Investing Classics: Portfolio ManagementSizing, Odds, Patience, and the Tax Drag

Buffett's counterintuitive point: concentration is not necessarily riskier, because it forces you to think harder — someone willing to place only five big bets scrutinizes each more carefully than someone scattering fifty. But concentration also amplifies idiosyncratic risk. The key variable isn't concentration itself but whether you actually have an edge: with an edge, concentration is leverage; without one, it just magnifies luck.

The 1964 "salad oil scandal": American Express had guaranteed a subsidiary that had faked inventory, taking a huge loss; the stock halved from about $60 to about $35. Buffett did his own fieldwork — sitting in restaurants and stores — and saw customers still swiping their cards, brand trust intact. Judging the scandal one-off and absorbable, he put about 40% of the partnership (~$13 million) into this single stock; over the following years it multiplied. A textbook case of "bet big once you understand" — with the hard part being that "understanding."

Concentration is poison for those without an edge. Bessembinder (2018), studying all US stocks 1926–2016: about 58% failed to beat Treasury bills over their lifetime, and nearly all net wealth creation came from about 4% of firms. Concentrate at random and you'll likely land on the mediocre or doomed side. Bill Miller beat the S&P for 15 straight years, then nearly blew up in 2008 concentrating in financials — concentration magnifies the outcome; it doesn't improve the quality of your judgment.

• Is this concentrated position based on a verifiable edge, or just "I feel confident"?
• If a single holding went to zero, would my portfolio take a permanent, unrecoverable hit?
• Do my few big positions secretly share one risk factor (same sector, customer, rate sensitivity)?
• If I have no special insight into this business, should I admit I belong to the "should diversify" majority?

The math came from John Kelly (1956): the optimal bet fraction is f* = p − (1−p)/b, where p is the win probability and b the odds (net payoff multiple on a win).

Kelly optimizes geometric growth (compounding), not arithmetic expectation. Investing is multiplicative, not additive: a single −100% wipes out everything before it. So the goal isn't maximizing a single bet's expectancy but the fastest growth subject to never going bust. Bet above the Kelly fraction and ruin risk spikes while long-run growth actually falls. Because true win rates are always estimates, practitioners usually use half-Kelly (½f*): sacrifice a little growth for a large cushion against estimation error.

Ed Thorp carried Kelly from the blackjack table into markets: his Princeton/Newport fund (~1969–1988) sized positions on Kelly logic, returning roughly 19% annualized with almost no down years — not from single windfalls but from the discipline of "size to the edge, never overbet." The mirror image is LTCM in 1998: at roughly 25:1 leverage it pushed convergence trades far beyond any prudent Kelly. The positive expectancy may have been real, but the position was too big; when correlations spiked together that summer, it lost about $4.6 billion within weeks, came to the brink of collapse, and needed a Fed-orchestrated rescue.

Markets aren't a casino: you don't know p or b — you only estimate them; returns have fat tails, and the same "bet" can't be repeated as a trial. Applying Kelly as a precise formula is dangerous — its real value is as a thinking framework: scale size to the edge, and overbetting will kill you.

• Is my "win rate" a grounded estimate, or confidence in disguise?
• Am I using fractional Kelly, leaving a buffer for estimation error?
• If I'm half wrong about the odds, can this position still let me survive?
• Am I using leverage to push the position beyond what any Kelly framework would allow?

Most people chase rallies and fear drops. The correct logic is the reverse — the lower the price, the wider the margin of safety, the heavier the position should be, provided the thesis is unchanged. Soros's partner Druckenmiller put it plainly: "It's not whether you're right or wrong that matters, but how much you make when right and how much you lose when wrong." But this demands you calmly separate a price drop (widening opportunity) from a broken thesis (value trap) — the first calls for adding, the second for leaving, and on the screen they look identical.

At the depths of the 2008–2009 crisis, most capital froze and only wanted to survive. Buffett did the opposite and "reached for the bucket": in September–October 2008 he put $5 billion into Goldman Sachs and $3 billion into GE (both 10%-dividend preferred plus warrants), and in 2009 acquired the BNSF railroad for about $26 billion. While others, gripped by fear, put their thimbles away, he made some of his biggest moves — not the luck of bottom-fishing, but placing size where the margin of safety was widest, not where emotion was most comfortable.

The most dangerous misuse of "buy more as it falls" is adding relentlessly to a deteriorating business — mistaking a value trap for a discount, sinking deeper, finally cutting at the bottom. A margin of safety presumes the value itself hasn't broken; when the drop comes from a damaged moat or a falsified thesis, discipline means trimming, not adding.

• Is this drop market emotion, or has a pillar of my buy thesis genuinely collapsed?
• Before adding, did I independently re-estimate intrinsic value, or just stare at a lower price?
• After adding to this size, if the thesis is ultimately wrong, can I bear the permanent loss?
• Am I betting heavy where the odds are best, or just being dragged along by price swings?

Buffett drove this home with a single number in his 1989 letter (see the table below).

Turnover is a hidden tax on compounding: each realized gain hands a slice of principal to the government, and the loss is magnified by every year of compounding that follows. Unrealized gains, by contrast, are an interest-free loan with no maturity, left whole inside the position to keep rolling — exactly the meaning of the enormous "deferred tax liability" on Berkshire's balance sheet: decades of unrealized gains working for shareholders, like float. Low turnover isn't idleness; it's the rational choice of using that interest-free loan to the fullest.

Don't let the tax tail wag the investment dog. Clinging to an overvalued or deteriorating stock just to avoid tax is a costlier mistake than the tax itself — the tax saved won't cover the collapse in value. Also: deliberately realizing losses in down years (tax-loss harvesting) is a legitimate offset; and in tax-free / tax-deferred accounts this principle weakens sharply, so decisions should return to pure value judgment.

• Do I want to sell because fundamentals changed, or just because my hands itch / I want to "lock in"?
• How much tax would this sale realize, and how many years could that principal have compounded?
• Am I clinging to something I no longer want to own, purely to avoid the tax?
• Is this trade in a taxable or a deferred account? Should the weight of tax be the same?