
Reality Check
The performances in Timing Model are theoretical in the sense that the
model's structure is tested and revised annually (at the end of each year) to
optimize return based on historical data. The revised model is then used
"live" the following year. At
the end of that year the model is revised once again, for use the following
year. This page checks reality by
describing the timing model's live performance, starting with its first
complete year in 1990.
|
|
Inflation |
T-Bills |
Buy & Hold |
Standard |
Aggressive |
|
1990 |
6.1% |
7.5% |
-3.2% |
11.9% |
18.1% |
|
1991 |
3.1% |
5.4% |
30.4% |
30.4% |
40.8% |
|
1992 |
2.9% |
3.5% |
7.6% |
7.6% |
6.5% |
|
1993 |
2.7% |
3.0% |
10.0% |
10.0% |
10.5% |
|
1994 |
2.7% |
4.3% |
1.2% |
-2.0% |
-13.6% |
|
1995 |
2.5% |
5.5% |
37.5% |
25.6% |
20.9% |
|
1996 |
3.3% |
5.0% |
22.9% |
17.4% |
16.3% |
|
1997 |
1.7% |
5.1% |
33.3% |
22.7% |
19.1% |
|
1.5% |
4.9% |
28.6% |
35.5% |
56.2% |
|
|
2.7% |
4.7% |
21.0% |
12.8% |
3.0% |
|
|
2000 |
3.4% |
5.8% |
-9.1% |
-3.8% |
-14.0% |
|
2001 |
1.6% |
3.4% |
-11.9% |
-4.3% |
-9.4% |
|
2002 |
2.4% |
1.6% |
-22.1% |
-7.3% |
-0.3% |
|
2003 |
1.9% |
1.0% |
28.6% |
18.9% |
15.4% |
|
2004 |
3.3% |
1.4% |
10.9% |
10.9% |
13.3% |
|
2005 |
3.4% |
3.2% |
4.9% |
4.9% |
4.2% |
|
2006 |
2.5% |
4.7% |
15.7% |
15.7% |
20.7% |
|
2007 |
4.1% |
4.4% |
5.4% |
5.4% |
4.6% |
|
|
|
|
|
|
|
|
Start |
$100,000 |
$100,000 |
$100,000 |
$100,000 |
$100,000 |
|
End |
166,200 |
206,900 |
604,500 |
673,700 |
608,518 |
|
Return |
2.9% |
4.1% |
10.5% |
11.2% |
10.6% |
|
|
|
|
|
|
|
|
Years to Double |
24.6 |
17.2 |
6.9 |
6.5 |
6.9 |
|
|
|
|
|
|
|
|
Risk |
|
0.0% |
3.8% |
1.8% |
3.0% |
|
Max Drawdown (Yearly) |
|
|
22.1% |
7.3% |
14.0% |
|
Max Drawdown (Weekly) |
|
|
11.6% |
6.0% |
9.0% |
|
Risk-Adjusted Return |
|
|
2.8 |
6.2 |
3.5 |
|
Standard Deviation |
|
|
16.9% |
12.1% |
17.5% |
|
Sharpe Ratio |
|
|
0.38 |
0.58 |
0.37 |
|
Return is total annual percent return, including any reinvested dividends, but not including expenses or taxes. Each account was updated (compounded) on a weekly basis. Risk is average under performance relative to the three-month T-Bill's annual return, a la Morningstar. Max Drawdown is the worst paper loss, based either per year or per week. Risk-Adjusted Return is return divided by risk (return per unit of risk). Standard Deviation is a statistical metric that describes variability, the standard measure of risk in financial analysis. Sharpe Ratio is another measure of risk-adjusted return, defined as excess return over risk-free return (T-bill return in this case) divided by standard deviation. Buy & Hold is buying and holding the S&P 500 Index through thick and thin, including reinvested dividends. Standard (model) is 100% in the S&P 500 during buy signals, including reinvested dividends, and 100% in T-Bills during sell signals. Aggressive is 150% long the S&P 500 (dividends not received) during buy signals and 100% short the S&P 500 (dividends not paid) during sell signals. The model's signals are based on week closings, but returns are calculated on next-day closings when switches are implemented, to more closely approximate the use of mutual funds and conform to reporting standards. |
|||||
The timing model's standard
portfolio returned more than a buy-and-hold strategy since 1990… and with
much less risk. The standard model's
annualized return of 11.2%
would double an investment in about 6.5 years. Its risk of 1.8% is less than half
that of buying and holding. It was also less risky from a drawdown
perspective. On an annual basis, its maximum loss was 7.3% (in 2002) versus
the 22.1% loss in 2002 for buying &
holding. Maximum weekly drawdowns were 6.0% (2003) for the standard model and 11.6% for buying and holding, in the week following
resumption of trading after 9/11/01.
Over the
actual life of the model (since 1990), the standard model versus buy and hold
ends with a larger portfolio, by about $69 thousand (11%), with about half the risk. Its risk-adjusted return, favored by many
analysts, more than doubles that of buy & hold, based on Morningstar
risk. Using the Sharpe ratio, its
risk-adjusted return exceeds buy & hold’s by
54%. Its intermediate to long-term
perspective issued just 25 signals over 18 years, or about three every two
years on average.
The
aggressive model can substantially boost a year's return, as it did in 1998,
but the price paid is higher volatility. The 14% drawdowns in 1994 and 2000 very
much hurt its performance. So far, it falls short of its theoretical
promise of decisively outperforming the standard model… a big
disappointment.
The hallmark of a very good timing model is to match or exceed its benchmark index over an extended period of time, but with much less risk. Sounds good, but it’s the rare money manager or newsletter writer that manages this feat. The standard model has achieved this goal easily in actual use; the aggressive model achieves the goal as well, barely.
The shortage of severe corrections
during the 1990s favored buying and holding strategies over timing systems. The
length of the expansion from October, 1990 to April, 1997 was without precedent
in not experiencing so much as a 10% correction. In April, 1997 we finally got
an 11% correction in the S&P 500, but it lasted all of one day! On a
weekly-closing basis, the correction was 9%, a nonevent from the model's
perspective. The nearly 20% S&P 500 declines in 1990 and 1998, the
49% S&P 500 bear market in 2000-2002, and the Nasdaq's bear markets of 35%
in 1998 and 78% in 2000-2002 all underscore the benefits of capital
preservation strategies. Note that years in which there were substantial
selloffs (1990, 1998, 2000, 2001, and 2002) are the years that the model's
standard portfolio beat the buy and hold strategy.
The breakdown by decades in
the accompanying tables tells the story. The standard strategy under-performed
buy & hold by about 1.5% annualized percentage points per year during the
1990s, ending with a portfolio amount some $6200 less, a cumulative 12%
deficit. Timing strategies simply
can’t beat a disciplined buy & hold strategy during an exceptionally
strong bull market. Adjusted for risk,
the standard model’s return exceeds buy & hold’s return by 100%
during the
1990s, but
this does not translate into real dollars, unless risk scares off the buy &
holder from practicing this tough, emotionless discipline.
And
then came the new millennium. As in the 1970s, cash has been king in
the 2000s, although both the standard and aggressive timing strategies beat
cash invested in 90-day Treasury Bills.
The total gain of an S&P portfolio over 2000-2007 was about 14%
compared to a 44% gain for the standard model.
We now have five successive years of positive returns for
our benchmark index, although it remains barely above water eight years into
the decade. Over the 2000-2007 span, the buy-and-hold strategy averaged an annualized gain of just 1.6%,
well below the cash alternative; the standard and aggressive models showed
annualized gains of
4.7%
and 3.7%,
respectively, with significantly lower risk than buy and hold. Buy and
hold actually marked a negative real return of -1.2%
per year over the current decade, after accounting for a 2.8% annualized
inflation rate. That is to say, the
buying power of a buy-and-hold portfolio is less now than it was at the
beginning of 2000. Cash has been king for the past eight years, models
excepted.
We have a
confirmed cyclical bull market from the bottom in October, 2002. And in May, 2007, that bottom was confirmed
as the end of a short secular bear market that included a tradable cyclical
bull market sandwiched between two cyclical bear markets. This type of intermediate to long-term market
is tailor-made for a good timing system to decisively beat buy-and-hold, as
seen in the second table at right and by the previous secular bear market that
spanned 1966-82, as the Dow flirted with 1000 for 16 bruising years. The
model's genetic heritage gives it a decisive edge over the theoretical buy and
holders during prolonged declines.
Note:
The model's signals use week-end (usually
Friday) closings but the reported live performances are based on next-day
(usually Monday) trading, to more closely approximate the reality of mutual
fund trading and better conform to performance reporting standards.
Live Market Timing PhasesNote:
Each signal date in the table is a
Sunday. Each index is a next-day closing. These are actual (not
theoretical) results. See download page for Excel
workbook that includes time series of S&P 500 and model histories.
|
Timing Model Signal Phases |
||||||
|
Dates |
Signal |
Positions |
DJI |
S&P 500 |
Nasdaq |
Comment |
|
10/29/89 |
Sell |
Cash/Short |
2603-2516 |
335-315 |
451-355 |
Stood aside during 13% S&P 500 rally from February to July, but avoided 20% bear-market decline into October. Overall, dodged 6% loss in S&P 500 and stiff 21% pummeling in Nasdaq. |
|
09/30/90 |
Buy |
Stocks |
2516-3593 |
315-439 |
355-727 |
Missed bear-market bottom by 15 S&P 500 points (on next-day closing basis) and two weeks. Rode primary bull market for about 3 1/2 years. Nasdaq was rocket over this period. |
|
04/03/94 |
Sell |
Cash/Short |
3593-4198 |
439-507 |
727-821 |
Poor sell signal. Up in air for nine months as markets traded within narrow ranges, but missed powerful rally that took averages to new highs in first quarter, 1995. Back to drawing board. Model substantially revised July, 1995, based on data through end 1994. New model would have given timely and LIVE buy signal in mid January, 1995. |
|
04/09/95 |
Buy |
Stocks |
4198-5350 |
507-630 |
821-1060 |
Good buy signal, with solid gains each index. In hindsight, this buy phase should have continued. |
|
07/14/96 |
Sell |
Cash/Short |
5350-5674 |
630-660 |
1060-1121 |
Unsuccessful switch-back signal, lasting only three weeks and giving up gains of 5-6% in indexes. Looked good for two weeks, but strong rally in third week flipped signal back to buy. Current model's technical indicators very sensitive to market's volatility. |
|
08/04/96 |
Buy |
Stocks |
5674-6556 |
660-762 |
1121-1251 |
Profitable, with okay gains all indexes. Ideally, this buy phase should have been continuation of preceding buy phase. |
|
04/06/97 |
Sell |
Cash/Short |
6556-7214 |
762-830 |
1251-1339 |
Another short sell signal with gains regretted. The 9% lost opportunity gain in the S&P 500 is higher than previous 5% worst case for this version of model. Stunning last week accounts for just about all gains, a lesson not lost in next revision of model. |
|
05/04/97 |
Buy |
Stocks |
7214-8787 |
830-1112 |
1339-1851 |
Fifteen-month phase yields good gains all around, although Dow significantly lags behind Nasdaq. In hindsight, this buy phase should have been extension of preceding buy phase. |
|
08/02/98 |
Sell |
Cash/Short |
8787-8466 |
1112-1062 |
1851-1649 |
Eleven-week sell signal modestly successful as 4 to 11% losses are avoided, while nestled in safety of money market during very volatile period that included bear market in Nasdaq. |
|
10/18/98 |
Buy |
Stocks |
8466-10861 |
1062-1348 |
1649-2619 |
Nine-month buy signal very profitable, especially Nasdaq gain. |
|
07/25/99 |
Sell |
Cash/Short |
10861-11008 |
1348-1402 |
2619-4096 |
Did a good job of identifying summer's 12% thirteen-week decline in its early stages, but overlooked subsequent 18% eleven-week reversal to the upside ending December 31. Gave up small gains in the Dow and S&P over its 26-week sell signal, but missed the Nasdaq blow out. |
|
01/23/00 |
Buy |
Stocks |
11008-10941 |
1402-1394 |
4096-3940 |
One-week switchback shows small losses. |
|
01/30/00 |
Sell |
Cash/Short |
10941-10815 |
1394-1468 |
3940-3822 |
Eighteen week sell signal okay for Dow and Nasdaq, but foregone gain for S&P. |
|
06/04/00 |
Buy |
Stocks |
10815-10543 |
1468-1455 |
3822-3912 |
Small losses in big caps, small gain in techs, in volatile markets. |
|
06/25/00 |
Sell |
Cash/Short |
10543-10647 |
1455-1476 |
3912-3980 |
Another short signal gives up small gains as model reacts to tape changes while looking to be on right side of any sustained move. |
|
07/09/00 |
Buy |
Stocks |
10647-10522 |
1476-1430 |
3980-3767 |
Three-week signal yields moderate losses as model gets whipsawed by trading-range behavior. |
|
07/30/00 |
Sell |
Cash/Short |
10522-11080 |
1430-1499 |
3767-3953 |
Model once again on wrong side of three-week trade, as momentum falters in continued trading range. Market reminiscent of 1994 market, but this model more sensitive than earlier version, reacting to head fakes, while likely ensuring early detection of persistent market direction. |
|
08/20/00 |
Buy |
Stocks |
11080-10808 |
1499-1439 |
3953-3741 |
Continued trading range promotes another losing trade. |
|
09/24/00 |
Sell |
Cash/Short |
10808-11338 |
1439-1313 |
3741-2306 |
Sell signal lasted eight months, during which S&P lost 9%. Bear market confirmed during this signal, at 28% below high. Nasdaq plunged 38% during signal. Its bear-market low stands at 68% below high. Dow was star here, actually rising 5%, recovering from near-bear market low of almost 20%. |
|
05/20/01 |
Buy |
Stocks |
11338-10645 |
1313-1208 |
2306-1989 |
Poor buy signal, with significant losses all around. Strong counter-trends, lasting just weeks, give model grief, until clearer intermediate to long-term trends emerge. |
|
06/17/01 |
Sell |
Cash/Short |
10645-9921 |
1208-1140 |
1989-1992 |
Six-month
sell signal moderately successful, although subsequent buy signal missed bottom
of ten-week tradable rally from bear-market lows of Dow 8236 (-30% from
high), S&P 966 (-37%), and Nasdaq 1423 (-72%) following 9/11. |
|
12/09/01 02/10/02 |
Buy |
Stocks |
9921-9885 -1% |
1140-1112 -2% |
||