Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

A Statistical Approach To Timing Entries And Exits In Index Futures.

|Includes: SPDR S&P 500 Trust ETF (SPY)
Summary

High Low Ranges can provide a great tool for timing trades.

The approach is applicable to both stocks and idexes.

Implied volatility is a great lagged range predictor.

Assume you are bullish or bearish a certain index or stock, When do you get in? My own approach combines two methods, one is based on deriving information from the positioning in the relevant options market and the second is a statistical approach that I will describe and provide the xls. file for in this article.

For simplicity’s sake let’s assume that you are bullish on the S&P 500 index, you would ideally want to get inside after a significant drawdown. We will thus attempt to quantify what a significant drawdown looks like. Then we will sort them by their percentiles in order to understand what a median, 75% or 99% drawdown looks like on the S&P500. This is accomplished easily by downloading Open/Close data and calculating consecutive declines and advances. We end up with the following table for each index (eg for the NASDAQ100).

Table of median, 75%, 95% and 99% declines/advances and ranges: NASDAQ 100

Now the remaining issue is how do you know if you need to wait for a median drawdown or a 99% one? Two possible solutions: you just stay cash waiting for a super opportunity that would be best represented by a 95% percentile decline or 99: Why not? those who did that this year have been greatly rewarded for example.

Description: C:\Users\Saad\OneDrive\Images\Captures d’écran\2018-06-14 (2).png

In Jan the NDX had an advance of around 7 to 8% which corresponds to a little bit more than a 99% advance in my table – if you were in the camp that such an advance is unsustainable and needs to be counterbalanced by a drawdown then this statistical table would have placed your short quite comfortably. However 99%-ile events are by definition rare and not everyone has the patience to wait for such an event and then go all in. The second solution is to use your own experience and judgment depending on what is happening in the market to see whether a median DD is warranted or more but this takes time to develop and your intuition has to be well calibrated to what you are trading.

Another statistical method I employ but that does not rely on intuition is predicting next day’s range using a lagged variable. I tested various potential predictors such as the previous days’ volume,30d volatility average, different measures of implied volatility, implied vol of implied vol etc..

I have had the best results with an index’s implied volatility (at close of previous day) as many of the variables I have cited are heavily correlated (which makes sense, the way the vix is calculated makes it a linear combination of moments and not just the volatility)

REGRESSION

/MISSING LISTWISE

/STATISTICS COEFF OUTS R ANOVA

/CRITERIA=PIN(.05) POUT(.10)

/NOORIGIN

/DEPENDENT RangeSPX

/METHOD=ENTER VIX.

Regression

Variables Entered/Removeda

Model

Variables Entered

Variables Removed

Method

1

VIXb

.

Enter

a. Dependent Variable: Range SPX

b. All requested variables entered.

Model Summary

Model

R

R Square

Adjusted R Square

Std. Error of the Estimate

1

,792a

,628

,628

66,358541160000000

a. Predictors: (Constant), VIX

ANOVAa

Model

Sum of Squares

df

Mean Square

F

Sig.

1

Regression

22166145,790

1

22166145,790

5033,807

,000b

Residual

13131105,750

2982

4403,456

Total

35297251,530

2983

a. Dependent Variable: Range SPX

b. Predictors: (Constant), VIX

Coefficientsa

Model

Unstandardized Coefficients

Standardized Coefficients

t

Sig.

B

Std. Error

Beta

1

(Constant)

-49,993

2,791

-17,911

,000

VIX

8,998

,127

,792

70,949

,000

a. Dependent Variable: Range SPX

Similar results are obtained when using other indexes with their implied volatility.

Result for NKY

When testing for which spot in the term curve of implied volatility gives you the best R^2 it appears to be the 10d implied volatility and hence this is what the model will be using. With this variable the SP500 regression reaches 69% R^2. Needless to say these models do not explain all of the variations in ranges and you can still have some statistically outrageous day such as when the NADAQ100 fell 4% because NVDIA said it would back down from research into autonomous vehicles but I have found them to be a very useful guide since I started trading, Especially if there is no significant event.

Link for excel INDEX RANGES.xlsx

You can use the list button in the index’s names to switch the index and then you have to input the day-1 implied vol level. Not all indexes are supported at this time but this file is regularly updated and already has all the big ones.

Description: C:\Users\Saad\OneDrive\Images\Captures d’écran\2018-06-17 (2).png

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.