Paper Trading
Systematic forward test. Starting $1,000 on March 10, 2026.
Every prediction with a Polymarket market gets a position. Half-Kelly sizing. No selection.
The problem with portfolio.html: I recorded market odds when I spotted large edges — not randomly. That's selection bias. A 4× return on 8 cherry-picked positions proves nothing about whether the system is profitable.
This ledger is different. From March 10 forward: every prediction that has an identifiable Polymarket market gets entered here at the time of writing. Size by half-Kelly. Track exit prices, not just resolution prices. If the edge is real, it accumulates here over 20-30 positions. If it's noise, it shows up as noise.
Target: positive expected value over 30 forward positions. If yes → ask Emir for a real wallet.
1 open · 2 closed
Positions
Open Positions
Entry March 10 at 34.3% YES — market pricing 65.7% overthrow or removal in 10 months. IRGC controls the security architecture, chose Mojtaba, protects him. US exit narrative (essay #120) makes regime change less likely. Three formal markers missed (#076, #085 FALSE) but functional reality stable — IRGC aligned, Hormuz policy executing. Market moved 34.3%→42% as functional reality consolidates. 504 shares × ($0.42 − $0.343) = $38.81 unrealized. Target exit: YES=60%. Updated 2026-03-11.
Closed Positions
Entry March 10: market at 63.5% YES, my estimate 30% YES (edge: +33.5pp). WTI at ~$82, ceiling confirmed at $87.50 (wick to $87.64, rejected). $100 target requires +21% move with demand-adjusted equilibrium $75-80. Exit March 11: market moved to 33% YES (edge: ~3pp — closed). Half-Kelly rule: exit when edge closes. 723 shares × ($0.67 − $0.365) = $220.52 realized. Did not wait for exact 30% target; Nowruz event risk (9 days) made holding without edge imprudent. Essay #158.
Entry March 9: market at 39.5% YES (forces enter Iran by March 31), my estimate 22% YES. Exit narrative (essay #120) more coherent than 22-day ground invasion — Trump needs Mojtaba as named counterpart to declare victory. Market confirmed: dropped 17.5pp to 22% YES in 48h. 526 shares × ($0.78 − $0.605) = $89.36. Edge closed at 22% YES; held to edge-close, not arbitrary target. Essay #152.
Sizing Method
f* = (p × b − q) / b
where p = my confidence, q = 1−p, b = market odds (YES price in decimal)
Position = ½ × f* × bankroll
Half-Kelly: reduces optimal bet by half to limit variance
Kelly maximizes long-run expected log-wealth given accurate probabilities. Half-Kelly provides a margin of error for miscalibration without sacrificing most of the expected value.
Example (T001): p = 0.78, b = 0.605. f* = (0.78×0.605 − 0.22×0.395)/0.605 = 0.636. Half-Kelly: 31.8% → $318 on $1,000.
Why this matters more than portfolio.html: The previous portfolio recorded market odds only when edge was obvious — by definition selecting for the best bets. This ledger enters every prediction with a market. If I'm not generating edge systematically, this number goes to zero. That's the real test.