IBD “Distribution Days” backtesting

So I have been going to the Investor Business Daily meetup in OKC. for the last few months. We usually have about 5 people in attendance and we talk stocks for a couple hours. The host, Raylon Rogers, reads a chapter from an IBD book and then we look at the charts to get our investment ideas.  This last week he talked about “Redemption Days”.  This is an IBD term to look for selling with volume.  IBD is pretty much all a technical analysis method (I perfer fundamentals).  If at the end of one trading day the price action was a lower close and the volume rose from yesterday than this is a distribution day.  Their sell signal is 5 distribution days out of the last 15 trading days.

How often should a redemption day occur?  If it is all truly random, you would expect 3.75 distribution days over any rolling 15 day period (volume can either go up or down, price can either go up or down — therefore a .25 probability that both price moves down and volume moves up on any given day).


So I decided to backtest.  I wrote a script in PERL and am hosting it on github if you would like to download and try yourself. It uses data from alphavantage.co API to get daily stock data.

What I follow, and what we talked about in class was the NASDAQ index.  Using the tool there have been a total of 4,484 trading days since Jan 1, 2000  (where the alphavantage data stops– I would really like to test this for 1997-1999, since I think that will be the best model for how this particular bull market ends).    Of those 4,484 days I can use 4,408 of them for backtesting — I can’t use Jan 1-15, 2000 (since I can’t compute the redemption days between Dec 15-31, 1999), and I can’t use anything after July 26th, 2017 (as I can’t compute the 90 day return after that date (yet)). I use 90 days because that gets a single earning release cycle and I think that is a good number to judge — oh yeah, the market tanked, and you can tell over the last 90 days.

So we are analyzing about ~4,400 days.  In that period for the NASDAQ there were 932 times we had 5 or more redemption days out of the last 15 trading days.   Of those 932 times, 90 days later the index was down 355 times (38%).

OK, seems decent. If I have a formula to keep me protected 38% of the time that might be worth something.  Is it? Let’s see the number of days out of those 4,400 when the return was negative 90 days out:  we get 1,609  (36%).  So in that light 38% is not statistically significant. I can simply use a rule that says always stay out of the market and I will avoid having negative 90 days returns 36% of the time.  Seems about as good as saying stay out when you have 5 redemption days over the last 15 and you will avoid negative returns 38% of the time.

OK, what about total return if you follow redemption day theory? Average return 90 days out from our 932 days of 5/15 redemption days is 1.89%.   This compares to an average return 90 days out from all 4,404 analyzed days of 1.25%    So what this is saying is you are worse than the market (note both returns are positive).  You would be better off buying on 5/15 redemption days.


So how about trying 6 or more redemption days out of 15 instead of just 5/15:

times with 6+ redemption days = 320 
times with 7+ redemption days = 106
times with 8+ redemption days = 14

8 is the most, no time since 2000 has there been 9 or more redemption days out of 15 in the IXIC (Nasdaq).  So analyze those 14 times with 8/15 redemption days. Average 90 day return is highly positive +7.17%: Here are the 14 times it happened and the 90 day forward return:

2000-04-18 = 7.9%
2001-09-24 = 30.8%
2001-09-25 = 31.6%
2004-03-15  = 2.3%
2009-11-06  = 3.1%
2010-06-03  = -3.0%
2010-06-04 = -0.5%
2012-04-18 = -3.5%
2012-04-23 = -3.9%
2012-07-27  = 0.6%
2016-01-22 = 6.5%
2016-01-25 = 7.6%
2016-05-05  = 10.5%
2016-05-06  = 10.3%

Of interest, not only is the 90 day return super positive, you are only saved from -3% moves down three times, and two times you miss 90 day moves of 30%. I definitely want the other side of this trade.  Also, following this theory you are not protected against the 2008 financial crisis or the 2000 dot com bubble.

So, safe to say this theory is totally busted as it relates to the NASDAQ index. How about individual stocks?  Let’s try a few:

distribution days = 960 days (21%)
  average 90 day return = +1.72% 
total days = totals 4,408 days
  average 90 day return = +1.02%
distribution days = 849 days (19%)
  average 90 day return = +1.56% 
total days = totals 4,408 days
  average 90 day return = -0.12%
distribution days = 766 days (17%)
  average 90 day return = +6.7% 
total days = totals 4,408 days
  average 90 day return = +7.08%

note – dividends are considered here for TGT and CSCO, we are using alphavantage.co TIME_SERIES_DAILY_ADJUSTED, not TIME_SERIES_DAILY.

Again, this distribution day theory is busted.  In fact, you want the other side of this trade, if anything (buy after a series of distribution days, don’t sell).


Want me to try it for your stock?  Let me know what symbol in the comments and I’ll post the results here.




College Football thoughts on the eve of the 1st rankings

playoffpredictor.com has as the computer top 4:

Screen Shot 2017-10-29 at 8.29.41 AM

Which really makes sense, if you think about it.  Georgia has a super-quality win over ND. Bama’s best win is over #38 Texas A&M.   Wisconsin looks solid and Penn State’s one loss (1 point on the road) is a much better loss than tOSUs (15 points at home) or OUs (7 points at home).

I am expecting the committee to come out with:

  • Bama #1
  • Georgia #2
  • Wisconsin #3
  • Ohio State or ND  #4

Right now there are 2 other unbeatens – UCF and Miami.   Miami and Notre Dame play each other on Nov 11.  What I am really interested to see is how the committee treats UCF.  My computer has them at #5 – they have some very good wins — in fact, their best win is much better than Bama or Wisconsin’s best win.   However, I suspect the committee will put them at about #20 in the initial poll. The way this season is shaking out UCF could be the only unbeaten in college football.  I will really like to see if a rematch with Memphis and them winning the AAC would be enough to get a mid-major in.