This page has been archived and commenting is disabled.

Did Grinch Steal The Rally: A Peek At The BS Behind "Big Data" Predictions

Tyler Durden's picture




 

Submitted by Salil Mehta of Statistical Ideas

Did Grinch steal the rally?

It was just before Thanksgiving 2014.  It should have been an ebullient time of year.  Markets were already up nearly 17% year-to-date.  Financial market junkies were quick to tell anyone with a wallet, that there would be a year-end rally.  The December signal was a strong one, a number of folks stated, leveraging anything resembling a probability idea to buttress such claims.  Adding to the recent mania, a high-tech era Stock Trader's Almanac -known as Kensho- made its glitzy and cute debut that very few questioned.  However they came out with a very definite and what would be a highly calamitous stock market call.  Instead they quickly compounded and magnified the flaw in such products.  On this blog, several weeks ago, we were properly cautious against such dangers of falling prey to data-mining excesses.  Now we are positioned to conduct a post-mortem on the outcome of following such faulty advice, put forth by those mostly interested in getting you to make (more) specific and ongoing stock trades.  Such excess trading is always to an individual's own financial detriment (it's perhaps amusing in the short run, but saving money over the long run should be more rewarding!)  As I write this on New Year's from Miami's airport's premium lounge, I can see a TSA employee walk past and purchase a couple Lottery tickets from a state-run kiosk at this international airport.  I wonder how much precious earnings of this government employee (a position created as a consequence of 9/11) and that of other colleagues were being routinely dumped into that poor game of chance, further trapping them in such an inelegant position.

Returning to our main topic here, on November 25, the S&P was at its own high level of 2065.  Kensho said, with misleading precision, that the typical return of the S&P (from then to year-end) was +2.2%.  Or a higher market level of 2110.

Using November 25, the amount of days to year-end happens to be just about a week lengthier than that of a typical month (so about 36 calendar days).  To then conform to the odd Kensho style, we looked at the previous 10 years of monthly data, standardizing each non-December "month" period to be the length of about 36 days.  The number of these monthly data collected therefore is high:

10 past years *11 months = 110 historical data

And the typical return over this history was +1%.  Or a market level of 2086 (1% higher than our base-line level 2065).  So for our analysis, breaking above 2086 would be our bogey.  Kensho also provided other frilly statistics to falsely satisfy your concern.  Such as the information that by then, 10 of the past 10 years the S&P was higher during this "Santa Claus rally", though markets as we noted above has an upward tilt, so we might expect our baseline "return" to be higher, greater than 50% of the time to begin with.

Now this is not to be critical of just Kensho, since as we noted earlier that they are just the latest in a long line of assumed cognoscenti to mistakenly clamor confidence in this so-called Santa Claus rally.  In any event, this all turns out to be a sad disarray for anyone who greatly bought into this craze by loading up.  If you lost your way before November 25 and chose then to double-down in order to tighten your relative performance drag, then the Santa Claus rally period did you no favors.  The markets closed the year at 2059, or 6 less than the 2065 we had on November 25.  The data-mined year-end market call also turned out to fall short on a relative basis (i.e., 2059<2086).  Lastly, given the extensive rerun of market panic at the start of December, anyone should easily concede that this trade would not have been a good idea on any risk-adjusted basis.

Let's give an illustration to show some of these statistics from a probability standpoint.  Look at the grey marker visually guiding you in this S&P illustration below.  This marker represents the starting point for where we started on November 25 (the level of 2065).  On the chart below, we also show the percentage change relative to this 2065 level.  In red we then show the 10-year historical distribution of outcomes we would expect for this year-end period length, from a simple buy-and-hold strategy (i.e., ignoring all of the ill-wisdom pundits are prone to throw together).  Of the 110 performance data, we highlight 70 (or 64% of 110) that are greater than the 2059 closing value.

Then see in purple where the Kensho predicted 2110 is.  That was all clearly set-up as false promise!  And finally see in green, where we are now in reality.  Not at all pretty, is it, for those bent on seeking Kensho-style data mined advice?

It is clear from a probability perspective, that the actual returns we just saw in the past half dozen weeks, is a product of just the normal market behavior.  It is not somehow the bad "luck" of a great Santa Claus rally, which Doctor Seuss' Grinch somehow pirated.  To re-express this comment and leverage the illustration above, it is clear that the green marker is much closer to the red marker versus to the purple marker!  But to show this with greater probability rigor, let's see the alternate chart further below.

We portray the Kensho proposed hypothesis that the baseline December rally should have produced a +2.2% return, or an S&P about 2110.  So on this chart we display the instead the percentage change relative to this 2110 level. In purple we then show the 10-year historical distribution of outcomes we would expect if the Santa Claus rally were a real strategy to advise people on.  Of course, in the 10 prior years, we highlight none (0% of 10) that are less than or equal to 2059 closing value.  This closing value, in green, is nearly -2.45% less than the data-mined median value of 2110 (with the right-skewed returns distribution, the average would show an even larger difference).  Finally, if we generously superimpose the broader 110 monthly periods we looked at above, onto the data-minded Santa Claus rally results, the results are the trade recommendation were still improper.  In red below (and using the right vertical axis), we highlight only 24 (22% of 110) that are less than or equal to 2059 closing value.

We use statistics "power tests" to learn the prospect that one hypothesis is better versus another.  We had a 64% likelihood that the rationale we put forth on this blog is correct and how the data-mined style that gave you the Santa Claus rally is wrong.  And we had something between a 0% and a 22% (though closer to the former) likelihood that the unimaginative, Stock Trader's Almanac style logic of investing is laudable.  From this probabilistic criterion, it's not even a close match for which of these two contrary ideas you should well champion.

So now, as we swing from this bad habit of trying to data-mine for great market trading insights, it is worth noting that we enter the New Year now.  Before we go into this wonderful journey, we should really think through some interesting tales from this recent experience, and experiences throughout 2014.  The first point is to start ignoring the 2015 year-end targets, which have come out in the past couple weeks, before this year has actually ended!

Would you ever total a restaurant bill, including gratuity, before you sat down for your meal?  Of course not.  There is uncertainty everywhere, and more so than ever, there is uncertainty with where markets will go over the next year, or so.  Who can ever consistently accurately call the market moves for stocks, bonds, and commodities?  

Another rhetorical question: and would their "charitable" public calls improve, if they made it in late November versus late December, in the year before?  The kookiest of the 2015 commentaries is from Art Cashin, seemingly revered New York Stock Exchange (NYSE) floor manager, who suggests 2015 will be an up-year for the market since years ending in "5" tend to be up years.  It's astonishing that this is the best insight one has, after just celebrating his 50-year anniversary on the Exchange.  But there you have it.  And there he is -on the right of the picture below- ready to dragoon the uneasy Saint Nicolas if he doesn't deliver his goods in the 11th hour.

The second point is that it is also desirable to revisit our blog post, "The zoot who didn't know", created a year ago.  There we note how much of the "correctness" of market calls also comes down to a small amount of risky events that can occur at the start or at end of the year.  This is what we wrote about a year ago, and that's what we then saw occur in 2014.  Not to mention these ideas are more theoretically and empirically grounded.

So with your money, don't listen to or become entrapped by financial advisors offering some great skill, in knowing the guide path for various asset prices, and therefore can guide you in when to buy and sell certain securities.  It's their only job to simply sell as much insane fantasy as possible, to anyone willing to buy.  And it's those same advisors -frenzied by their battalion of technology resources- who are likely uneducated on the lessons contained in the straightforward story about the Grinch, to provide a better light for anyone in their winter voyage ahead.

 

- advertisements -

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
Fri, 01/02/2015 - 09:42 | 5614336 Groundhog Day
Groundhog Day's picture

Isn't all this data and charts etc etc, also just noise cuz no one can really predict asset classes...especially when when their is an invisible hand day in and day out.  seems like a whole lot of work for nothing  whit the blues, greens and purples

Fri, 01/02/2015 - 09:48 | 5614349 SheepDog-One
SheepDog-One's picture

At best, all people could do is attempt to guess what type and flavor of manipulation they'll do to the markets...may as well be Clark W. Griswold playing 'guess which number I'm thinking of' in that dive casino in Vegas Vacation.

Fri, 01/02/2015 - 11:28 | 5614621 Wahooo
Wahooo's picture

Yes. At the end of the day, invest in management.

Fri, 01/02/2015 - 10:27 | 5614343 LawsofPhysics
LawsofPhysics's picture

Where is Tiny Tim and his "arsonists"?  Is arson a good thing now?

Retribution time motherfuckers.

Fri, 01/02/2015 - 09:51 | 5614356 q99x2
q99x2's picture

The only bankers I saw in NYC New Year's Eve that were jumping were jumping for joy. Just my luck. I least I got some photos of seemingly friendly police this time.

Fri, 01/02/2015 - 09:51 | 5614357 tc06rtw
tc06rtw's picture

 WHO is  THE GRINCH ?

Fri, 01/02/2015 - 10:11 | 5614404 SAT 800
SAT 800's picture

I gave up on my S&P500 shorts; as I had the chance to buy them back at zero=zero; this morning. They were as much as 25 pts. out of the money during the recent mania. I'll just watch the fun from the sidelines for a while.

Fri, 01/02/2015 - 10:21 | 5614423 Atomizer
Atomizer's picture

No one ever talks about the capital gains taxation due April 15, 2015.

Fri, 01/02/2015 - 10:29 | 5614445 LawsofPhysics
LawsofPhysics's picture

Gains?  What gains?  Once those gains go into a tax-haven island bank account, they don't exist.  Fire your tax attorney, I don't think he understands how mark to fantasy accounting works.

Fri, 01/02/2015 - 10:38 | 5614470 Osmium
Osmium's picture

Did the Grinch steal the rally?

 

Nope, looks like the "markets" are rallying hard today.

Fri, 01/02/2015 - 11:16 | 5614586 Pwdrhead1711
Pwdrhead1711's picture

Data does not alter belief systems; humans intepret results to reinforce aleady held beliefs and ingnore data that contradicts it...

Fri, 01/02/2015 - 11:36 | 5614643 22winmag
22winmag's picture

Isn't it "The Grinch"?

 

Sort of like "Ukraine" or "The Ukraine".

Fri, 01/02/2015 - 12:40 | 5614967 bid the soldier...
bid the soldiers shoot's picture

Art Cashin is too polite to tell us the real reason. 

Art Cashin, seemingly revered New York Stock Exchange (NYSE) floor manager, who suggests 2015 will be an up-year for the market

He will tell us equally off-the-wall reasons for the markets to go up in 2016, 2017, etc.  

The only reasons for the markets to go down are Doomsday, a Martial coup and rule, and World War III.

Barring those 3 events, the American markets are on their way to 'infinity and beyond'. 

Almost forgot this: if the Fed's printer runs out of ink.

Do NOT follow this link or you will be banned from the site!