Broad Scanner

Casting a wide net across all categories to find the best risk-adjusted opportunities

The case for broad scanning

Most trading strategies on prediction markets start with a category focus — weather markets, political markets, or sports. A broad scanning approach deliberately avoids this specialization, instead evaluating every active market against the same criteria and letting the data determine where the best opportunities are.

The logic is straightforward: if a trader only watches weather markets, they will miss a mispriced political contract. If they only watch high-volume markets, they will miss a wide-spread opportunity in a niche category. Breadth ensures the best risk-adjusted trades surface regardless of category.

Diversification in prediction markets

Harry Markowitz's Modern Portfolio Theory (1952) established that diversification reduces portfolio variance without necessarily reducing expected return. The key insight: combining uncorrelated assets produces a better risk-return profile than concentrating in any single asset.

In prediction markets, this translates directly. A portfolio of trades across weather, politics, sports, crypto, and economics is less correlated than a portfolio concentrated in one category. A weather forecast bust doesn't affect political market outcomes; an election surprise doesn't affect temperature brackets.

Correlation within categories

Within-category correlation is often high:

• Weather — Temperature outcomes for cities in the same region are correlated (a cold front affects the entire Northeast)
• Politics — Markets on related political outcomes (e.g., Senate control and individual Senate races) are highly correlated
• Crypto — Bitcoin and Ethereum price brackets tend to move together
• Economics — GDP, unemployment, and inflation markets share macro drivers

A broad scanner naturally mitigates this by distributing trades across categories with low cross-correlation.

Systematic vs. discretionary scanning

A broad scan can be discretionary (manually browsing Kalshi's market list) or systematic (applying quantitative filters programmatically). The systematic approach has several advantages:

• Completeness — Kalshi lists hundreds of active markets at any time. Manual browsing inevitably misses opportunities.
• Consistency — Quantitative filters apply the same criteria to every market, removing the bias toward familiar categories or recent winners.
• Speed — A systematic scanner can evaluate all markets in seconds, while manual review takes minutes per market.

Manski (2006) noted that aggregating information across prediction markets reveals patterns not visible in individual markets — a principle that applies equally to a trader scanning across categories.

Ranking and filtering

A broad scan generates a large candidate set. The challenge is ranking these opportunities effectively. Common ranking dimensions include:

• Expected edge — The estimated difference between the contract price and its true probability. This is the most important factor but also the hardest to estimate accurately.
• Liquidity — Volume, open interest, and spread width. More liquid markets are easier to enter and exit.
• Time to settlement — How soon the market resolves. Faster settlement means faster capital recycling.
• Category — Used for deduplication — limiting exposure to any single category or event family to maintain diversification.

The relative weighting of these dimensions reflects the trader's priorities. A broad scanner typically uses balanced weights rather than heavily favoring any single factor.

Deduplication

One pitfall of broad scanning is over-concentration in a single event. For example, a weather event in Dallas might generate 10 different bracket markets that all look attractive. Without deduplication, a broad scanner might recommend filling the entire portfolio with Dallas temperature brackets — defeating the purpose of diversification.

Effective broad scanners limit the number of positions per event family (e.g., at most 2 brackets from the same event) and per category (e.g., at most 5 weather markets total).

Key considerations

• Edge estimation is harder across categories — A weather specialist may accurately estimate edge in temperature markets but poorly in political markets. Broad scanning requires either general-purpose edge models or acceptance of less precise estimates.
• Information overload — Scanning 500 markets generates noise. Effective filtering and ranking are essential to surface actionable opportunities.
• Shallow domain knowledge — A generalist scanning all categories may have less insight into any individual market than a specialist. The tradeoff is breadth vs. depth.
• Execution complexity — Managing positions across many categories requires tracking different settlement schedules, market structures, and risk characteristics.