Offbeat Ideas and Commentary from the Depths of Gil's Trading Notes
When I ran money for Bill O'Neil, I also managed the Institutional Services Group that was responsible for advising over 700 institutional clients ranging from outfits like Fidelity, Soros Asset Management, Gabelli Asset Management, Strong Funds, Mass Mutual, Tudor Asset Management, etc. One of the services we sold to these institutional clients was our New Stock Market Ideas, known internally by the acronym NSMI and the nickname, "Nizmee."
Institutional clients who wanted these ideas ponied up $150,000 a year to take the service, which was essentially our stock ideas book. Every week clients received a book of charts, known as DataGraphs, containing our best stock ideas along with real-time fax alerts whenever we added a new name to the list. The book was printed on blue paper (to prevent copying) with a daily chart on the left, a weekly chart on the right.
The ideas were generated by Bill and his team of William O'Neil + Co., Inc. (WON) internal portfolio managers, which included some notable names from the organization like yours truly, Chris Kacher, Lee Freestone, Mike Webster (a rookie at the time), and others. WON internal portfolio managers would submit ideas to Bill for inclusion in the service, and if one wanted to buy a particular stock in their firm portfolio they had to convince Bill to put that name on the NSMI service's recommended list. We were only allowed to buy stocks on the NSMI recommended list.
One of my primary tasks when I worked at WON was my weekly Sunday night call with Bill where I would go over new ideas with him. Chris Kacher would call ideas into me before my call with Bill and then I would call Bill. My call with Bill was always scheduled for 7:00 p.m. Pacific, and I did that pretty much every Sunday during my eight years at the firm. We would discuss things for about an hour or so, and then Bill would call Lee Freestone to finalize whatever names we were adding (or even taking off) of the NSMI recommended list.
In August 2004, a little company known as Google (GOOG) had come public using a Dutch Auction method to price the IPO, something that was somewhat unique at the time. Just before the IPO I got a call from a reporter at the San Jose Mercury News, the big mainstream newspaper in Silicon Valley. They wanted me to comment on the GOOG IPO and what I thought about a) the company, b) the unique Dutch Auction format, and c) whether investors should buy it on the first day of trading.
My comments were brief: the company had a compelling business (Google's name itself had already become a verb in the recently emerged internet age), and while the Dutch Auction was interesting we were not really focused on that. We would simply let it start trading and wait for the first base to form and then look to buy it at that time, perhaps. A few days later on August 19th, the stock came public at $85, ran up to a peak of $113.48 in its first three days of trading and then pulled back to form a short U-shaped flag over the next three weeks.
On Sunday night, September 12, 2004 I brought GOOG up to Bill as a new stock idea during our usual Sunday night market pow-wow. I told him this was an IPO U-Turn formation, similar to other patterns we had seen during the dot.com boom in other big internet IPOs like eBay (EBAY), for example, a pattern that was first noticed (we could say "discovered") by two WON institutional sales reps, Mike Lowrey and Steve Greenfield.
This is what GOOG (now GOOGL) looked like back then on the weekly chart. You can see how I've drawn an additional portion of the price bar for the first week of trading as a dotted orange line showing the first-day gap-up from an IPO offering price of $85 to the first print of $100 even when it started trading on August 19th.
When I told Bill about the company's story (he was not that familiar with it at the time) and the stock's IPO U-Turn base pattern he was skeptical, "What's an IPO U-Turn? That's not a base! That's a nothing!" He was not interested - it had to form a "real base," he said. Suddenly, I had a revelation. I would translate it all into terms that Bill could understand, so I pointed out to him that it had come public at $85 and traded up to $113, after which it formed a short three-week base, therefore it was....drum roll please...a High, Tight Flag!!
Now, that was something he could understand - his beloved HTF! He thoughtfully murmured "Hmmmm..." but ultimately ended up rejecting GOOG as a candidate for the NSMI recommended list that Sunday night. In a quick change of heart, two days later he called me up at the office and said to me, "You know that Google you mentioned, why don't you go ahead and add it to the NSMI." Bingo. The stock was now on the list and I could buy it as a WON portfolio manager for the firm account that I ran as well as my own personal account.
I gobbled up shares between $111 and 113 and the rest was history. In 2004 I owned equal amounts of Apple (AAPL) and Google (GOOG) on full margin and ended up 168% by year-end with those two big-stock winners.
Here's the real rub in the whole story, however: GOOG was not a high, tight, flag (HTF) at all, because if it had opened at $85 and traded up to $113 before forming the short IPO U-Turn base, the "flag" portion of my hypothetical HTF, it had only rallied 33% up the "flag pole." Allegedly, at least according to the definition in How to Make Money in Stocks (HTMMIS), an HTF occurs after a rally of 100% or more. This is the definition, verbatim:
“A ‘high, tight flag' price pattern is rare, occurring in no more than a few stocks during a bull market. It begins with the stock moving generally 100 to 120% in a very short period of time (four to eight weeks). It then corrects sideways 10% to 25%, usually in three to four weeks. This is the strongest of patterns, but its also very risky and difficult to interpret.”
Bottom line: GOOG was not a high, tight flag, but by speaking Bill's "language" and using a chart pattern that was part of his well-worn vernacular and thus more acceptable to his mental processes, I got it by him and I got it on the NSMI list, so I could buy it. And all I really wanted was to be able to buy the damn thing, so a little deception in this regard was a worthwhile maneuver that worked.
Imagine my surprise, then when I saw GOOG used as an HTF example in the orange-cover edition of HTMMIS, one of only 15 whopping examples of HTFs going all the way back to 1915. This particular edition of the book was published well after I departed from the firm in 2005. Here's the chart as shown in HTMMIS:
A funny story as I remember it, but perhaps a revealing one. If you wonder why I find the HTF pattern to be a mythical beast, both anecdotally and statistically, consider that despite nearly 100 years of market history, only 15 examples could be found for inclusion in the book, with the final one, GOOG, not even fitting what an HTF is supposed to be. I understand its supposed to be "rare," but when determining whether a certain pattern has any predictive value one needs to have a sufficient sample size to do so, rare or not.
Any statistician worth her salt knows that a minimum sample size should consist of 100 or more observations. 15 certainly doesn't cut it, especially when the last one doesn't even fit the definition of the pattern, and if you examine closely the other 14 HTF examples in the book, you'll notice there are a few others that don't fit the definition either, which you can investigate on your own. Enjoy.