Single-trade skill does not guarantee account survival. Six reasonable edges in the same cluster can lose together on one headline. Portfolio construction decides how many bets, how correlated they are, and what fraction of bankroll faces settlement, dispute, and platform risk at once.
This chapter unifies correlation intuition, Kelly caps, contract shapes, and venue roles into one risk budget.
You can execute passive entries beautifully and still blow up if five entries share one catalyst. Portfolio construction is the account-level answer to a question every earlier tactic ignores: “What happens if I am wrong about the theme, not just the ticket?”
Primitives
Bankroll is capital earmarked for prediction markets only—not rent. Open risk is what you lose if every open position fails. A cluster is events that move together: US election legs, rates complexes, crypto policy themes. A venue slice is exposure per platform after you chose primary and secondary homes.
Each leg remains a binary or tree claim; portfolio math does not remove resolution risk.
Correlation in plain language
Senate control, a tipping state, and presidential odds for the same party are not four independent trades. Treat correlation above roughly 0.5 as one bet for sizing unless you model the joint distribution explicitly. Sports legs in the same league and macro legs tied to the same Fed path behave similarly.
A $30,000 bankroll with four election YES positions near $5,000 each can put three-quarters of the account at risk on one bad night. The fix is cluster caps—perhaps 25% open risk in a theme—with per-leg ceilings near 8%.
Kelly across correlated legs
Two binaries in the same cluster might each imply 12% Kelly in isolation. Summing to 24% overbets. Assign a 12% cluster budget and split by edge weight: the eight-cent edge leg might take half the cluster slice, the four-cent legs a quarter each. Recompute when any probability updates; the portfolio is a living spreadsheet, not a snapshot.
Venue allocation
Most bankroll usually sits on a primary execution venue where depth and rules fit. A secondary slice holds float for arbitrage. Signal-only venues inform without sizing. Experimental listings might cap at a few percent. Regulatory access loss, withdrawal delay, oracle dispute, and illiquidity stack—do not ignore them because edge looks juicy.
Contract-type diversification
Binaries are core. Multi-outcome slates concentrate risk in one narrative. Index contracts can hedge macro. Conditionals carry higher model risk—size down. Bundles help when native combos reduce leg risk. Five binaries that secretly tell one conditional story are still one bet.
Sleeves and caps
Many hobbyists use a core sleeve for passive and Kelly-sized holds, a tactical sleeve for momentum and contrarian with daily loss stops, an arb sleeve with separate float, overlays for hedges on core only, and cash dry powder around fifteen percent uncommitted.
Suggested retail starting caps: 8% single leg, 25% cluster open risk, 85% on one venue, combined tactical exposure on the same event around 10%, arb notional near 30% with buffer, dispute-heavy listings 5% total—halve while learning.
Time horizon mix
Short-dated catalysts can dominate open risk even at small size if five resolve the same night. Spread resolves across weeks: maybe half of open risk in the one-to-three-month bucket, thirty percent inside a week, twenty percent long-dated—with lower caps on dispute-heavy long tails.
Stress the portfolio: if every US-election leg loses, is a 25% bankroll drawdown acceptable? If not, cut before adding.
Common mistakes
Ticket count mistaken for diversification, ignoring venue caps, full Kelly on each leg, netting arb plus directional on the same headline without a chart, all-long YES in one macro story, and treating play-money win rate as permission to supersize USD.
Building the book step by step
Start from cash. List open legs with cluster tags and max loss dollars. Sum cluster exposure; compare to caps. Only then add a new ticket. If the new ticket breaches cluster cap, hedge, shrink, or skip—do not “just make it smaller later.” Weekly review compares open risk to bankroll, not account balance highs.
Arb sleeve isolation
Keep arb float separate mentally. Directional heat and arb heat should not share the same emergency buffer. When one arb leg sticks, the recovery trade is operational, not a reason to double directional size elsewhere.
Expected value aggregation caution
Portfolio EV is not the sum of leg EVs when legs correlate. Stress all-cluster lose scenarios in dollars, not in percent hit rate. If one bad night can end the account, the portfolio is misbuilt regardless of per-leg brilliance.
Cash as a position
Dry powder is not idle failure—it is capacity to enter when edge appears after a crash. Books that are fully deployed into one cluster have no flexibility when contrarian opportunities arrive.
Election night rehearsal
Imagine every open US-election leg losing at once. Sum premium at risk. If that number exceeds twenty-five percent of bankroll, you are not diversified—you are speculating on one night. Cut or hedge before the night, not during the cascade.
Key ideas
Clusters beat ticket counts. Kelly applies to budgets across correlated legs. Venue, contract shape, and catalyst calendar are risk dimensions beside probability.
What comes next
You know how much to hold; the next chapter decides how you get filled.
Tactical sleeve interaction
Momentum and contrarian tickets should share a tactical budget inside the cluster cap. If tactical spend is exhausted, sit out the headline—even when FOMO is loud. Portfolio construction is the enforcement layer for bias and sizing rules.
Misconceptions
“More tickets equals diversification” is the deadliest. “Arb is risk-free so ignore cluster” nets arb with directional accidentally. “I will hedge later” without capacity is hope. Portfolio work is boring and prevents exciting blowups.
Treat open risk as a single number beside bankroll every Sunday.
Next: Using Limit Orders vs. Market Orders