1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

June 15, 2016 - Kevin Lee - Facts and Myths - how IV impacts your butterfly trades

Discussion in 'Round Table Presentations' started by tom, Jun 16, 2016.

  1. tom

    tom Administrator Staff Member

    This forum thread is to discussion Kevin Lee's presentation "Facts and Myths - how IV impacts your butterfly trades."

    Last edited: Jun 16, 2016
  2. TH Lui

    TH Lui New Member

    Hi Tom, will Kevin's presentation via webex be posted as video for us for view?
  3. tom

    tom Administrator Staff Member

    I just added the YouTube video to the first post in this thread. Enjoy!
  4. TH Lui

    TH Lui New Member

    Thank you Tom. Appreciate it!
  5. Boomer34

    Boomer34 Well-Known Member

  6. William Smith

    William Smith Member

    Holy smokes! Great stuff. Extremely useful thinking.
  7. DavidF

    DavidF Well-Known Member

    Agree, thank you Kevin, you´re always extremely generous with your considerable knowledge. It´s clearly complex, especially since it seems like the skew doesn´t necessarily flatten in response to a spike in VIX.

    A question: Kevin has demonstrated that the BWB performs better than a standard M3 on a drop in volatility when the far right leg gets crushed. Accordingly the M3 performs better on a down move with a spike in volatility. However, if VIX had already spiked and you think "Ok I´ll put on a BWB so my far right leg doesn´t get crushed on rebound and declining vol", isn´t this essentially the same as shorting volatility after a spike in VIX? (which is expressed via a flattening of the skew curve)

    in backtrader I placed either a SPX BWB or M3 on the 21st of August 2015 after VIX has already spiked with vicious drop on the Friday. With Monday´s continued down move the M3 still performed better than the BWB.

    So even with the knowledge of how skew affects butterfly prices, do you still end up market timing based on whether you think vol is expanding or contracting and how this will affect skew?
    Last edited: Jun 18, 2016
    Rakesh, N N and Gail like this.
  8. Kevin Lee

    Kevin Lee Well-Known Member

    Hi David,
    I'm not advocating market timing because IV skews may move in ways that surprise us. However, IV skews, in a way is more predictable than general market movements because they tend to mean revert and certain skew changes are guaranteed. For example, If IV skew flattens during a down move, it's most likely that it will steepen when the down move stalls or reverses.

    In addition to underlying movements creating a IV skew change, the passage of time will cause IV skew to steepen too. This IV skew change is guaranteed and we can and should plan for it. Therefore there are logical trading decisions which we can take that leverages on our knowledge of IV skew.

    FYI..... I just created a video in response to some of the questions I received about M3 vs BWB. I introduced a different way of analyzing options positions, a simplified way. I created a separate post for it. Go take a look ...... https://forums.capitaldiscussions.com/threads/m3-vs-bwb-a-reductionist-approach.620/
    Rakesh, Wolfgang, PKMTrader and 3 others like this.
  9. N N

    N N Well-Known Member

    Fantastic information.
    Question: When looking at the slope %, what levels are considered 'expensive skew' and what is considered 'cheap skew' based on your calculation?
  10. Kevin Lee

    Kevin Lee Well-Known Member

    Hi NN,
    I think there isn't one number or even a range of slope % to benchmark against. What is a reasonable slope % depends on : (1) which underlying, (2) DTE, (3) absolute IV level and (4) recent market environment. So the way I use this information is more visual than a specific numerical number, ie I look at the graph visually to see if it has been flattening or steepening.
  11. Todd

    Todd New Member

    I really appreciate you sharing your knowledge Kevin. Very interesting and helpful to understand what is happening with these trades. Makes me feel the need to learn about trading the M3.
    Kevin Lee likes this.
  12. guwe75

    guwe75 Active Member

    Hi Kevin,

    very interesting topic.
    One possible way to standardize the measurement of the skew would be moneyness or delta.
    If you measure the difference in % between the Mid IV 90 vs. Mid IV 110 (or Mid IV Put delta 25% put vs. Mid IV Call Delta 25%), the slope and slope changes could be compared among underlyings and over time in terms of
    a) flat or steep skew
    b) change of skew
    c) DTE and absolute IV level
    Rakesh likes this.
  13. Kevin Lee

    Kevin Lee Well-Known Member


    I don't think I get it. Can you give an example how do you translate the difference in IV % of the two strikes to the slope of the curve?
  14. AKJ

    AKJ Well-Known Member

    Kevin, if you measure the skew by translating the strikes as a percentage of the underlying, this will give you a measure that is comparable across time.

    As an example, suppose you have the following two options.

    50 DTE, 1140 strike put trading at 25% IV (underlying at 1200)
    50 DTE, 1200 strike put trading at 20% IV (underlying at 1200)

    One measure of the slope is (25 - 20)/(1140 - 1200) = -.0833333. This measure is not comparable across time since the underlying price will change making the denominator different across times.

    A better measure (in my humble opinion) is (.25 - .20)/(.95 - 1.00) = - 1. The denominator here translates the strike prices into a percentage of the underlying. This allows for a metric that is more comparable across times.

    A few caveats to note:

    1) As you have demonstrated, skewness tends to steepen as the option chain approaches expiration. This is a market behavior that is called sticky delta. Because of this, any measure comparing skew slope across time would have to look at options with similar DTE. As an example, you might consider comparing skews across time for options with 30-45 DTE.

    2) The volatility skew is not linear, so one linear measure of skew slope may not explain the behavior of the entire option chain. I have found in my own analysis though that the non-linearity of the IV curve, especially when there is sufficient DTE (say more than 14 DTE), tends to kick in well above the spot price. For the structures that I trade, I tend not to have options on the IV curve where the slope has gone positive, so a linear measure of skew slope has been more than sufficient as a single metric to compare option skews across time.
    Rakesh likes this.
  15. guwe75

    guwe75 Active Member

    Attached Files:

    Last edited: Jun 21, 2016
    Rakesh likes this.
  16. DGH

    DGH Administrator

    One of the best studies on the subject of volatility skew, IMO, is Scott Mixon's academic paper published in the Summer 2011 issue of The Journal of Derivatives entitled "What Does Implied Volatility Skew Measure". Interestingly, he found that, of all the possible ways of determining skewness, the most useful for him is: (25 delta put IV minus 25 delta call IV)/50 delta IV. From his article, "Perhaps the most interesting chart is the bottom right one. For the (25-delta put volatility–25-delta call volatility)/50-delta volatility, there is virtually no dependence on the ATM volatility. Roughly speaking, more left skewness means a higher value for this measure of skew, irrespective of the level of volatility. This ease of interpretation makes it a very attractive measure according to the theory." He reviewed several other methods of determination of skewness, most being suitable only for quants with big computers and big paychecks.
    Bax, Rakesh, DavidF and 1 other person like this.
  17. Kevin Lee

    Kevin Lee Well-Known Member

    @Andrew John, @guwe75 - thank you very much for the detail explanation.

    To be frank, I haven't spent much time thinking about the slope comparability over time and over different underlying values. I agree if that is the objective, then we need to normalize with moneyness or delta. Neither have I put much emphasis on standardization of the measurement. Perhaps these are good candidates of future projects for me :)

    My main interest in using the slope is to determine how my position has been affected by IV skew change and to make better trading decisions on a day to day basis. So, I do not compare slopes of positions across wide time scale, eg. today's slope vs few years of historical data to determine if it's flat or steep. Therefore, I'm more concern with how the slope has changed from yesterday or last week. I usually rely on a quick visual inspection of the excel graph and an analysis of my excel table. If upper strikes IV increased more than lower strikes, that's flattening and vice versa. I don't really pay much attention to the actual slope % value.

    Moreover, IV movement is only half the equation. IV will only affect the time value of options and time value depends on how close the strike is to money. Therefore, even if there could be a bigger IV movement for a particular strike, but if carries tiny time value, it's also less relevant to another strike with less IV movement but bigger time value.

    In addition, I only look at the part of the IV curve that is relatively linear, ie the lower strikes to about slightly above money. The non-linear portion, ie upper strikes where the U shape is will distort the slope calculation. I exclude those strikes from calculation. That's also because I don't trade butterflies above market. Hence that portion is less important to me.

    This is a good discussion. Thanx.
    Last edited: Jun 21, 2016
    Rakesh likes this.
  18. vega4mike

    vega4mike Well-Known Member

    For those interested in the paper, here's a copy.
    Read & Enjoy:)

    Attached Files:

    DavidF likes this.
  19. Steve S

    Steve S Well-Known Member

    I decided a long time ago that people spend to much time and energy on identifying one or two numbers to measure the skew (10-delta puts, 25-delta calls, etcetera). It turns out to be very workable to design an approach that can be used to measure skew for any strike relative to the atm, and more important for any two strikes on the curve (very useful for identifying certain overpricings/underpricings of butterflies).
    The basic "building block" local model for the equity skew is the very simple log-linear curve: vol(x) = vol(x_atm) + Beta * [log(x/x_atm) / vol(x_atm) * sqrt(T)].
    Here Beta is interpreted as "change in skew per unit change in Black-Scholes standard deviation". Note that in using this model we are NOT postulating incorrectly that the skew is actually log-linear (although normally this is a very good approximation to the put skew) ... rather, we are simply using this as a local model that can explain and measure a lot of the skew behavior in useful ways ... basically this model is used to explain/remove so-called "first-order effects", whereupon the more subtle items of interest become more easily apparent.
    This elemental model captures to first order the 1/sqrt(T) time-steepening behavior of the skew and the 1/vol(x_atm) vol-up-flattening behavior and vol-down-steepening behavior. It is first-order independent of DTE; that is, using this elemental model automatically goes a long way towards apples-to-apples comparison for different DTEs. The Betas are easy to calculate on a spreadsheet, both for strikes relative to the money [Beta(x, x_atm)] and local Betas between any two strikes [Beta(x1, x2)].
    Example of Beta's usefulness for butterfly trading: Say the put fly strikes are x1, x2, x3 in descending order. The "basic skew" for the trade might be measured as Beta(x_atm, x3), the "basic skew of the fly" would be Beta(x1, x3), and the concavity/convexity of the fly (which is critical for fly pricing) would be Beta(x1, x2) - Beta(x2, x3). Instead of using one artificial skew measure, you are using skew measures customized to the particular trade.
    So instead of hanging your hat on one number like 25-put minus 25-call, it only takes a little more effort, and some experience time, to get a feel for what Betas are normally and what abnormal Betas are ... you replace "one skew number" of limited usefulness with "one skew formula" of very extensive usefulness.
    One more thing: You can use simple mathematical tricks to "translate" Betas to something you are more familiar with ... I usually look at Betas in terms of the old standard ratio 10-delta-put-vol / atm-vol, where something like 1.3 is flat and 1.6 is steep ... so when I started using Betas there was no experience-time learning curve as they were stated in terms of a number I already had a handle on.
    My 2 cents.
    Murphy Tan, Capt Hobbes and vega4mike like this.
  20. DGH

    DGH Administrator

    Very interesting expression for skew, Steve. I am 50 years (yes, really) out of calculus, but I basically "see" it and the rationale behind the equation. If I transposed correctly, Beta=[vol(x)-vol(x_atm)] / [log(x/x_atm)/vol(x_atm)*sqrt(T)]. Is this LOG base 10 or natural logarithm (ln)? Also, the equation as shown is returning a negative number; the absolute values are in the 2.0 to 2.5 range, suggesting increasing steepness, of course. Am I interpreting the formula correctly? Why the negative number?

    Thanks for sharing this.

Share This Page