29Nov06 Letter - introducing HotHands

The SweetSpot Investment Letter

November 29, 2006

Subject: SweetSpot plus HotHands

Our back-test taught us a few things, not least of which is that SweetSpot investors have been spoiled. While we can hope never to suffer an extended period of poor performance, we can't expect it. Even the very best investment strategies don't perform well every year, our experience to date notwithstanding. I decided to search for "insurance" in the form of an investment strategy that:

-- consistently beats the market;
-- differs in style from SweetSpot; and
-- requires little time and effort.

Good luck, I thought. Then I remembered an article from last January that purported to offer just what I wanted:

http://tinyurl.com/tl56j

The "Hot-Hands" strategy tells us to buy last year's best-performing exchange-traded fund (ETF) and hold it for one year. I had previously ruled out this approach for the very reason that it seemed contrary to what we were doing with SweetSpot. When I took a closer look, however, there were several things to like about it, the first being its name. Not to be inappropriate, but putting your "sweet spot" together with "hot hands" just seems to be something you would want to do if you could.

Second and third: The strategy is simple and easy. But most impressive are the numbers:

Hot-Hands ETF Strategy 2001-2006:*

Average one-year return: 26.6%
Average one-year market return: 3.9%
Average annual excess return: 22.7%

*Through November 27, 2006.

As promising as those numbers look, they haven't been back-tested. Or rather, they are the back-test. The article's author, Jim Lowell, first suggested Hot-Hands as an ETF strategy in January 2005. ETFs are a fairly new investment vehicle with a limited history, so they can only be tested going back to 2001. Yet there are no meaningful differences between ETFs and sector funds. Why not use sector funds to test the strategy going back to, say, 1989? Mr. Lowell publishes a newsletter called the Fidelity Sector Investor. Wouldn't he be expected to see the value of a real back-test?

Then there's this from Mr. Lowell's description of his methodology:

"I excluded single-sector (Select, Real Estate, Utilities) ETFs [and] the geographically non-diversified single-country) international ETFs."

Huh? He ruled out the ETFs that correspond to most of the funds in our universe and didn't even say why. Still, the article offered an approach (henceforth "HotHands") whose merit I could test. Our SweetSpot strategy looks at annual performance, so I had the necessary data at hand. I input the next-year performance of each year's best-performing Fidelity fund for the period 1989 through year-to-date 2006:

Next-Year Performance of Best-Performing Funds:

Average one-year return: 23.9%
Average one-year market return: 12.9%
Average annual excess return: 11.0%

These numbers compare favorably with our SweetSpot strategy. The sample size here is only 18, however. So how did the second- and third-best performers fare the following year? That's another question that apparently did not occur to Mr. Lowell. (Maybe after excluding so many ETFs he was left with too few to consider picking more than one each year.)

Next-Year Performance of Second-Best-Performing Funds:

Average one-year return: 15.7%
Average one-year market return: 12.9%
Average annual excess return: 2.8%

Next-Year Performance of Third-Best-Performing Funds:

Average one-year return: 18.4%
Average one-year market return: 12.9%
Average annual excess return: 5.5%

And we know that Mr. Lowell didn't wonder about also-rans, of which there were 15 [1]:

Next Year Performance of Also-Rans:

Average one-year return: 34.7%
Average one-year market return: 20.6%
Average annual excess return: 14.2%

By far, the best HotHands performers were... the also-rans, just as they were in our earlier back-test of the SweetSpot strategy. I'm choosing to believe that this is a fluke and that the differences in returns among the groups are due to small sample sizes. I couldn't help noticing, however, that all four groups showed a profit and beat the market. When we combine them, we have 69 trades from 1989 through year-to-date 2006:

Consolidated HotHands Trades:

Average one-year return: 22.7%
Average one-year market return: 14.6%
Average annual excess return: 8.1%

HotHands beat the market by twice as much as SweetSpot! How can that be? We have traded unloved funds for years and done phenomenally well. Are the loved funds even better? The numbers say yes, and when we apply the same criteria we used in our SweetSpot back-test, we have no choice but to incorporate HotHands picks into our investment strategy. Doing so will both lower our risk (by diversifying our investment styles) and improve our expected returns.

I hope you're not taken aback. There have been signs that both loved and unloved sectors perform well. The sectors to avoid, it seems, are those that are neither loved nor unloved; they provoke nothing but indifference. We also should stay away from sectors that are loved "too much" as they can be expected to break our hearts (see Consecutive HotHands Picks, below)...

We used to employ a sell signal that was triggered when a SweetSpot pick became a cash-inflow leader while we held it. Our sell signal failed over time, I believe, because when large cash inflows translate into price gains [2], they can become a magnet for new cash from the many traders who chase performance. Moreover, the gains often continue for years after the "breakout." Hence a three-year holding period for SweetSpot trades (which we enter before the breakout year occurs) and a one-year holding period for HotHands trades (which we enter only after the breakout year is behind us).

Symbiotic Synchronicity

When I combined the HotHands trades with our list of back-tested SweetSpot trades, I verified that there was no commonality between SweetSpot picks and HotHands picks the year they were picked. In fact, it's highly unlikely that a fund would be picked by both SweetSpot (as a cash-outflow leader) and HotHands (as a best performer) in the same year. It never happened during the back-test period.

There is some overlap between the two strategies, however. Fifteen times from 1989-2006, almost once a year, HotHands picked a fund we held the year before as a SweetSpot pick. That makes sense. We buy SweetSpot funds based on prior weakness but their performance after we buy them has been strong. HotHands identifies last year's best performing funds -- when we already own them, we get the benefit of that performance...

A while back I depicted the "sweet spot" graphically as an "X" on the right side of a bell curve, near the bottom where the curve flattens. SweetSpot predicts that the line will begin to climb up the left side of a new curve. HotHands insures us against wrong SweetSpot predictions, identifying investments only after the upward trend is well underway.

As it happens, the relationship is symbiotic: SweetSpot also insures us against wrong HotHands predictions, the "one-year wonders" that we buy only after they have already completed their last (or only) leg up. Ten times in our back-test those funds morphed into SweetSpot trades, giving us three more years to try for market-beating returns. We almost always succeeded.

Nine times we would have held a fund as a HotHands pick that was still a SweetSpot holding. What happens when two winning but disparate strategies both tell us to hold the same investment? In gambling parlance, that's an overlay. As a group, these trades should do better than those picked by only one strategy or the other. They do:

Simultaneous SweetSpot and HotHands Positions:

Average one-year return: 31.5%
Average one-year market return: 14.5%
Average annual excess return: 17.0%

Given these numbers, we will add a second position whenever HotHands identifies a fund that SweetSpot has us holding at the same time.

Six times HotHands picked a fund based on its performance during the third and final year we held it as a SweetSpot pick. Thus it became a buy under one strategy just as we were due to sell it under the other. Now that they were no longer SweetSpot picks, these funds beat the market by about the same margin as the average HotHands pick. Still pretty good, so we will hold onto these funds for one more year, unless they are:

Consecutive HotHands Picks

Three of Mr. Lowell's six ETF trades were repeat picks: small-cap value in 2003 and Latin America in both 2005 and 2006. Latin America gained 50.5% last year, more than the market by tenfold. Year to date, it is again beating the market by a lot. But should it even have been bought after 2004? Remember that when HotHands makes a pick, the fund has already moved one leg up the left side of the curve; a fund picked a second straight year has moved two legs up, and so on... The more legs up, the greater the risk of future losses. The numbers say we are not rewarded for taking that risk: Fourteen repeat Hot-Hands picks trailed the market by an average of 5.3%.

Accordingly, we will not buy consecutive HotHands picks, [3] with one exception. In 1996 electronics was both a repeat HotHands pick and a simultaneous holding (got that?). This will happen when HotHands picks a fund that SweetSpot picked the year before, then picks it again after its second year as a SweetSpot holding. Being consecutively picked in that situation is perfectly fine; we expect SweetSpot holdings to go up year after year. Given the special returns that simultaneous holdings produce, we won't disqualify them from purchase just because they were HotHands picks the year before. Strangely, we will only buy repeat HotHands picks when we already own them as SweetSpot picks. [4]

When we remove consecutive, non-simultaneous HotHands picks, our back-test is left with 55 trades:

Adjusted HotHands Trades:*

Average one-year return: 25.1%
Average one-year market return: 13.6%
Average annual excess return: 11.5%

Combining HotHands with SweetSpot gives us 238 "position-years" for the period 1990-2006 [5]:

SweetSpot plus HotHands Trades:*

Average one-year return: 17.9%%
Average one-year market return: 11.7%
Average annual excess return: 6.2%

*As of November 27, 2006

And these improved returns come with lower risk...

All of this begs the question, however: Why not abandon SweetSpot altogether and just go for the 25% annual returns that HotHands promises? There are lots of reasons why not, and I swear they don't have anything to do with any emotional attachment I might feel toward our sweet, sweet, SweetSpot strategy.

Our improved risk profile comes from combining two strategies. Trading HotHands alone would be high-risk, with only three or so positions each year, all of which have already been bid up when we buy in. Our "turnover rate" would increase from 33% to 100% as we would be selling and replacing all of our holdings every year.

The HotHands numbers are hypothetical, whereas we have been trading SweetSpot in real time since 1998. [6] Real-time results often bear little resemblance to back-tested numbers, in part because successful strategies tend to stop working once they become widely known. As a one-year trade, HotHands is more likely to catch on as a strategy than our three-year, ugly-duckling SweetSpot trades, which seem to fly under the radar...

In summary, we will trade our "new-and-improved" strategy as follows:

-- SweetSpot component: As always, each January we will buy the three funds (plus also-rans) that saw the greatest cash outflows during the year just ended. We will hold these positions for three years.

-- HotHands component: Each year we will buy the three sectors (plus also-rans) that performed the best during the year just ended. We will not buy repeat HotHands picks unless they are also funds we are holding as SweetSpot picks. We will hold HotHands positions for one year.

At any given time our Sweet-Spot/Hot-Hands ratio will be around 3/1. The mix could be adjusted to suit personal preferences, but to optimize our risk profile I would buy all picks if possible.

During the back-test period, we would have held as many as 18 funds at any one time (this year we would be holding 13). That's a lot of funds. Yet we are not required to make a leap of faith that we can hold so many funds and still expect to beat the market. Seeing is believing and these numbers don't lie... they sing!

Cheers,

Neil