09/19/2019 at 5:41 AM #107908
I’m a Hedgeye subscriber and they use proprietary ‘risk ranges’ for the main ETFs and Indices. The risk range is calculated across 3 dimensions, namely, price, volatility and volume. The CEO will not give away the secret sauce, but they seem to work well. In a downtrend you wait for XXX ETF to reach the top of the risk range, and then sell; vice versa for an uptrend.
Their levels for the risk ranges yesterday on a couple of securities were:
SP500 – Bullish trend, 2924-3021
VIX – Neutral – 13.36-18.19
Gold – Bullish, 1490-1535
UST10 yield – bearish, 1.41-1.96%
Now, I presumed it was some sort of ATR band, or Bollinger Band, but these levels don’t seem to match. I know they look at the IVOL/RVOL discount and premiums in the options market, so maybe this could be their Vola input.
My question is, does anyone know of an existing indicator that I could use on my charts to indicate similar risk ranges? Could one be built/collaborated on?
Many thanks09/19/2019 at 10:06 AM #10792609/19/2019 at 10:21 AM #107927
Is it this one https://it.tradingview.com/script/nCL4EJSq-eksOr-Risk-Range/?Risk Range12345678910111213141516171819202122232425//@version=2//@author eksOrstudy("eksOr - Risk Range", overlay=true)lookback_length = input(defval = 4, title = "Moving window size", type = integer)offset = input(defval = 0, title = "Offset", type = integer)std_close = stdev(close, lookback_length)upper_bound = highest(high, lookback_length)lower_bound = lowest(low, lookback_length)mid = (upper_bound + lower_bound) / 2if close > midupper_bound = upper_bound + std_closeif close < midlower_bound = lower_bound + std_close// recalcuate midmid := (upper_bound + lower_bound) / 2plot(upper_bound, color=#ce4b1f, title="Upper Range", offset=offset)plot(lower_bound, color=#ce4b1f, title="Lower Range", offset=offset)plot(mid, color=#DFB122, title="Mid Line", offset=offset)09/19/2019 at 3:13 PM #107959
They don’t apply it to the charts sadly. The Risk Ranges are sent out daily; all I know is that they are a derivate of the aformentioned price, volume and volatility. Do we know of an indicator that takes these metrics into account? As far as I know, ATR bands or Bollinger bands are purely 2-demensional, being price and volatility; they exclude the volume element?03/06/2020 at 8:53 PM #12141003/06/2020 at 8:57 PM #121412
That’s the closest think that I come up with.
Because IV is determining the range of the move.
But I don’t know how to incorporate the volume in the formula.
Volume can be incorporated in the trend for example for QQQ is VXN so on tradingview type vxn/QQq or QQq/vxn and you got the trans.05/01/2020 at 3:34 AM #129150
Subscriber here as well. I’ve been playing with the numbers for weeks now trying to see what I can figure out.
I’m starting to think that volume has nothing, (or at least very), little to do with it. In my stats, I’ve found that the correlation between volume and the real price range is less than statistically significant. I think he may simply be keeping an eye on it, but I can’t imagine where this could appear in the math. I think it could possibly be a purposeful misdirection.
I’d love to put our heads together on this if anyone is interested.05/01/2020 at 2:58 PM #129231
Hi guys and thanks for the replies. I have been doing some more digging and it appears that like you say John, the volume might not be that important. A Cambs Uni study on volumes in the Futures market showed that using volume didn’t lead to better edge. I know from listening to Keith on the daily Macro Show, he highlighted the other day that OVX at 198 allowed him to lower his risk range on WTI Oil that day to $7.24. So this clue shows that it must be a function of the volatility (using VIX, OVX, GVZ etc) and price. I think for individual stocks maybe he uses the IVOL/RVOL discount or premium as taken from the options market.
Using the WTI above, the risk range on 29 Apr was 7.24 to 17.31. and OVX was 194 with WTI opening at 12.08. Maybe that’s all we need. I know Keith plugs the values into an Excel doc for his calcs daily. Not sure how we could incorporate into PRT other than manually update the indicator daily?
I know that Keith also says the ‘Machines’ use 30 day vola as that’s all they care about.
1 user thanked author for this post.05/01/2020 at 4:47 PM #129271
I am a Hedgeye subscriber as well and have been trying to work out some risk ranges that are more useful for my purposes since the frequency of Real Time Alerts is too high for my objectives. As to the volume question, I believe it is just a validating factor (price up + volume up = bullish, etc.) and may not be used in the actual risk range calculation. Keith recently mentioned specifically that the risk ranges are based on “rescaled range analysis”. You probably know that Keith is a big fan of fractal analysis and Benoit Mandelbrot’s “The Misbehavior of Markets”. Chapter 9 begins a discussion of Edwin Hurst’s work on the Ashwan High damn in Egypt. I dug into this a little further reading about the Hurst Exponent which is an indicator of persistence (or lack thereof) in a trend. Mandelbrot used this concept in analyzing financial instruments. In short, a rescaled range is range divided by standard deviation which would incorporate both price and volatility, but not volume. All that being said, I have not come up with anything useful yet and am continuing to work on this. I came across this forum doing a Google search and wanted to post what little I know about the topic. I also found this link explaining rescaled ranges and the Hurst Exponent which was very helpful. https://blogs.cfainstitute.org/investor/2013/01/30/rescaled-range-analysis-a-method-for-detecting-persistence-randomness-or-mean-reversion-in-financial-markets/ Any other thoughts would be welcome.05/01/2020 at 6:23 PM #129291
@derrick: I have not heard of rescale range analysis, that seems very useful, I’ll see what I can figure out in that direction. Thanks for that.
@grim: You’re sure he uses Excel? That’s good to know, it at least means it’s not as complex as it might seem. I’ve felt like I was hitting the limits of Excel when looking for clues here, and have been doing more digging using Python’s matplotlib. My stats show a very clear correlation between the VIX and the risk range of the S&P, far more so than using just the implied volatility or the standard deviation.
I assume for assets that don’t have a volatility index, like Tesla or Apple, he must be reverting to the standard deviation. Or maybe he’s replicated the VIX equation and applied it to other assets, but that VIX equation is pretty hardcore math, if he replicated that he’s a lot smarter than he lets on. In case you haven’t seen it, here’s the white paper on the VIX; be warned it’s not easy to follow. https://www.cboe.com/micro/vix/vixwhite.pdf
Let’s keep this going, we can figure this out.05/01/2020 at 6:28 PM #129292
@grim: “I know that Keith also says the ‘Machines’ use 30 day vola as that’s all they care about.”
Also good to know. I’ve been using his trade, trend, tail timeframes, but I’ll see what I can figure out with 30 day timeframes.05/04/2020 at 10:37 PM #129931
KM mentioned in the MacroShow today (concerning Gold) that the risk ranges were a great product esp for gold with GVZ (Gold Vix) down below 30 and the trend being up. There was nothing about volume.
Today Spot gold opened at 1700. GVZ was 24. The risk range was Bullish at 1678 to 1760. So 1700 minus the 24(GVZ) gives us our 1678, but not sure how 24 helps us get to 1760? Maybe as it’s bullish trend he uses Spot open minus the Vola (GVZ) for the low end and maybe a 2 or 3 standard deviation of that GVZ to the upside. Reverse obvs for down trends??
When I have tracked his trend signals before, they have correlated with the last major weekly swing high or low of the charts, and the trend seems to change when price violates these levels.
Maybe more clues?!
Best. Grim05/04/2020 at 10:50 PM #129932
Hmm. Can’t be that simple as WTI spot (June Futs) closed Fri at $18.61. ^OVX closed at 110. Today’s Risk Range was 9.75 to 20.70. The simplicity of the above doesn’t work. Maybe the vola element comes from the IVOL from the options market??05/07/2020 at 11:56 AM #130421
Hedgeye subscriber here. The more I look at this, I get the feeling that its better to start with Donchian channel and apply fractal maths (Hurst Exponent/power law) etc. For example. DC indicator adds subtracts 1 STD to the upper & lower channel depending on price action. We can perhaps add/subtract 1STD along with log(IV) or something to make the numbers come as close to Keith’s range as possible. We can simply follow DC’s +- formula for low volatility bucket, different +- approach for slow volatility and a completely different +- approach for the fuck bucket. A few trail and errors can help; this is not rocket science.
A simple DC with a daily interval of 9 seems to be close to Keith’s Gold daily range. At the end PNL is all that matters. Stick to bull markets, buy at the low end of the range or range breakout and reduce size when the price nibbles at the top end of the range. FYI: Watching the price within the bounds of DC is putting my emotions in check.05/08/2020 at 3:29 PM #130641
Not sure if that was one of you guys who asked that question on the Macro Show this morning but if that was any one of you, hats off to you. Someone asked, “why does the Risk Range model use durations of 3-something *or more*…, so 3 weeks or more, 3months or more, 3 years or more. Why aren’t the trade, trend, tail numbers static?”
So apparently Keith doesn’t use the same ranges each time, sometimes a trend range is more than 90 days, sometimes it’s “119 days” as he said. And he also reiterated that Hurst exponents are an important factor, as @derrick was saying. Hurst exponents are where I’m currently putting most of my focus.
@grim: At least with the S&P, I know there’s definitely a strong correlation between Keith’s numbers and the VIX, but I have yet to figure out exactly where he fits them into the equation. And that’s a good point about oil, that number can be crazy high but oil’s price would never reflect a change in the 200s.
@p2: I have never heard of Donchian channel but I will look into that as well. Thanks for the heads up on that.