Two toy instruments face the same seeded storm. The Middle earns ×1.01 in calm and takes −60% when a storm lands; the Barbell floors most of its wealth at ×1.00, bleeds in calm, and is paid in storms — same ensemble mean by construction, opposite shape.
Both instruments are fed the same per-player storm stream — common random numbers, the fair-comparison design choice: the only thing that differs is the structure. And honestly: B = 9.85 is solved so the two ensemble means coincide — by construction, and only at w = 0.90, p = 0.01. Move either dial and the match breaks, and the page says so.
Terminal wealth after 200 periods, 1,000 players each — the shape difference is the exhibit: the Middle’s skew is negative, the Barbell’s positive.
The game, exactly: each of 1,000 seeded players holds each instrument for 200 periods. The Middle multiplies wealth by ×1.01 in a calm period and by ×0.40 — a −60% period — when a storm lands (chance p per period). The Barbell holds a fraction w at a floor of ×1.00 and the rest in a risk sleeve at ×0.95 calm / ×9.85 storm, rebalanced back to w every period (the formulas are wrong without that phrase): net ×0.995 calm, ×1.885 (+88.5%) storm at w = 0.90. The middle’s per-period ensemble mean is 1.01 × 0.99 + 0.40 × 0.01 = 1.0039 exactly, and B = 9.85 is the exact solution of 0.9 + 0.1(0.95 × 0.99 + 0.01·B) = 1.0039 — so both instruments share the ensemble mean 1.0039 per period, 1.0039200 ≈ 2.178 over the full game, by construction and only at w = 0.90, p = 0.01. Same mean, opposite skew: the middle’s calm is rented against the −60% period; the barbell’s bleed buys a bounded worst case plus the storm payoff — which is why every barbell player who ends above start got there through storms: calm periods only bleed.
The floor is arithmetic in a sense worth being precise about. The −0.5% worst period at w = 0.90 comes from the sleeve’s specified ×0.95; the floor weight alone guarantees — even if the sleeve were wiped to zero in a period — a loss no worse than −(1−w), i.e. −10% at the default. Luck decides when storms land, never how deep a single period can cut. The median texture is exact too: over 200 periods at p = 0.01 the median storm count is exactly 2, and because each instrument’s terminal wealth moves monotonically with the storm count, the two-storm value is the exact median of terminal wealth — middle 1.01198 × 0.402 ≈ 1.1475, barbell 0.995198 × 1.8852 ≈ 1.3170. What that is not is a typical sequence: storm timing varies from player to player, and timing is what shapes each path’s drawdowns.
And the honest cost, stated plainly: with probability 0.99200 ≈ 0.134, a full 200-period run contains no storm at all. Then the barbell decays to its pure bleed, 0.995200 ≈ 0.367 — a −63.3% game — while an untouched middle just compounds its ×1.01 the whole way. Positive skew is paid for in patience: roughly one seeded player in seven waits out the entire game for a storm that never comes.
The structure is the barbell of Nassim Nicholas Taleb’s Antifragile: Things That Gain from Disorder (Random House, 2012), ch. 11, “Never Marry the Rock Star”: two extremes kept deliberately apart — extreme safety on one side, small aggressive bets with open upside on the other, the middle avoided — so the downside is clipped while the good surprises stay live (paraphrased; his prose stays his). Read through this wing’s convexity lens, the barbell is convexity bought for a bounded premium — the fragility mathematics behind that reading lives on that page.