Monday, January 14, 2008

IBD Follow Through Days pt. 1 - Are they predictive?

In my last entry I provided a quick overview of IBD Follow Through Days and posed a number of quantitative questions I intend to tackle. This will be the first installment in the series.

Since the market may be on the verge of providing investors with an IBD Follow Though Day I thought it best to tackle the most important question first. Are IBD Follow Through Days predictive of a new bull rally? If so what is the success rate? According to Investors Business Daily, Follow Through Days carry a success rate of between 70%-80%. To test this we need to first define some terms and make some assumptions:

1) What determines a “significant decline”? Declines can be determined many ways. Some may say it would require a series of lowers lows and lower highs. Others might say a certain amount of time should be involved. Others would simply look at a percentage drop to determine significance. I’m going to keep it simple and just look at percent drops. What percent is most appropriate is also arguable. For today’s test I chose 8%. In future studies I will look at multiple levels, so don’t worry if you don’t like my choice. There are two primary and subjective reasons I chose 8%. First, I wanted a number that wasn’t too small that I was testing every minor correction. A 5% drop could just be a few bad days so that seemed too small. Second, I didn’t want a number too large that IBD followers would tell me about all the great rallies my study missed. Since there weren’t any 10% declines in the S&P 500 from March 2003 through 2006, and Investors Business Daily published several Follow Through Day calls over that period of time, 10% seemed too large. I picked somewhere in the middle – 8%.

2) Investors Business Daily originally stated a Follow Through Day should be a rise of at least 1% in one of the major averages accompanied by an increase in volume over the previous day. A few years ago this number was changed to 1.7%. Their explanation was that volatility had increased in the market and a 1% move was no longer as significant. Whatever the reason I decided to test it both ways. As you’ll see, it made no difference.

3) For the test I used the S&P 500 as the “tradeable index”. Determination of success or failure was based on the movement in the S&P 500. I did this because of the three major indices (Dow 30, Nasdaq, and S&P 500), the S&P was the broadest and generally considered the most representative of the overall market. What should be noted, though, is that I allowed a Follow Through Day to be triggered by any of the three above listed major indices, as per William O’Neil’s definition.

4) Success and failure were the most difficult things to define. Here I wanted to be as liberal as possible to give Investors Business Daily the benefit of the doubt. IBD stated that failure could be defined by either “multiple signs of distribution – significant down days in higher volume” or “if one of the major indices undercuts its recent lows”. I was less stringent and said that the S&P 500 specifically would have to CLOSE below the INTRADAY low of the bottom prior to the Follow Through Day. (This decision incidentally made the 08/06/2002 Follow Through Day a success whereas most people would have labeled it a failure – and some the October 2002 bottom a failure.) Until that happened, it still had a chance to succeed. IBD has never offered a clear definition of success, so here I was on my own. I first decided that if you were going to use the Follow Through Day to make money then there should be a good portion of the move remaining. Therefore the target for success was set at twice the distance from the close of the Follow Through Day to the low of the potential bottom day. As hard as bottoms are to pick, I believe tops are even more difficult, so if you lose a third of the move off the bottom, you may also lose a third off the top. Therefore I wanted the meat of the move in the middle (potential reward) to be at least as much as the initial thrust (potential risk). There were a few instances where the market actually made new highs without fulfilling this requirement – so I made things even easier. I  said that any new 200-day high would also signal a “successful” Follow Through Day. This benefited several Follow Though Days. Instances that went from “failure” to “success” include the 08/11/1986 Follow Through Day and the 10/19/1989 Follow Through Day.

I ran the tests back to December of 1971. Since the Nasdaq began trading in 1971 and I wanted to include that as a possible trigger, it made 1971 a reasonable starting year. I needed about 200 bars of data to run some of the calculations and that is why the test only goes back to December of that year. While Investors Business Daily’s research undoubtedly goes back further, 37 years of data is plenty for me. Success or failure prior to that doesn’t concern me greatly.

The Results
I was unable to duplicate IBD’s success rate of 70%-80% even though I made the definitions of “success” and “failure” as liberal as I could.

Using a 1% upmove as the minimum thrust for a Follow Through Day, since December 1971 through January 11, 2008, 35 of 64 possible Follow Through Days were successful for a success rate of 54.7%.

Changing the minimum thrust from1% to1.7% as IBD has done in recent years resulted in 29 of 52 possible Follow Through Days being labeled “successful”. This equals a 55.7% success rate.

Rigging the definition of success to provide the IBD Follow Through Days the benefit of the doubt still didn’t allow me to approach their claims of 70%-80%. In fact the Follow Through Days would have been 50% or less accurate without my beneficial tweaks.

So back to the original question: Are IBD Follow Through Days predictive of a new bull rally? Well, somewhat. A coin flip is not exactly the kind of quantifiable edge I look for unless rewards are substantially higher than risks – which I will address in another post.

Are they as good as advertised? Not any way that I was able to find. Perhaps they’re running their tests differently than I. Unlike them though, I’m willing to back up my claims with some hard evidence (link to trades table).

In the coming days I’ll be answering many more of the questions I posed last night about Follow Through Days. By the time I reach the end of this series you should hopefully have a solid handle on what kind of quantifiable edge they really provide.


Bill Luby said...

Hi Rob,

Some truly superb work in this space. Count me as an early fan. In fact, I like what you've done so much that I have waived my usual 30-60 'trial reading' period and added you to my blogroll today.

Looking forward to more...



Anonymous said...

ditto Bill! (including the 30-60 part ;-)

Dave M. said...

Perhaps you missed some key factors. Did you account for the fact that the follow-through day must be between the 4th and 10th trading day after the bottom day? IBD always says history shows that a follow through day after day 10 is much weaker as an indicator.

Also, what factors did you use to define a bottom day?

Rob Hanna said...

Dave M,

The count was done as follows...

"Here's how it works: The first up day from a bottom in the major indexes is Day 1 of an attempted rally. If the previous low is undercut, the count returns to zero. If the low isn't undercut, the count for Day 2, Day 3, etc. continues. If on Day 4 or later, the market closes up sharply on higher volume than the previous session, then the market has registered a follow-through day."

I only allowed for Follow Through Days to qualify starting Day 4 as per IBD. While IBD claims FTD's after day 10 are less reliable, they were counted since IBD still counts them. One part of my series on FTD's will address specifically FTD's after day 10 and if in fact they are less reliable. (Probably Study #5 as I layed out in the intro piece.)


Pradeep Bonde said...

The follow through day definition.

* Market in downturn.
* First there is a up day which closes higher than previous day. Volume is higher than previous day. That is when you start counting. (Attempted rally day 1)
* From that day 4th to 10th day a 1.7% plus up day on higher volume than previous day and higher than 50 day average volume is confirmed follow through day.
* And during the period between the 1st rally attempt and the follow through, the market should not trade below the first rally day attempt low.

Rob Hanna said...

Pradeep -

Thanks. Your definition matches mine except for two points. First, there is no volume requirement for Day 1. You may refer to the IBD links I included in the post or O'Neil's book for confirmation of this. Second, Day 10 is not an absolute cutoff. I will be examining FTD's after Day 10 in detail in a later post.


Pradeep Bonde said...

If you read his first or second book the 1st attempt day has to be up day with volume higher than previous day and preferably above 50 day average volume is what he says.

Rob Hanna said...


That is not what his book says. From "How To Make Money In Stocks", second edition, copyright 1995, page 59:

"The bottom day in the Dow Jones OR the first strong day up after a major decline is usually the first indication of a possible bottom. A good follow-through...accompanied by an increase in the daily volume from the day before, will usually be on the fourth, fifth, sixth, or seventh day of the attempted rally. This is your second confirmation and main buy signal."

Volume is neccessary on the follow-through day. Not Day 1. Please refer to the links to IBD articles I provided to confirm this further.


Pradeep Bonde said...

See "24 essential lessons for investment success"
Page 76 to 81

Rob Hanna said...

Pradeep -

This will be my last attempt to explain this. Day 1 is the day of the potential bottom. Any time the market rises after it has been downtrending it is possible a bottom is being made. No one is trying to predict a bottom at that point. They are just labeling Day 1 in order to start the count in search of a FTD. I do not have "24 Essential Lessons" handy but I am confident that "How To Make Money In Stocks" along with several articles written by Investors Busines Daily confirming this should be enough to satisfy anyone. I will stand by my study regardless of your interpretation of O'Neil's writing.


Dave M. said...

Thanks for the clarification Rob. I used to read your posts on Nice work. it's nice to see a factual study of a methodology.

Even though these results of the FTD are less than impressive, I still like the IBD method of stock picking since it combines basic company fundamentals and basic technicals, nothing fancy. Growing fundamentals with growing demand for the stock makes the most logical sense to me.

I'd like to see someone tackle the enormous task of verifying the success of a 8+ week or more cup/handle breakout on big volume. I always think that this is a valid and very profitable pattern, but I have to allow that I might have a bias in my observation.

cloudsandskye said...

IBD using a 1.7% gain for the follow through day is a recent criteria - possibly started in 2004. IBD used 2% for a few years - possibly from 2000-2003. The first rally day does not have to close higher than the close of the previous day, but rather can just close in the upper half of the range. A follow through day can occur on day three if days 1-3 all have strong price gains and strong volume. See pages 64-67 of How to Make Money in Stocks (third edition, copyright 2002).

Tim said...

Even with a success rate of approx 50%, no bull market has started without a follow through day. That alone is worth following this indicator. Also, I believe it was Bernard Baruch that said something along the line that a person only needs to be correct 3-4 times out of ten to yield a fortune if they keep the losses small and winners large.

Good work though.

Anonymous said...

Following up on Tim's comment, I'm curious if you are interested in verifying the O'Neil claim that no bull market has ever started without a follow-through day (specifically on days 4 - 10, since obviously a bull market would eventually clock a gain on high volume).


Rob Hanna said...

I just wanted to note that I've needed to make a small change to this study due to some bad volume data I had which was recently noticed. In the original post, the "success" rate with my favorable tweeks was about 52%. It is now 55%. There were a small number of Dow FTD's that didn't originally appear. This has been fixed and 1% FTD's are now 35 of 64 instead of 33 of 63. Without my favorable tweaks, success is still about 50%.


Anonymous said...

I was attempting similar research but stopped due to the subjective nature of FTDs. For example, the 1/31 Nasdaq jump of 1.7% on higher volume did not trigger a FTD due to leading stocks being nowhere near forming constructive bases.

Anonymous said...

I know this is an old post but intriquing nonetheless. I am curious in you study of IBD follow through days. IBD uses them in part to declare a "confirmed rally" after a downturn -found in their "The Big Picture" column. Have you compared your identification of a "Follow-through" day to IBD changeover to a "confirmed rally"?