Double Butterfly spread

greg25 · Jan 4, 2016

Hi All,

Bone, or anybody how used TT Autospreder or CQG. Is worth to pay for it???

I mean I am trading spreads, flys and condors. I like the idea that i can do it synthetic. Nat Gas Flys, or condors, same Cl flys and condors.. or brent ones...

I am also looking for same LME spreads, the inter market one.. Comex vs LME.. or same Coffe europe vs ny

But as i see the autospreder is very expensive.. 3K (the platform, market fees, and connection fees). Is worth that money ?

Thanks

Greg

bone · Jan 4, 2016

Greg, I used to use it all the time and for the past several years I do not. YMMV. For the past several years I have also not been trading at high intraday frequencies like I used to. Now I swing trade spreads.

My advice would be first things first: check the exchange messaging restrictions regarding your products of interest.

greg25 · Jan 5, 2016

Ok, thanks.

I am not a intraday trade, I mean I am looking for same intraday but i am also a swing trader, so maybe just "cts sniper" is enough.

Thanks a lot, i will check the market restrictions.

Greg

i960 · Mar 10, 2016

So revisiting this after some time...

Wingz said:
When you get into a fly you buy and sell an equal number of spreads. Going long a 6 month spread is the sum of going long six 1 month spreads, you could build it that way by buying each month individually but then you'd be paying more commissions.

If I take what you say literally, we have:

dec-jan
jan-feb
feb-mar
mar-apr
apr-may
may-jun

Which of course simply cancels out to just dec-jun.

Wingz said:
So in order to hedge that you need to equivalently sell 6 lots of a 1 month, that's all it is, you just need your exposure to be equal.

However, we know that between dec-jun there could be all manner of stuff going on that could dislocate the curve, such that individual 1-month spreads would have different characteristics than the 6 month. True, while summed they'd equal out (ignoring friction), but individually they represent something different of course. Based on what you said in a previous post:

Wingz said:
If I'm looking at a block of 6 months (in energy markets I'll trade in DecJun blocks) and I think... Hmm FebMar seems a little lower then the others I'll buy 6 FebMar spreads and sell 1 DecJun spread.

What I'm saying is that relative to all the other spreads in DecJun I think that FebMar will rise more. So im essentially exposing myself to 6 long FebMar contracts and I'm short 1 of every spread within DecJun (including FebMar so really only 5 contracts of FebMar against 1 of each of the other 5 spreads in the DecJun group).

Hope this makes sense. They're some of the most ranging mean reverting spreads I've found.

In essence it seems like what you're trading here is normalization of the curve rather than how the curve itself changes over time. But to do that, it assumes a predictable and consistent curve shape (dare I say smooth and logarithmic), right? How would you even apply something like that to NG or HO for example which tend to have various disruptions in the curves? I guess one could have an idea of how normal said "disruption" actually is based on history but then why not just target that portion of the curve explicitly? Additionally, by being +6 feb-mar and -1 dec-jun aren't you actually really exposed to that feb-mar portion of the curve if some kind of seasonal or otherwise dislocation happened?

I feel like there is a hidden risk here in that 6 x feb-mar doesn't completely translate to 1 x dec-jun even if dec-jun can be mathematically composed of 6 individual 1 month calendars, inclusive, because it assumes a smooth normalized curve is the "mean." Maybe in markets where the latter months tend to be proportionally affected by "disruptions" of earlier months this translates well, but I'd think markets which can get hit with heavier seasonal effects there could be potentially other concerns. Am I missing something?

e.g. non-"smooth/logarithmic" curves:

And a typical (for these days) Brent curve:

Given what you said earlier, I'd imagine you'd be targeting jan/feb'18 here (a week ago)?

cdcaveman · Mar 10, 2016

the position that i first described just for everyone's information did work out..

the x16z16f17 fly this was trading at -.14 now mid is -.04
the the double fly the vxzf.. this was around -.12 at one point for a long time it was -10 now trading at -.01

bone · Mar 10, 2016

Now this is just my philosophy - but all I am looking for is convergence or divergence. In other words, there are dislocations in the forward curve induced by commercial order flows. Especially as you transition from the Demand aspect of the curve to the Supply aspect of the forward curve. Doesn't really matter to me if it's a calendar or a butterfly or a condor intramarket or intermarket.

Wingz · Mar 10, 2016

i960 said:
So revisiting this after some time...

If I take what you say literally, we have:

dec-jan

jan-feb

feb-mar

mar-apr

apr-may

may-jun

Which of course simply cancels out to just dec-jun.

However, we know that between dec-jun there could be all manner of stuff going on that could dislocate the curve, such that individual 1-month spreads would have different characteristics than the 6 month. True, while summed they'd equal out (ignoring friction), but individually they represent something different of course. Based on what you said in a previous post:

In essence it seems like what you're trading here is normalization of the curve rather than how the curve itself changes over time. But to do that, it assumes a predictable and consistent curve shape (dare I say smooth and logarithmic), right? How would you even apply something like that to NG or HO for example which tend to have various disruptions in the curves? I guess one could have an idea of how normal said "disruption" actually is based on history but then why not just target that portion of the curve explicitly? Additionally, by being +6 feb-mar and -1 dec-jun aren't you actually really exposed to that feb-mar portion of the curve if some kind of seasonal or otherwise dislocation happened?

I feel like there is a hidden risk here in that 6 x feb-mar doesn't completely translate to 1 x dec-jun even if dec-jun can be mathematically composed of 6 individual 1 month calendars, inclusive, because it assumes a smooth normalized curve is the "mean." Maybe in markets where the latter months tend to be proportionally affected by "disruptions" of earlier months this translates well, but I'd think markets which can get hit with heavier seasonal effects there could be potentially other concerns. Am I missing something?

e.g. non-"smooth/logarithmic" curves:

View attachment 162758

View attachment 162759

And a typical (for these days) Brent curve:

View attachment 162760

Given what you said earlier, I'd imagine you'd be targeting jan/feb'18 here (a week ago)?

Hey i960,

My process is that I've got all the possible spread permutations that I trade manually set up as charts that I can double click and scroll through multiple times a day. In Brent I look through around 150 different charts for around 50% of the time I'm in front of the screen every day.

I base my decisions mainly on price action and the context of the market, is it trending, reversing, ranging, mean reverting. Looking at correlations, hedges, paper moves. Taking into account the timing structure of the market, settlement times, expiry dates and any patterns I can pick up on.

I honestly have no idea what the curve looks like in the way you charted it, with term structures. Which is a big red flag in my head, it's another area I will take steps to improve upon. So given that, I'm not sure I'm in the best position to shed light on your insights.

However, in terms of dislocations (or things going out of whack), through continually looking at the charts and ladders I can see those dislocations taking place and have a general (probability based) idea of where they're likely to extend to based on experience and forward testing.

An example right now is that the last few days over settlement there's been a huge seller of the dec16-jan17 spread over settlement which has huge ripple effects across the curve. I don't know why he's deciding to sell it, but I understand the implications and the effects on the relative value of the other spreads and combinations around it.

I don't have any positions on past 2017 so I wouldn't have seen anything in 2018, however now that you've given me a nudge, I might be able to automate some of my manual processes so I can look at a broader spectrum of trades.

I'll take a look at the structures, cheers for the heads up!

bone · Mar 10, 2016

Ummmm..... If you are limiting yourself to 2016 and 2017 you are truly missing most of the fun. Seriously. And that goes for any listed product with decent open interest 2018 and beyond. You'll have to wait for prime trading hours but the spreads are always tighter and more liquid than the out rights.

Wingz · Mar 10, 2016

bone said:
Ummmm..... If you are limiting yourself to 2016 and 2017 you are truly missing most of the fun. Seriously. And that goes for any listed product with decent open interest 2018 and beyond. You'll have to wait for prime trading hours but the spreads are always tighter and more liquid than the out rights.

Haha, I know. There's been a nagging feeling in my head for a while to focus more on the backs. I'm comfortable in what I'm doing now, so the intensity has gone down and I have more time to allocate.

I think ill have to finally put some motivation behind finding a few new strategies in the backs. A very different ballgame, but like you said more based on prime trading hours, so something I can develop over time.

Wingz · Mar 14, 2016

i960 said:
So revisiting this after some time...

If I take what you say literally, we have:

dec-jan

jan-feb

feb-mar

mar-apr

apr-may

may-jun

Which of course simply cancels out to just dec-jun.

However, we know that between dec-jun there could be all manner of stuff going on that could dislocate the curve, such that individual 1-month spreads would have different characteristics than the 6 month. True, while summed they'd equal out (ignoring friction), but individually they represent something different of course. Based on what you said in a previous post:

In essence it seems like what you're trading here is normalization of the curve rather than how the curve itself changes over time. But to do that, it assumes a predictable and consistent curve shape (dare I say smooth and logarithmic), right? How would you even apply something like that to NG or HO for example which tend to have various disruptions in the curves? I guess one could have an idea of how normal said "disruption" actually is based on history but then why not just target that portion of the curve explicitly? Additionally, by being +6 feb-mar and -1 dec-jun aren't you actually really exposed to that feb-mar portion of the curve if some kind of seasonal or otherwise dislocation happened?

I feel like there is a hidden risk here in that 6 x feb-mar doesn't completely translate to 1 x dec-jun even if dec-jun can be mathematically composed of 6 individual 1 month calendars, inclusive, because it assumes a smooth normalized curve is the "mean." Maybe in markets where the latter months tend to be proportionally affected by "disruptions" of earlier months this translates well, but I'd think markets which can get hit with heavier seasonal effects there could be potentially other concerns. Am I missing something?

e.g. non-"smooth/logarithmic" curves:

View attachment 162758

View attachment 162759

And a typical (for these days) Brent curve:

View attachment 162760

Given what you said earlier, I'd imagine you'd be targeting jan/feb'18 here (a week ago)?

To follow on from the 'paper seller' I saw in DecJan previously. I got short that spread today right before settlement expecting trading on the bids and anticipating a tick or two. However, it was like a western movie with dust clouds brushing along the horizon, he was nowhere to be seen. I quickly hedged with the 6 month DecJun and put in my bid to scratch in the DecJan with priority. Managed to get the DecJan back and make 2 ticks on the DecJun, so got out with a 1/6ish of a tick loss net overall (on the 1 month, given costs of a half tick and a 1/6th allocation to the 6 month).

However given that information I went long the 3 month fly DecMarJun which was at range lows. The settlement seller wasn't there so the DecMar area was not going to be supressed, it was at lows in a range so a long over settlement made sense as a high probability trade into the afternoon session.

For me these kinds of observations provide the most edge in my own trading with the spreads. However as bone pointed out the backs are really the most exciting part of the curve, there's a whole lot less competition and free trades being handed out left and right for a market maker.