Skip to content
Final StateWhen Uncertainty Pays You
VOL. I  ·  NODE 002▢  ATLAS

TWO BETS

When Uncertainty Pays You

A field rule for uncertain bets: size the total loss to survive, keep the upside genuinely open, pay the carrying cost, then test the curve over the relevant domain.

JENSEN

The chord is the test; a real bend can lift expected value

Jensen chord diagram for two equally likely outcomes, showing their average payoff above the payoff at their average state on a convex curve.The figure demonstrates the equal-probability case of Jensen's inequality: moving two equally likely outcomes apart creates a gap where the average endpoint payoff exceeds the payoff at the average state.THE MIDDLESPREAD THE OUTCOMESJENSEN GAPAVERAGEOF OUTCOMESOUTCOMEAT AVERAGEJENSEN · 1906SPREAD CAN PAY
Source: J. L. W. V. Jensen, Acta Mathematica 30 (1906), 175-193, doi:10.1007/BF02418571.

This is the actual convexity test over the relevant domain. Capped loss plus open upside neither proves convexity nor is required for it; those terms describe a practical asymmetric profile.

  • Jensen, Acta Mathematica 30 (1906), pages 175-193
  • For convex f: f(lambda x + (1-lambda)y) <= lambda f(x) + (1-lambda)f(y)
  • The gain is strict only where the payoff actually bends

Terms screen the bet; curvature tests convexity; odds price it

Use probabilities where a reliable reference class exists. A loss cap and open upside screen for practical asymmetry, not mathematical convexity. Test curvature separately: over the relevant domain, the payoff at every weighted-average state must not exceed the same weighted average of endpoint payoffs.

THE BARBELL

Safe mass, wild slice, no middle

  1. 01Taleb, Antifragile (2012), chapter 11: keep the bulk low-risk and liquid
  2. 02Cut a slice you can lose entirely
  3. 03Scatter that slice across many open-ended bets
  4. 04Avoid the deceptively stable middle

Source: Taleb, Antifragile (2012), chapter 11, "Never Marry the Rock Star," section "Seneca's Barbell."

Barbell allocation diagram with safe mass, empty middle, and small wild slice.The figure shows the Taleb-style barbell: most resources are kept safe, a small slice is exposed to open-ended bets, and the fragile middle is left empty.BARBELL ALLOCATIONSAFE · EMPTY · WILDSAFE MASSLOW RISK · LIQUIDWILDBETSNOMIDDLESMALL SLICE · OPEN UPSIDE

AFFORDABLE LOSS

Write off the whole stake before you start

Two asymmetric payoff sketches showing an affordable-loss floor holding versus a guarantee leaking through it.The figure shows that the downside cap must include cash, time, guarantees, and reputation; hidden liabilities turn a nominally loss-capped attempt into open downside without determining whether its curve is convex.NO CAPPED DOWNSIDENO CONVEX BETINTACT FLOORFLOOR WITHA HOLEAFFORDABLE LOSSFLOORUPSIDE CAN RUNHIDDENGUARANTEEDOWNSIDE LEAKSCOUNT CASH + TIME + GUARANTEES+ REPUTATIONFLOOR REQUIRED; CURVE TEST SEPARATE
  • Sarasvathy, Academy of Management Review 26(2), 2001, page 252, doi:10.2307/259121
  • Count cash, time, attention, guarantees, and reputation
  • A capacity cap or partner veto cuts the ceiling, not the loss floor

Affordable loss is a sizing rule, not a curvature test. Sarasvathy states the principle on page 252: predetermine what loss is affordable. If total failure leaves you unable to try again, the stake was never affordable.

THE TURN

The curve works in clear weather; the rule earns its keep in fog

Split diagram between risk with a reference class and uncertainty with no reliable class.The figure distinguishes knowable risk from Knightian uncertainty: one side has comparable cases for odds, the other lacks a reliable reference class.RISK HAS ODDSUNCERTAINTY MAY NOTRISKKNOWN CLASSUNCERTAINTYNO CLASSNO CLASS · NO HONEST ODDS
Sources: Black & Scholes, 1973, doi:10.1086/260062; Knight, Risk, Uncertainty and Profit, 1921, Part III, chapter VII. No reliable reference class is not the same as low probability.
  • Black & Scholes, Journal of Political Economy 81(3), 1973, pages 637-654
  • Knight, Risk, Uncertainty and Profit (1921), Part III, chapter VII
  • Known odds: price them. No reliable class: do not invent them

Knightian uncertainty is where no reliable reference class is available. Convexity does not create odds there. Affordable-loss sizing limits what a bad estimate can cost; curvature separately tells you whether dispersion can help.

BRING A VERIFIER

Price the proof into the bet

Sizing ledger that subtracts checking cost and time from gross upside to show collectible upside.The figure treats verification as a holding cost: a gross possible win becomes collectible upside only after the money and time required to prove it are deducted.PRICE THE PROOFINTO THE BETGROSS UPSIDEPOSSIBLECHECKMONEY+ TIMENET UPSIDEAFTER CHECKGROSS - CHECK COST= COLLECTIBLE UPSIDETOO COSTLY? WALK
  • Gross upside minus checking cost equals collectible upside
  • Count verification time against the expiry clock
  • A convex curve can still be a bad trade after checking cost

Verification has its own economics. For sizing, count the full cost of checking before you call the upside open.

Expiry turns attention into a carrying cost

Bowman and Hurry, Academy of Management Review 18(4), 1993, page 763: a shadow option must first be recognised before it can be struck. For sizing, fund the monitoring before the right expires unseen.

WHEN IT PAYS

Survivable, open, and cheap enough to hold

  • Survivable: does total failure leave you whole enough to try again?
  • Open: after capacity limits and vetoes, can the upside still run?
  • Holdable: can you carry enough attempts, checks, and monitoring for long enough?

The three gates identify a practical asymmetric position; they do not prove convexity. Affordable loss sets the size, while the chord/Jensen inequality tests curvature over the relevant domain. Once a position passes both screens, ask where the system must abstain rather than invent odds.

Read the transcript

01 · TWO BETS

A freelance studio has ten free weekends. It can pour all ten into a year-long, fixed-fee retainer. The fee cannot rise, but the revisions can. Or it can spend one weekend on each of ten tiny software tools and keep the right to sell every copy. Nine can die; each costs one weekend. One can keep selling without rebuilding the other nine. Same starting budget, different terms. By winter one position fears surprise. The other has bought room for it.

02 · JENSEN

Across the bottom runs the outcome of the world; up the side, your payoff. The convex curve bends gently on the left, steeply on the right. Take two outcomes with equal probability and pull them apart while keeping their average fixed. For a convex payoff, the average endpoint payoff cannot be lower than the payoff at the average outcome. It is strictly higher only where the curve actually bends. Johan Jensen published the result in 1906. More generally, for any two outcomes and any probability weights, the payoff at the weighted-average outcome cannot exceed the weighted average of their payoffs. That chord inequality, across the relevant domain, is the definition we need.

03 · SHAPE, NOT ODDS

Nobody forecasts a rare win perfectly. The probability may be a guess wearing a number. But odds do not disappear. When a reference class is reliable, use it to price the bet. The terms do a different job. Ask what this can cost, including the price and the waiting, and what ceiling could stop the win. A capped loss and open upside describe a useful asymmetric profile. They are neither sufficient nor necessary for mathematical convexity. To call the exposure convex, run the separate curvature test across the relevant domain: every chord must sit on or above the payoff curve. Then ask whether the price is still good.

04 · THE BARBELL

In chapter eleven of Antifragile, Nassim Taleb gives this posture a portfolio rule and calls it the barbell. Keep the bulk of what you hold low-risk and liquid. Cut a thin slice, only what you can lose outright, and scatter it across many cheap, open-ended bets. Avoid the middle, where a position can look steady while hiding a ruinous break. The heavy end is meant to keep a shock from ending the game. The wild slice gives one good surprise room to matter.

05 · AFFORDABLE LOSS

That thin slice has a name among founders: affordable loss. On page two hundred fifty-two of her 2001 Academy of Management Review article, Saras Sarasvathy contrasts expected returns with an effectual rule: predetermine how much loss is affordable. Ask a smaller, answerable question. If this fails completely, will I be whole enough to try again? Count cash, time, attention, guarantees, and reputation. A hidden liability punches through the floor. A capacity cap or a partner's veto does different damage: it cuts off the upside. Affordable loss makes the downside survivable. It does not establish convexity; curvature remains a separate test.

06 · IT NEEDS THE FOG

Now the turn. The curve does not need fog. Options with known distributions can gain value from volatility too, all else equal; Black and Scholes formalised that pricing in 1973. Fog is where this sizing rule earns its keep. In 1921 Frank Knight separated risk, where probabilities can be calculated or estimated from a sound class, from uncertainty, where no reliable class supports the number. Convexity does not manufacture odds at that edge. Affordable-loss sizing limits what a bad estimate can cost. The separate curvature test tells you whether dispersion can help.

07 · Advertisement · Bubble AI App Builder

Some ideas do not need another document before they become testable. With Bubble AI, you describe the app you want, and Bubble creates a working starting point: interface, data, and logic you can inspect. From there, you refine visually, connect AI models and services, and turn the first version into something real enough to use.

08 · BRING A VERIFIER

Now count the cost of collecting the win. A possible upside is not the same as a collectible one. A proof may check a construction. A lab may measure a binding. A registry may confirm a filing. Each check costs money and time, and the expiry clock keeps moving. Subtract that cost before you value the upside. If proving the result costs more than the result is worth, the curve may be convex and the trade still wrong. Verification has its own economics. The sizing move is simply to put the full check on the bill.

09 · SEEN FIRST

One last item belongs on that bill: attention. On page seven hundred sixty-three of their 1993 article, Edward Bowman and Dileep Hurry describe shadow options that must first be recognised before they can be struck. A right can be cheap to own and still expire before anyone acts. Here the lesson is narrow. Monitoring and exercise are carrying costs. If the position needs a weekly watch, a filing, or a decision before a deadline, fund that work when you size it. An option you cannot afford to monitor is not an affordable option.

10 · WHEN IT PAYS

Go back to the studio's ten weekends. Do not ask only whether one tool could take off. Run three gates. Survivable: if every tool fails, counting hours, guarantees, and reputation, are you whole enough to try again? Open: after capacity limits and vetoes, can the upside still run? Holdable: can you carry enough attempts for long enough, including checking, monitoring, and expiry, without invading the safe mass? Those gates identify a practical asymmetric position. They do not prove convexity. Run the chord test separately across the relevant domain. Where trustworthy odds exist, they still shape the price. The position worth holding is survivable, open, cheap enough to carry, and honest about its curvature. Affordable loss sets its size; the Jensen inequality tests its shape.

01 / 10 · TWO BETS0:00 / 7:19