Aug

12

An unusual consilience of 6 consecutive 20 days maxs in S&P futures has occurred in the last days.

It is interesting to put some stats on table for frequencies of how often such events have occurred in last 17 years.
 

Number of consecutive 20 day extremes                                          

           min                          max                                          

1         204                     313                                          

2          95                     165                                          

3          37                      88                                          

4          17                       50                                         

5           9                      17                                          

6           5                         9                                        

7          0                         4                                          

8          0                         2                                           

9                                    1                                          

10                                   1                                         

11                                   0                                         

In case my formatting didn't come across on your screens, the number of consecutive 20 day minima starting from 1 and going to 8 was 203, 95, 37, 17, 9, 5, 0 , 0

The number of consecutive 20 day maxima starting from 1 and going to 11 was 313,165, 88, 50, 17, 9, 4, 2, 1, 1, 0.

It is interesting to note the falloffs in consecutive maxs from 4 to 5 and the fall offs in consecutive minima from 2 to 3.

Steve Ellison writes: 

If markets were efficient, one would expect a 50% probability that a run of n consecutive highs/lows would become a run of n+1. In this dataset, of 1017 runs of n, 500 (49.2%) became runs of n+1. That doesn't seem far off 50%, but p=.31 by the binomial distribution.

There was an upward bias to the S&P 500 in the past 17 years, so not surprisingly the results are more extreme when split into highs and lows. Only 163 of 367 runs of n consecutive lows became runs of n+1, an apparent p of .02, if one neglects to detrend the data. 337 of 650 runs of n consecutive highs became runs of n+1.
                                                                                  


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search