You crossed who superforecasters are, how to decompose, outside versus inside view, Bayesian practice, dragonfly aggregation, calibration training, daily habits, and why experts fail on leaderboards. Module nine closes the loop that began when mispricing audits asked for a defensible p′: you now have the habits and scoreboards to supply one.
The question is no longer whether superforecasting sounds noble—it is whether you will lock numbers and let reality grade them. The closing question is practical: can you learn this, or is it fixed like height?
Research and market experience converge on a nuanced yes: calibration and updating improve with training and feedback; not everyone reaches elite ranks; markets punish overconfidence as much as undertraining.
Three skill layers
Mechanics—books, courses, drills—are highly learnable. Calibration improves when Brier is visible and bins are reviewed. Market edge is harder: net expected value after fees and liquidity is the trader's exam. Speed under news improves somewhat with templates and tier discipline.
You can learn honest probabilities without beating consensus—that still helps risk control and any reporting role you play.
What Good Judgment–style evidence suggests
Training helps structured teams beat untrained aggregates on many questions. Aggregation—median of independent estimates—helps further. Practice plus feedback moves Brier. Fluid intelligence helps but is not required for solid calibration. Domain expertise helps until it becomes hedgehog overreach. A ceiling remains: top two percent mix trait and practice.
Training does not guarantee positive EV on efficient marquee contracts where price already embeds serious money.
Illustrative learning curve
Untrained individuals might sit near mean Brier 0.23; after a twelve-week course 0.20; trained team median 0.17; market at lock 0.19. Dropping Brier by 0.03 is meaningful—keep journaling. Still losing P&L might mean edge was fees or liquidity, not belief. Beating market Brier over fifty locks is grounds to cautiously scale—not to lever narrative.
On a real bankroll year: Q1 many trades, worse Brier than market, small loss—stop sizing up. Q2 fewer trades, shrink map live, flat P&L. Q3 fewer still, beat market Brier, modest gain. Q4 fewer again, better process, better return. Trade count fell as skill rose—fox passing is growth, not cowardice.
Feedback latency
Learning speed depends on how fast you see outcomes. Tournament questions resolve on known dates; your trades resolve the same way—use that clock. Do not wait for "enough feelings"; wait for scored events.
Methods that work
Deliberate practice on fifty to one hundred scored binaries per year. Immediate feedback when events resolve. Spaced Sunday bin reviews. Ban verbal buckets in the journal—probabilities only. Buddy median monthly. Domain focus on two sectors max for segmented Brier. Daily shadow forecast without money.
Methods that disappoint
Reading without scoring. Twitter conviction loops. Backtesting price paths without locked p. Copying whales on different liquidity. More trades always—fees eat miscalibrated edge. Ignoring resolution text—disputes destroy samples.
Talent versus practice (honest)
High curiosity with average math can reach strong calibration. Quant background without humility dies on ninety-percent buckets. Domain experts need fox habits or hedgehog bleed ruins bins. Part-time traders: shadow plus quarterly size. Full-time: dossier factory plus strict mispricing gates.
Stop rule: after two hundred resolved forecasts, if you are more than 0.03 Brier worse than market and P&L negative, treat consensus as default belief or reduce to passive participation until bins recover.
Twelve-month learning contract sketch
Months 1–2: FORECAST on every trade, full journal compliance. Months 3–4: sixty public locks, compute Brier. Months 5–6: reliability diagram v1, fix worst bucket. Months 7–8: shrink plus mispricing only on finalists, cut trade count ~30%. Months 9–10: beat rolling market Brier over fifty, cautious sizing. Months 11–12: read decentralized module with shadow on-chain forecasts—separate chain risk from event p.
Handoff to blockchain and decentralized markets
Honest event probability informs oracle dispute priors and governance fights—but smart contract, oracle, and gas risk are separate models. Superforecast the world outcome; separately ask will this contract pay.
Module ten adds transparency, on-chain creation, oracle types, dispute layers, settlement without intermediaries, liquidity incentives, scalability, major platforms, and failure modes—none of which replaces calibration; all of which can void a correct event forecast.
Full arc in one breath
Foundations and mechanisms taught what markets are. Probability and contracts taught math and instruments. Platforms and strategies taught frictions and execution. Signals taught sensors and mispricing hunts. Superforecasting taught trustworthy belief. Retail edge, if it exists, lives where hunts find candidates, process validates probability, discipline sizes and executes, and venue choice matches goal.
Separating learnable calibration from efficient markets
You might improve Brier every quarter while marquee contracts stay unprofitable. That is normal if consensus embeds serious money. Learning still matters: smaller sectors, earlier listings, shadow books, and risk control when you do trade. Training is not a promise of riches; it is a promise of honest uncertainty language.
What elite rank might still require
Top tournament tiers may combine practice with traits like curiosity, cognitive reflection, and tolerance for boredom. You do not need elite rank to trade better than you do today. You need better bins than last quarter and fewer unforced errors in resolution reading.
Integration with the full curriculum
Signals without superforecasting is hunting mispricings with unscored beliefs. Superforecasting without signals is academic accuracy without venue awareness. Strategy without calibration is size on ego. Together they describe a professional retail loop: sense consensus, audit mispricing, lock f, size with discipline, score outcomes, shrink biases, repeat.
Closing checklist before you scale size
At least one hundred resolved forecasts logged. Brier within 0.02 of market or better. Worst bin gap under ten points. Rule violations trending down. Positive net EV on audited subset after fees. Healthy pass rate—not zero trades, not every itch.
Missing several? Keep learning; do not lever story.
What comes next
Superforecasting completes the probability side of the curriculum for retail traders. The blockchain module shifts infrastructure: censorship resistance, on-chain event creation, oracles, disputes, and the ways a correct forecast can still lose money if the contract does not pay.
Key ideas to carry forward
Calibration is trainable; elite rank is competitive. Learn Brier before you lever story. Twelve-month contract: journal, public locks, shrink, fewer trades. Module ten adds contract risk on top of event p.
If you remember one sentence from Module nine: lock probabilities, score them, shrink overconfidence, trade only when economics agree. Everything else is elaboration.
Module ten will ask the same discipline about contracts and oracles—different failure modes, same need for honest uncertainty.
Next: Why Blockchain for Prediction Markets? Transparency and Censorship Resistance