, but it might be after, not before, the front month price tick propagates.

Patrik
Founding Member

Total Posts: 1338
Joined: Mar 2004
 
Posted: 2016-04-05 16:28
@EspressoLover: all fundamentals of NG are inherently seasonal, so expect everything about it's pricing to be seasonal as well :)

Capital Structure Demolition LLC Radiation

radikal


Total Posts: 258
Joined: Dec 2012
 
Posted: 2016-04-05 17:35
The implieds are also worse Q than the true outrights, no? It's been so so so many years since I've traded on CME.

I used to trade far dated calendars occasionally on CME many years ago, though in products without implieds. (FX/Equity)

Your ability to "pick off" "stale" orders in the illiquid is, to some degree, a function of your pricing in the calendar. If the calendar is 2 ticks wide, but you're willing to trade aggressively to buy the middle tick, you might be able to hop the resting order before the crowd. This leads to weird game theory with potentially flashing an order in the mid tick of the calendar or something..

There are no surprising facts, only models that are surprised by facts

Nonius
Founding Member
Nonius Unbound
Total Posts: 12699
Joined: Mar 2004
 
Posted: 2016-04-10 23:45
I've a friend next to whom I sat (not the MMaker I mentioned in another thread) who was a specialist in doing a sort of agency business in dividend futures. He was "amazed" at the fact that when he hit/lifted one contract in the curve, amazingly, and automatically, the shorter /more liquid contract would trade in lockstep and full cadence. he demonstrated this amazing fact to me by click trading one contract. I told him my theory that some fuck was just doing a sort of "pairs cointegration" semi hit semi quote trading strat by quoting the illiquid stuff at the longer end and eating the more liquid end when he was hit/lifted. and doing it automatically. fairly sure I'm right but the guy was sort of perplexed.

Chiral is Tyler Durden

radikal


Total Posts: 258
Joined: Dec 2012
 
Posted: 2016-04-11 03:59
That's how most 3rd party / off-the-shelf systems by-default work. You specify that you're willing to do the spread for 30 ticks and the tool automatically moves quotes in the illiquid and auto-hedges in the liquid upon fill.

This is circa 2006-2007. People generally try to "squeeze" and not pointlessly cross the liquid unless it seems necessary these days. (Conditional auto hedge)

There are no surprising facts, only models that are surprised by facts

rp6


Total Posts: 24
Joined: Jun 2012
 
Posted: 2016-04-21 16:38
You're possibly attributing too much intelligence to the majority of the crowd! I haven't done this since 2012, but have put some energy spreads on since then. I imagine behind the vast majority of the orders on the screen are an ecosystem of arcade punters using a basic TT/CQG/Stellar autospreader with a few placebo-inducing bells and whistles thrown in by the sales rep. Spreading mainly flies and condors, maybe 12m calendars include the back; some spreading against other parts of the energy complex even where no fundamental arb relationship exists*; (plus some spreading in vain against copycat products?). I imagine there is very little outright activity beyond the first two contracts and the first set of quarterlies, so all the volume you see beyond that is a ricochet of autohedges setting off other spreaders.

I like the game theory aspect mentioned here, in the back calendars you often see a 1 or 2 lot flash repeatedly to improve the bid/offer, which i think was a strategy to job all the spreaders out of their queue position and/or see where the resting orders were.

bramj


Total Posts: 48
Joined: Sep 2009
 
Posted: 2016-04-25 19:13
If you have access to quotes on strips and are willing to assume that price changes are jointly normally distributed, it's possibly to analytically determine the expectation of each future price conditional on the strip value. This requires having a good estimator for the covariance matrix, and from what I recall on doing on this, it helped to deseasonalize data first.

The advantage of this approach is that it's fully analytical so very quick and it can be made to work with having multiple quotes available; so e.g. it can be combined with the bid-offers you determine by applying some shortest path algorithm to all spread prices, or incorporate summer, winter and quarterly strips if you have them all. I never got this to work in full generality with using both bid and offer information though, and then there is the normality assumption...

rftx713


Total Posts: 98
Joined: May 2016
 
Posted: 2016-08-05 01:32
Patrik's posts definitely match up most closely with what I saw on the energy trading floor, but also some pretty interesting ideas in here as well.

Sorry, didn't realize I revived a 4-month old thread.... my bad
Previous Thread :: Next Thread 
Page 1 of 1