Double Butterfly spread

cdcaveman said:
All this information in your post are things I've picked up over time separately... And through some pain... Pascals binomial expansion is exactly the way risk is hedged with futures...
More...

Right, I was simply responding to your reply to @Wingz specifically with regards to Pascal's triangle and also since you had stated in the first post of the thread "what does this (double butterflies) isolate?" The guy had pointed out:

pascal_triangle_excerpt4.png


All of this is interesting for understanding the principles at work, however it looks pretty academic beyond a double butterfly. But in keeping things in the spirit of full retard, here's a synthetic 13th order Triple Rainbow Guy spread of WTI crude from M6->M7:
https://www.youtube.com/watch?v=OQSNhk5ICTI
cl_13th_order_triple_rainbow_spread.png

https://www.youtube.com/watch?v=OQSNhk5ICTI
Even if we pretend that chart had any semblance of accuracy, we can safely assume slippage would put an end to that "idea" pretty quick. Assuming an average of 5 ticks of slippage per lot (and that's probably optimistic) and perfect entry (but of course), 2048 lots later you'd have burned 10,000 ticks in slippage just getting the position on for a 1.5 year duration 7000 tick move. Sign me up! :D

Here's a double butterfly which seems a lot more reasonable:

cl_db_spread.png


And even a "4th-difference spread of boxes" (1:-4:6:-4:1) for fun:

cl_tb_spread.png


All of these are of course synthetic charts of exchange traded calendars, so the dubiousness probably increases as we go down the rabbit hole - but hey, slow weekend and I'm the curious type.
 
BTW: I noticed in the ICE SPAN guide that there's actually another variant of overlapped flies that involves sharing only the outer wing of each fly as compared to sharing the body and wing (which is a double butterfly/DF). They call it an extended double butterfly and the construction is generally +1/-2/0/+2/-1 where the center exp is canceled out due to the wing overlap. In a way this is slightly more "condorish" in nature as the body isn't shared but it's still spreading two adjacent flies against each other.

image.jpeg


IB even recognizes this with Euribor combos but it should be generally usable across anything.

Note, this extended double-fly can also be creating with a full body subtraction of 2 condors (similar to fly-fly == DF) or sum of 2 double-flys like so:

Code: 
CF-CF(3) => 1:2:0:2:1 (8 => 12, EF)

a-b-c+d

 -b+c+d-e

a-2b+0c+2d-e

DF+DF(3) => 1:2:0:2:1 (16 => 12, EF)

a-3b+3c-d

   b-3c+3d-e

a-2b+0c-2d+e
 
Last edited:
Would love to hear some thoughts on when someone would use this... I imagine you could +1/-2/0/0/-2/+1 as well.. but when does this make sense
 
cdcaveman said:
Would love to hear some thoughts on when someone would use this... I imagine you could +1/-2/0/0/-2/+1 as well.. but when does this make sense
More...

I'm not 100% sure myself on what exactly it's doing vs a DF, but I'd imagine it's doing somewhat less intra-spread hedging as only the wing is shared vs body+wing in the DF case and seems somewhat condorish.

However, given some basics and some brainstorming:

Fly is the difference of two overlapping calendars:
Code: 
SP-SP(1) => 1:2:1 (4 => 4, BF)
a-b
 -b+c
a-2b+c

Condor is the difference of two non-overlapping calendars:
Code: 
SP-SP(0) => 1:1:1:1 (4 => 4, CF)
a-b
   -c+d
a-b-c+d

Condor is also the *sum* of two overlapping flies:
Code: 
BF+BF(2) => 1:1:1:1 (8 => 4, CF)
a-2b+c
   b-2c+d
a-b-c+d

Double fly is the difference of two overlapping flies:
Code: 
BF-BF(2) => 1:3:3:1 (8 => 8, DF)
a-2b+c
  -b+2c-d
a-3b+3c-d

Extended double fly is the difference of two overlapping flies (wings only):
Code: 
BF-BF(1) => 1:2:0:2:1 => (8 => 6, EF)
a-2b+c
    -c+2d-e
a-2b+0c+2d-e

Extended double fly is also the difference of two overlapping condors and the sum of two overlapping double flies:
Code: 
CF-CF(3) => 1:2:0:2:1 (8 => 12, EF)
a-b-c+d
 -b+c+d-e
a-2b+0c+2d-e

DF+DF(3) => 1:2:0:2:1 (16 => 12, EF)
a-3b+3c-d
   b-3c+3d-e
a-2b+0c+2d-e

I guess one would might say that if a condor is the sum of two overlapping flies then similarly an EF/EDF is the sum of two overlapping double flies and if we see "sum" it really means "concurrent" in this context (just like being concurrently long or short outright in two different expirations of the same instrument but with more hedging). I think breaking things down into calendars might make it easier to conceptually visualize (let alone execute):

If we take the sum of two flies example (which results in a condor) and break it down into calendars, it looks like this:
Code: 
(a-b)-(b-c)             (+fly)
      (b-c)-(c-d)       (+fly)
(a-b)      -(c-d) or +a-b-c+d

Which essentially would be a play on two (almost adjacent) calendars vs each other, or alternatively long the wings and short the body like a fly but in this case the body isn't just 1 exp but 2 adjacent expirations.

Since we know that an EF is also the difference of two condors, using the calendar math (note, I pre-negated the other side so that we always use addition):
Code: 
(a-b)-(c-d)             (+condor)
  -(b-c)+(d-e)       (-condor)
(a-b)-(b-c)-(c-d)+(d-e) or +a-2b+0c+2d-e

One kinda notices that there's a bit of chaining between months going on here and that in a way it resembles a "condor of calendars" (the +:-:-:+ relationship) or a "fly of calendars" where the body is not a shared calendar (which results in a double fly) but two adjacent calendars (just like two adjacent outrights in the condor example). This seems conceptually similar to a condor in nature and knowing what we know from the previous example where a condor is the sum of two overlapped flies it now makes a bit more sense why the sum of two double flies creates an extended double fly. To me, it appears to be more condor in nature than a fly (due to the less shared body) so I'd expect there to be similar behavior to a DF in the way a condor can be similar to a fly.

Here's a CL Z17M18Z18M19 DF on the daily:

Chart_16-10-16_14-20-07.png


And here's a EF for the same general months:

Chart_16-10-16_14-19-33.png


Also, Euribor weekly charts DF:

Chart_16-10-16_14-25-10.png


vs EF:

Chart_16-10-16_14-25-22.png


Some ED/GE weekly charts (EF):

Chart_16-10-16_14-50-28.png


Chart_16-10-16_14-50-01.png



BTW, one can also experiment with different takes on where things are overlapped or subtracted from each other, even coming up with things that don't even have a name:

Condor vs condor but using less overlap:
Code: 
CF-CF(2) => 1:1:2:2:1:1 (8 => 8, kinda like a DF but no shared legs)
a-b-c+d
   -c+d+e-f
a-b-2c+2d+e-f or +(a-b)-2*(c-d)+(e-f)

CF-CF(1) => 1:1:1:0:1:1:1 (8 => 6)
a-b-c+d
     -d+e+f-g
a-b-c+0d+e+f-g

Weird fly and condor sums:
Code: 
BF+BF(1) => 1:2:2:2:1 (8 => 8)
a-2b+c
     c-2d+e
a-2b+2c-2d+e

CF+CF(2) => 1:1:0:0:1:1 (8 => 4, CF [widen center])
a-b-c+d
     c-d-e+f
a-b-0c-0d-e+f

CF+CF(1) => 1:1:1:2:1:1:1 (8 => 8)
a-b-c+d
       d-e-f+g
a-b-c+2d-e-f+g

I suspect that what's really going on here in all of this is control over the shape of how the body and/or wings are hedged vs normal flies.
 
i'm thinking mostly you have found a part of the curve you would like to isolate in a more hedged way... like a 1331 or a 14641 the farther you go out pascals trangle the more nuetrality you get.. this 12021 fly might likely be as stable as a 1331 fly, i would think it would be more about the month or months you are trying to isolate in relation to other spreads
 
One advantage I can see is that if you can get a good price on a single cal spanning the 2 middle cals then you only need to execute with 3 cals rather than 4 as the 2 middle cals normally cancel out the center leg when executed. For instance, if it were a 6 month span for everything you could handle the 2 center cals with a single 12 month cal spanning the width. Assuming the 12mo is priced fairly and has a decent bid/ask this may be cheaper overall due to execution costs. However once the legs start getting near 12mos it'll get more difficult as you'd need a 24mo cal to do the same thing.

Also even when some markets provide exchange traded double flies (Eurodollars) that doesn't necessarily mean that's going to be the cheapest way to get in and out. A lot of these more complex exchange traded spreads either have implication disabled or at best only imply based off outright prices. Almost all the time within the energy space it's more efficient to get into a fly, condor, or beyond by using cals only. Even though the exchange provides a native fly book they don't necessarily imply based off the tighter cal bid/asks making for a worse price when using the fly book directly. This also makes risk management more of a hassle then it needs to be.
 
i960 said:
Almost all the time within the energy space it's more efficient to get into a fly, condor, or beyond by using cals only.
More...

FWIW, I will personally use the cals to leg into more complex spread expressions. Quite frequently, in fact.
 
Another obvious note that might be helpful to the newer traders: Even if a spread expression is not listed on the exchange margin sheets, for intra market spreads the exchange will almost always calculate a margin credit offset for you.
 
i960 said:
I'm not 100% sure myself on what exactly it's doing vs a DF, but I'd imagine it's doing somewhat less intra-spread hedging as only the wing is shared vs body+wing in the DF case and seems somewhat condorish.

However, given some basics and some brainstorming:

Fly is the difference of two overlapping calendars:
Code: 
SP-SP(1) => 1:2:1 (4 => 4, BF)
a-b
-b+c
a-2b+c

Condor is the difference of two non-overlapping calendars:
Code: 
SP-SP(0) => 1:1:1:1 (4 => 4, CF)
a-b
   -c+d
a-b-c+d

Condor is also the *sum* of two overlapping flies:
Code: 
BF+BF(2) => 1:1:1:1 (8 => 4, CF)
a-2b+c
   b-2c+d
a-b-c+d

Double fly is the difference of two overlapping flies:
Code: 
BF-BF(2) => 1:3:3:1 (8 => 8, DF)
a-2b+c
  -b+2c-d
a-3b+3c-d

Extended double fly is the difference of two overlapping flies (wings only):
Code: 
BF-BF(1) => 1:2:0:2:1 => (8 => 6, EF)
a-2b+c
    -c+2d-e
a-2b+0c+2d-e

Extended double fly is also the difference of two overlapping condors and the sum of two overlapping double flies:
Code: 
CF-CF(3) => 1:2:0:2:1 (8 => 12, EF)
a-b-c+d
-b+c+d-e
a-2b+0c+2d-e

DF+DF(3) => 1:2:0:2:1 (16 => 12, EF)
a-3b+3c-d
   b-3c+3d-e
a-2b+0c+2d-e

I guess one would might say that if a condor is the sum of two overlapping flies then similarly an EF/EDF is the sum of two overlapping double flies and if we see "sum" it really means "concurrent" in this context (just like being concurrently long or short outright in two different expirations of the same instrument but with more hedging). I think breaking things down into calendars might make it easier to conceptually visualize (let alone execute):

If we take the sum of two flies example (which results in a condor) and break it down into calendars, it looks like this:
Code: 
(a-b)-(b-c)             (+fly)
      (b-c)-(c-d)       (+fly)
(a-b)      -(c-d) or +a-b-c+d

Which essentially would be a play on two (almost adjacent) calendars vs each other, or alternatively long the wings and short the body like a fly but in this case the body isn't just 1 exp but 2 adjacent expirations.

Since we know that an EF is also the difference of two condors, using the calendar math (note, I pre-negated the other side so that we always use addition):
Code: 
(a-b)-(c-d)             (+condor)
  -(b-c)+(d-e)       (-condor)
(a-b)-(b-c)-(c-d)+(d-e) or +a-2b+0c+2d-e

One kinda notices that there's a bit of chaining between months going on here and that in a way it resembles a "condor of calendars" (the +:-:-:+ relationship) or a "fly of calendars" where the body is not a shared calendar (which results in a double fly) but two adjacent calendars (just like two adjacent outrights in the condor example). This seems conceptually similar to a condor in nature and knowing what we know from the previous example where a condor is the sum of two overlapped flies it now makes a bit more sense why the sum of two double flies creates an extended double fly. To me, it appears to be more condor in nature than a fly (due to the less shared body) so I'd expect there to be similar behavior to a DF in the way a condor can be similar to a fly.

Here's a CL Z17M18Z18M19 DF on the daily:

View attachment 167355

And here's a EF for the same general months:

View attachment 167356

Also, Euribor weekly charts DF:

View attachment 167357

vs EF:

View attachment 167358

Some ED/GE weekly charts (EF):

View attachment 167359

View attachment 167360


BTW, one can also experiment with different takes on where things are overlapped or subtracted from each other, even coming up with things that don't even have a name:

Condor vs condor but using less overlap:
Code: 
CF-CF(2) => 1:1:2:2:1:1 (8 => 8, kinda like a DF but no shared legs)
a-b-c+d
   -c+d+e-f
a-b-2c+2d+e-f or +(a-b)-2*(c-d)+(e-f)

CF-CF(1) => 1:1:1:0:1:1:1 (8 => 6)
a-b-c+d
     -d+e+f-g
a-b-c+0d+e+f-g

Weird fly and condor sums:
Code: 
BF+BF(1) => 1:2:2:2:1 (8 => 8)
a-2b+c
     c-2d+e
a-2b+2c-2d+e

CF+CF(2) => 1:1:0:0:1:1 (8 => 4, CF [widen center])
a-b-c+d
     c-d-e+f
a-b-0c-0d-e+f

CF+CF(1) => 1:1:1:2:1:1:1 (8 => 8)
a-b-c+d
       d-e-f+g
a-b-c+2d-e-f+g

I suspect that what's really going on here in all of this is control over the shape of how the body and/or wings are hedged vs normal flies.
More...
When I look at many of these charts, for example CL Z17M18Z18M19 DF, it appears to be much more volatile, i.e. ragged, than the underlying outright CL so why trade this as opposed to the outright as with this volatility your stops would have to be quite wide.
 
You must log in or register to reply here.
Back
Top