ATTENTION, PEOPLE WHO HAVE DISCOVERED A CAN’T-MISS STRATEGY ON A TUESDAY!
PEOPLE WHOSE SPREADSHEET HAS JUST EXPLAINED WHY HISTORY WAS WRONG TO HAPPEN ANY OTHER WAY! PEOPLE ABOUT TO TELL A GROUP CHAT ABOUT A BACKTEST WITH NO LOSING MONTHS!
Are you tired of not having the secret market system?
Have you looked at a chart rising from lower left to upper right and felt the ancient human urge to call the broker, sell the boring thing, and buy a ticket on the MOMENTUM ROCKET DELUXE before lunch?
Then congratulations.
You qualify for the Five Number One Rules of Investing When Markets Are Being Markets — five controls for a long horizon, and no chrome-plated promise that a portfolio will stop behaving like a portfolio.
“Wait,” you’re shouting at the screen, “they can’t all be Number One!”
That’s exactly what Big Numbering wants, while it sells one heroic stock pick as a substitute for an entire investment process.
Investment mistakes are specialists.
Diversification manages co-movement — but a cheap fund can still be concentrated.
Low cost prevents compounding drag — but it can’t restore an allocation that wandered off.
Rebalancing restores exposure — but it won’t tell you whether an estimated edge is real.
Cautious sizing limits the damage from a bad estimate — but it can’t make a backtest with a bad reference class trustworthy.
Concentration, annual drag, drift, estimation error, and noise wearing a small historical costume. Five separate problems.
The market issues no bonus points for having one immaculate principle while the other four are out back licking a battery.
So here they are.
Five rules.
Five Number Ones.
No substitutions. No “but this fund has a great story.” No asking a chart to provide a character reference for itself.
RULE #1: OWN RISKS THAT DO NOT ALL FAIL TOGETHER
Introducing DIVERSI-FLYER 4000™, the portfolio system with so many ticker symbols it can no longer remember which of them sell breakfast cereal.
But quantity was never the product. Different sources of risk are.
Diversification is a curve, not a sacred holding count carved into an index card.
The research says roughly the same thing from several directions: most of the available risk reduction in an all-stock portfolio arrives early, within the first handful of holdings, and the benefit flattens out somewhere in the range of fifteen to thirty names.
Those studies don’t fight to the death in a parking lot. They report different knees in the same curve, using different stock universes, periods and definitions of “most.”
The practical instruction survives all of them intact.
The first holdings matter enormously more than the later ones — provided they aren’t tightly correlated with each other. Past roughly twenty or thirty individually selected names, the next one usually adds more monitoring burden and transaction cost than variance reduction.
Which is why a single low-cost broad-market index fund can hold hundreds or thousands of underlying securities and land you on the flat part of that curve without any of the bookkeeping.
Now the part the sales material omits.
Longin and Solnik found that major international equity correlations tend to rise during high volatility. And during the 2008–2009 recession, broadly diversified multi-asset portfolios still lost roughly 20% to 30% over a few months.
The correlation assumption weakened at precisely the moment the portfolio most wanted it to stay friendly. That isn’t an argument against diversifying. It’s an argument against expecting a guarantee from it.
Diversification does not remove market risk. It refuses to be paid for the same risk twice.
A crowded portfolio is not a diverse one. It’s just crowded.
RULE #1: PAY ATTENTION TO COST BEFORE STORY
And now, from the makers of “THE MANAGER HAS A VERY COMPELLING SLIDE DECK,” comes FEE-OF-THE-MONTH CLUB™.
Every year, a small percentage leaves the account in a tasteful envelope. After several decades, the envelope owns a porch.
Cost isn’t an opinion about whether active managers are good people with excellent neckties.
It’s arithmetic, and the arithmetic is brutal.
Start with $100,000, a 7% gross annual return, and thirty years.
At a 0.05% fee, you compound at 6.95% and finish around $750,600.
At 1.05%, you compound at 5.95% and finish around $566,300.
One percentage point. About $184,300 less — roughly a quarter of the final wealth — and the expensive fund didn’t underperform before fees at all. It simply took a smaller net rate and let compounding do the exhausting work in the other direction.
A cost paid once is an annoyance. A cost taken from every year’s base hires time as its accomplice.
And the performance record doesn’t supply an automatic rescue. SPIVA, which adjusts for survivorship bias, has repeatedly found large majorities of active managers trailing their benchmarks — around 79% of active managers underperforming the S&P 500 in 2021, and only about a quarter beating passive counterparts over the decade ending mid-2021.
Those are named report windows, not a permanent metaphysical verdict on every manager who ever lived. What they establish is the burden: an attractive story has to overcome an annual cost that compounds whether the story delivers or not.
A fee is a forecast of a smaller net return, and it gets to be right every year.
A 0.05% expense ratio requires no exciting story at all. That’s rather the point of it.
RULE #1: REBALANCE BY A STATED RULE
WELCOME TO ALLOC-O-MATIC: NOW WITH LESS “I’LL GET TO IT!”
A portfolio starts with target weights. Markets drag those weights around the room by one ankle. And an intended allocation quietly becomes a before-and-after photograph.
Three methods. Pick one on purpose.
Calendar rebalancing returns to target weights on a fixed schedule — most commonly annually, whether the drift was one percentage point or fifteen. Its virtue is that it’s simple enough to actually happen. Pick a recurring date and act. In a U.S. taxable account, an annual rhythm can also line up with the long-term capital-gains holding period.
Threshold rebalancing trades when an asset class moves outside a stated band, regardless of the calendar. Five percentage points is the commonly cited one, so a 60% equity target calls for action outside roughly 55–65%. It ignores trivial drift and responds faster when a sharp move materially changes portfolio risk — at the cost of requiring monitoring between dates.
Cash flow is the third, and the late-night announcer has somehow forgotten to charge for it. Direct new contributions to the underweight asset class. Draw withdrawals from the overweight one. This corrects in the same direction as a trade without realising a capital gain, and it’s especially useful while contributions are still large relative to the drift.
None of these is a universal winner. Calendar favours simplicity. Threshold favours responsiveness. Cash flow avoids a sale.
The failure mode isn’t picking the wrong holy calendar date. It’s having no stated mechanism at all — at which point every trade becomes a mood with transaction costs attached.
Rebalancing is not a prediction that prices will reverse. It is a promise to restore the risk that was chosen on purpose.
A band you announce after the move is a reaction, not a rule.
RULE #1: SIZE THE BET FOR UNCERTAINTY
BEHOLD KELLY KABOOM™ — the formula that hears “maximum geometric growth” and immediately tries to borrow a megaphone.
The maths is real. So is the uncertainty in the inputs a human feeds it.
Under its assumptions, the Kelly fraction maximises long-run geometric growth. Genuinely. It’s a theorem.
But investing is not a laboratory where expected return and variance arrive with a warranty card. A practitioner estimates both, and an error in either can turn a beautiful fraction into an instruction to take a drawdown personally.
Which is why practitioners who use Kelly-style sizing overwhelmingly use half-Kelly, or quarter-Kelly.
Half-Kelly gives up roughly a quarter of the theoretical maximum growth rate while cutting variance roughly in half. That’s an extremely good trade, and it isn’t a beginner’s diluted version of a braver method — it’s the uncertainty control, used precisely because full Kelly assumes a confidence in inputs that real estimates rarely deserve.
The absurdity detector here is worth keeping. Plug plausible numbers for a passive S&P 500 position into full Kelly and it can return something like 117% of capital — an instruction to use leverage.
That output is not a coupon for leverage. It’s a signal about the distance between a model’s assumptions and any individual investor’s ability to know expected return and variance.
Apply the fraction to a particular position or asset class inside a broader portfolio, not automatically to total net worth. And shrink the multiplier further as the estimates get shakier.
Sizing is where an opinion about an edge meets the price of being wrong about it.
The formula won’t grade its own derivation. It will, however, collect the loss.
RULE #1: DEMAND A REFERENCE CLASS FOR ANY BACKTEST
NOW AVAILABLE: ZERO-LOSS LEGEND™ — the backtest that opens with “nothing went wrong” and desperately hopes nobody asks how many independent chances did it have?
A chart with no losing instances is not a theorem. It’s an event count.
Here’s the useful tool. When something occurs zero times in n independent trials, the rule of three puts a 95% upper bound on the true rate at roughly 3/n.
Run that on a backtest and the spell breaks immediately.
Thirty independent trials with no losses? The true loss rate could still be as high as about 10%.
Ten trials? Around 26%.
Five? You’ve learned essentially nothing, and the approximation itself starts falling apart down there.
Zero doesn’t mean safe. Zero means the remaining uncertainty depends entirely on n.
Which is exactly why n cannot be summoned by slicing one favourable market history into decorative fragments. The rule assumes independent trials.
So a credible backtest names its reference class out loud. Which periods count. What event counts as a loss. Which comparable opportunities were included. And why adjacent or overlapping observations are independent enough for the calculation to mean anything.
Without that, zero losses may simply mean the search procedure went looking for non-losses and succeeded.
This is also why n ≥ 30 isn’t a stamp of statistical adulthood. That convention has no primary derivation that holds regardless of population shape, and heavy skew can require observations in the hundreds.
A familiar threshold is not a substitute for the assumption that gives it meaning.
A backtest earns confidence from its reference class, not from the tidiness of its equity curve.
“What could this have missed?” remains the investment question.
BUT WAIT, THERE’S MORE!
“What if I just own a broad index fund instead of individual stocks?”
Same rules. One fund can hold thousands of securities and its exposures still need examining. Cost still compounds, the allocation still drifts, and any historical claim still needs a reference class.
“What if the account is taxable?”
Same rules. A calendar approach can align with the long-term holding period, and contributions can correct the allocation without selling anything. Tax changes the operating details, not the reason to maintain an allocation by a stated method.
“What if a strategy had no losses across every period tested?”
Same rules. Define the eligible periods, test whether the trials are actually independent, and apply the rule of three. No loss in a selected record does not demonstrate a zero underlying loss rate.
“What if the model says the optimal position uses leverage?”
Same rules. That’s an estimate under model assumptions, not a personal mandate. Reduce the multiplier as the estimates get less certain, because sizing controls estimation error rather than wishing it away.
The details change.
The architecture doesn’t.
THE FIVE, WITHOUT THE EQUITY CURVE
Own risks that don’t all fail together — and remember that correlations rise exactly when you needed them not to.
Pay attention to cost before story. A fee compounds whether or not the story delivers.
Rebalance by a stated rule chosen in advance: calendar, threshold, or cash flow.
Size the bet for uncertainty in your own estimate, not for confidence in your own judgment.
Demand a reference class for any backtest. Zero losses is an event count, not a guarantee.
ACT NOW, BEFORE THE GROUP CHAT DOES
Tonight, open the actual portfolio rather than the idea of it.
List what you own and ask a harder question than “is this diversified?” Ask which of these would fall together in the same bad month. That’s the only version of the question the 2008 data answers.
Then find the expense ratios. All of them. Multiply the difference by thirty years and look directly at the number — it’s the cheapest, most certain improvement available to you, and it requires no forecast about anything.
Write down your rebalancing rule tonight, while nothing has moved. Calendar, threshold, or cash flow. Put the date or the band in writing. A rule invented after a drop is a reaction wearing a rule’s clothing.
If you hold any position sized on a claimed edge, write down what evidence produced that estimate and how many independent observations stand behind it. Then apply a fraction, and make it smaller than feels necessary.
And for any backtest currently doing persuasive work in your thinking, write its reference class: which periods, what counts as a loss, how many genuinely independent chances it had.
If that count is small, you now know what the zero was worth.
For the low, low price of refusing to let a chart vouch for itself, the complete Five Number One system is yours.
No leverage. No secret indicator. No monthly newsletter explaining why the last twelve months were unusual.
Just five controls that refuse to cover for one another.
Operators are no longer standing by.
The operator is the person setting the rules before the market asks for them.