Wednesday, October 8, 2008

CBI Finally Spikes

The CBI (Capitulative Breadth Indicator) finally spiked above 10 today and hit 12. For those who are unfamiliar with my CBI indicator, it basically uses a proprietary calculation to determine how much capitulation is evident among large-cap stocks. Spikes of 10 or higher in the past have led to market bounces on a fairly consistent basis. Those who would like more info on the CBI may want to read the intro post here or the full post history here. Until Tuesday the CBI had sat relatively dormant. I had thought the market might rebound before the CBI ever hit 10 this time, but this market seems bent on marking every extreme.

Until July, buying the S&P any time the CBI hit 10 or higher and selling it on a return to 3 or lower had a perfect record. The July trade turned out to be a loser. Below are statistics going back to 1995, which is as far as I was able to accurately reconstruct the indicator. The CBI has been tracked live for about 3 years now. All trades assume $100,000 into the S&P 500.

Impressive stats, but it’s important to keep in mind that the current environment is unlike anything we’ve seen since before data exists on this indicator.


Anonymous said...

Any chance you could revisit some T2116 analysis which you did on July 7th? That was a fascinating and instructive read. Many thanks, Steve

Anonymous said...

I respect your work and diligence, but do you really think that a system that gives less than one trade a year is reliable?
There was a reporter in the 70's who accurately predicted the stock markets direction for the year for 17 out of 18 years. His trick was that if the NFL won the super bowl, it was bullish. If the AFL won it was bearish.
I'm sure you of all people know the value of having a large enough sample size - do you think about coincidence vs. causality?

Rob Hanna said...


See today's post for a mention of T2116. There's really no testing to be done since there's only one other instance.


I absolutely believe the CBI has predictive value. Your comparison to the Superbowl indicator is absurd.

When observations are under 30, as they are in the staistics shown for spikes to 10 or higher, you may use a T-table statistic. A T-table may be found in the May 27th post titled "Significance". According to the T-table, 17 for 18 implies above a 99.9% chance that the results are significant.

Of course if readings of 10 were the only ones that worked then the odds of the CBI having significant value would be greatly diminished. That is why I have shown performance at numerous levels in the past. The January 6th, 2008 post "My Capitulative Breadth Indicator" layed out results for readings of 5 and 7 as well as 10. Many more instances occur when looking at these lower values. Feel free to check the T-table against them. They are all greater than 99.9% significant.


Roger said...

CBI at 48!? Can you give us an idea of what that means, if anything at all given that we are not in a normal market?