RIsk Ranges
- This topic has 111 replies, 20 voices, and was last updated 2 years ago by
.
-
Hi Guys I also have been trying to Figure Out HE Risk Ranges.
I’m pretty sure he is using Implied Vol for each asset. The average Implied Vol of the SPY for example is basically the VIX . For instance the VIX closed today at 31.89 and the ImpVol for SPY close is 32.18. I have begun collecting this data daily for around 50 assets.
What I have come up with so far is to Convert the implied Vol to a Daily number (Square root(1/252)) and +/- from the previous days close. This creates a 1 Std Dev range around the previous days close. This has been getting me close with some of the assets. Today HE SPX was 2780 – 2899. Using my Ivol method got me 2794 – 2911. He somehow weighted the Lower End of the Range 1.23 Std Dev and the Upper End .0.79 Std Dev.
I will begin researching Hurst exponents since he is somehow weighting the Upper and Lower Ends of the Ranges.
I was looking into Skew and Kurtosis, but I didn’t get far.
5 users thanked author for this post.
I tried the the approach you are taking with the implied vol (Daily number (Square root(1/252)) and +/- from the previous days close). The issue I had was that while the risk range would expand as IVOL goes up, the top end of the range would go up and the bottom of the range would go down, essentially the range would get wider around the previous day’s close. At around the 7 minute mark of the April 28th macro show Keith said, “If I were to take the VIX in my model, which I am using around 31-32, inside of 30 I could get at least to 2925” (vs the published top end of the risk range of 2899). In the past Keith has said as the VIX goes up the risk range expands so using these two ideas the calculation has to both expand the range and drop the top end of the range when the VIX goes up and do the opposite when the VIX goes down.
One interesting thing to note that I had not heard previously, or I may have just missed it, is that he said he was using 31-32 in his model to calculate the top end of the range. The VIX had closed at 33.29 the previous day while the published risk range on the VIX was 31.94 to 47.36. This leads me to believe that whatever model Keith uses to calculate the actual risk range is based on the low and high end of the Volatility range for the security. I think you have to use the Vol of Vol for a security (for the SPX it would be the VVIX) to generate a risk range of Volatility (VIX) then use the risk range of the volatility (VIX) to calculated the risk range of the security.
If you go back to the post I made last week in the notes I attached it shows the volatility factors and they definitely use skew and the cost of hedging. Using the skew, both the trend of the skew and the skew across different strike prices for the same expiration, are signals of bullish/bearishness that others have used. This is something I found (not Hedgeye) when reading about skew, “By comparing the current 30-Day Constant Maturity Skew to historical values, we can know whether the market is shifting in a bearish or bullish trend. When 25-Delta put skew moves from being +10.0% historically to being +15.0% currently, that would indicate that the markets are more bearish than usual, as traders are paying a higher relative premium for the expectation of a downside move.” Based on all of the macro shows I have listened to this sounds just like something Keith would say.
4 users thanked author for this post.
Glad to see all the new faces in this thread… blue mannequin faces but still. More hands make less work.
Question for @Smokeshow and @JL.
With this (Daily number (Square root(1/252)) formula, is Daily number the IVOL for that day? Would you guys mind elaborating on that formula?Also @Smokeshow, I think you may be right about how Keith uses Hurst exponents. The rescaled range by itself doesn’t produce numbers that I’ve been able to make use of, but the Hurst exponent being above or below 0.5 would be useful to determine whether the price was reverting to the mean or if a trend was continuing. He may use it as another indicator, like volume, but not necessarily something that’s used in the range equation. @derricksikes posted a paper that I still need to look into on creating buy/sell signals from the hurst exponents. Here it is again.
I was able to build a fractal calculator in Python that can tell you rescaled range, hurst exponents, r-squared, all that good stuff. Shoutout to @derricksikes for providing the article that made that possible. The author provides an example Excel spreadsheet that made it much easier. It shouldn’t be hard to turn it into a simple website that generates CSVs that you guys can use in your own research, I’ll see about making that happen.
In response to @KillJoy on the Donchian channels, I agree with you. It will be misleading on its own. But it seems like one of the best ideas we’ve put forward as of yet for determining the skew of the data (%upside and %downside).
2 users thanked author for this post.
@JohnLawsCarriage that IVOL number is a daily vol. To make sense of this I would refer you to the law of 16 here: https://tickertape.tdameritrade.com/trading/rule-16-volatility-skew-vix-16768
Simple way to think of it is the VIX @ 16 implies a 1% move in the S&P, VIX @ 32 implies a 2% move. using the sq route of 252 trading day (roughly the number of market days in the year) gives you an est of daily IVOL.
The offer still stands I have a bbg terminal so any data needed please let me know and I can provide.
2 users thanked author for this post.
@Smokeshow, ” The issue I had was that while the risk range would expand as IVOL goes up, the top end of the range would go up and the bottom of the range would go down, essentially the range would get wider around the previous day’s close. ”
Have you have tried dividing the newly calculated range with a function of the rate of change of ivol/rvol and readjust the upper & lower bounds accordingly? I notice that when the prior days close crosses Keith’s upper bound, bouth his bounds seems to be increasing and vice versa. This tells me that his risk ranges don’t change that much as long as the close is within the range, but when the close crosses the upper/lower boundaries, BOTH tend to change.
Stuck in a remote placed because of this lock down… I need to start playing with these numbers on Jupyter. Is there a good website that I can get IVol data? Ivolatility.com seems to be a good website for historical ivol data … Any other recommendations?
Also, I will need realtime ivol data, what is the most recommended website that I can use to get ivol data using an API? IVolatility.com’s API uses COM… I prefer a linux API… wondering if there are other sources that I use to get realtime ivol data? thanks.
2 users thanked author for this post.
I have not tried to divide the calculated range by IVOL/RVOL yet. I will as I move forward.
After writing the post last night I started thinking about something I had noted that Keith said on 4/28, he was using 31-32 in his model when the published risk range on the VIX was 31.94 to 47.36 with the previous day’s close at 33.29. I decided to go back and look at the published VIX range vs. the published SPX range. Looking back to May 1st it seems like the relationship is inverse and that the VIX risk range is the only thing that goes into the calculation of the SPX risk range. The reason I say this is that it matches what he said on 4/28 and if you go back and look at the low end of the SPX risk range on 5/11 & 5/12 they are both 2806 while the top end of the risk range on the VIX was 39.85 and 39.84, respectively.
This leads me to believe the key to the model is figuring out the relationship between the Vol of Vol and the Volatility and that any use of the volume is option volume and not volume of the primary security. I think he might use the volume of the primary security as a confirming factor but not in the actual calculation.
Just as a recap of where I am on this, none of this is proven fact just educated guesses.
The rescaled range (Hurst exponent) is used to determine trend, mean reverting or Brownian motion
The risk range of the primary security is driven off the risk range of the volatility of the security.
The risk range of volatility is driven off the volatility of volatility
The bullish/bearish trend is based on skew, cost of hedging and may include IVOL/RVOL premium/discount.
A few notes:
IVOL is not a daily number. It can be calculated on any time frame where options are available. You can use options that expire in 30 days to get a 30 day IVOL number or use options that expire in 90 days to get a 90 day IVOL number. IVOL is the “unknown” value in the options pricing model and, since all of the other inputs are known, it can be backed out. VIX is the 30 day IVOL of the SPX and VVIX is the 30 day IVOL of the VIX.
Rather than focus on individual securities I am focusing on the SPX. I figure if I can figure out the model for the SPX it will be easy to apply it to individual securities. There is far more information on the SPX, both market data and from what HE publishes, than individual securities.
Some of the things I am looking at as possible models that are being used are:
1) The Directional Movement Index combined with Momentum. In an old post DJ mentioned that Bruce Babcock was the closest practitioner to what HE was trying to do. Unfortunately Bruce Babcock died so his web site is not up anymore but I did find some old articles he had written and he mentioned Edgar Peters books and that he used the Directional Movement Index with Momentum and and open/closing price indicators. I have ordered some old books that Bruce Babcock wrote to see if there are any clues.
2) Fractal Adaptive Moving Average (FRAMA) model. I know Keith does not follow the “moving monkey” but does use support and resistance levels that tend to flow from moving averages so a moving average based on The Brot’s fractals would be somewhat of a fit.
3) Fractal chaos bands and fractal chaos oscillator. I have done the least amount of work on this one but it did come across my radar while doing my research.
5 users thanked author for this post.
I wanted to add two things both hurst exponents and how they are used to evaluate trend strength (attachment 1) and the IVOL/RVOL calc – The last fund I worked at did a lot of options trading and this is how we would calculate it (attachment 2). For the hurst exponent a reading about .50 is a strong trend reading below is a weak trend. For the Vol screen I have 30 day put implied vol vs 30 day realized. Let me know if you have any questions.
4 users thanked author for this post.
I was looking into what you said, about the VIX risk range effecting the SPX risk range. I backtested it back to February 21, definitely a strong positive correlation there. Check out this graph.
How I made this graph:
Grabbed risk ranges (RR) for SPX and VIX.
Subtracted the lower RR number from the upper RR number to create a range of the RR, (so if top of the RR is 40 and bottom is 27, the range of the RR = 13).
Got the natural logarithm for both SPX and VIX RR ranges to easily compare the two.Pretty strong positive correlation, when running =CORREL() in excel, r = 0.77
2 users thanked author for this post.
Great discussion. Also a HE subscriber that has been trying to reverse engineer the RR. I see the volume point has come up a few times. Has anyone tried using VWAP, which combines volume and price?
1 user thanked author for this post.
Hey @JohnLawsCarriage
check out the attached bbg implies the same IVOL for both puts/calls so can just use one of the columns.
3 users thanked author for this post.
I thought a SPX chart with the risk ranges might be a helpful visual, so I (manually) created one for the last 12 months. It was as much fun as it sounds. I generated one for VIX as well. The PDFs are attached. The SPX chart is a little hard to read because of the large scale, so I am attaching the Excel spreadsheet so you can “zoom in” on a time period of either low or high volatility by changing the referenced data range. I color coded bullish trends green, bearish red and neutral yellow. I am new to PRT and just downloaded the trial version a few days ago. Do any of you know how I could bring a fixed data set like the SPX risk ranges into PRT as a custom indicator to compare it to the test ones I have created in PRT?
3 users thanked author for this post.
I know very little about PRT and just downloaded the trial version a few days ago. Attached is a simple custom indicator I have been tinkering with based on the rescaled ranges spreadsheet. It certainly does not match the Hedgeye risk ranges and it doesn’t scale well at all. As you increase the number of days, the risk range becomes very large. I am wondering if using hourly intervals or something like that is worth looking at or if I am just way off track here. What are your thoughts?
1 user thanked author for this post.
Hi all,
Great discussion here and proof that others want to use the Risk Ranges for stocks/securities that KM doesn’t cover!
I think we have to look at this in terms of how ‘the machine’ operates. KM advises a number of major hedge funds, so he will know from this and his time running funds, how many of the algos operate. They can only really operate in 2D, using maths. I know from previous experience that many algos will use the VWAP, so despite KMs hatred of moving averages, maybe this is the only real way to incorporate volume. Maybe using a short VWAP such as 1 day, or a lower timeframe such as H4?
KM has said on numerous occasions that the high print in the Vix RR is linked to the low in the SPX RR. As others have found, many times just taking the higher bound number from SPX that day gives the low end of the SPX range, but the same doesn’t work for the high? Maybe the hurst function is used for a multiple here? Or maybe the Hurst is simply used for KMs trend (ie 63 day (3 trading months) hurst for the trend). So if 63day Hurst in .50 or higher we are trending. When it breaks below the .50 line we have a trend change?
Volatility is the key to cracking this. I will re-read Mandelbrots book as I’m sure the clues are there. Maybe we collectively use the Macro Show questions to try and chip away each day to build a knowledge base?
Let’s keep digging!
2 users thanked author for this post.
This morning on The Macro Show one question (around minute 20) was about the fact that the SPX risk range spread increased to 184 from 129 although the VIX was down which he thought would narrow the spread. KM’s response was “…I didn’t change the volatility parameter in the model…”. I was under the impression that volatility changed daily with price, but that is obviously not the case. When I was working on my previous post with the attached Excel spreadsheet, I did notice a couple of occasions when one of the risk range parameters did not change.
2 users thanked author for this post.
Find exclusive trading pro-tools on 
