A Synthesis of Academic Papers on Market Efficiency, Behavioral Biases, and Exploitable Edges

Executive Summary

This paper synthesizes actionable trading strategies from 20 academic studies on prediction markets, betting market efficiency, and behavioral finance. The research spans work from 2006 to 2026 and covers election prediction markets, sports betting exchanges, parimutuel systems, experimental market designs, on-chain Polymarket orderbook data, and Kalshi NFL microstructure.

Ten distinct, empirically-grounded strategies emerge from the literature. Each exploits a documented market inefficiency rooted in either structural features of prediction market design or systematic behavioral biases among participants. The strategies range from simple contrarian approaches (betting against longshots) to more sophisticated multi-market arbitrage, information-flow techniques, on-chain combinatorial arbitrage on Polymarket, passive liquidity underwriting on Kalshi, and dynamic hedging strategies that lock in profit from odds movement without taking outcome risk.

The strongest edges are found in five areas: the favourite-longshot bias at contract price extremes, mispriced hedging instruments, informed-participant flow signals near event resolution, intra-market and combinatorial arbitrage on Polymarket (which has historically paid out approximately $40M to sophisticated participants), and passive liquidity provision on Kalshi (effectively underwriting rather than market-making, with aggregate positive returns across an NFL season in the $29M range).

Strategy 1: Favourite-Longshot Bias

Source Papers: Buhagiar, Cortis & Newall (2018); Ottaviani & Sorensen (2009); Franck, Verbeek & Nuesch (2010)

Expected Edge: 2-5% over market average per contract

Confidence: Very High (replicated across dozens of studies and markets)

The Inefficiency

The favourite-longshot bias is the single most robust finding in prediction market research. Contracts on low-probability outcomes (longshots) are systematically overpriced, while contracts on high-probability outcomes (favourites) are systematically underpriced. Buhagiar et al. confirmed this using data on thousands of soccer matches: bettors who placed longshot bets lost significantly more money than those who bet on favourites, even after accounting for bookmaker margins.

The bias arises because participants overweight the attractiveness of large potential payoffs from unlikely events. A contract at $0.05 (5% implied probability) feels like a lottery ticket, and demand from entertainment-seeking participants pushes the price above fair value. Conversely, contracts at $0.85 or $0.90 feel unappealing because the payoff-to-risk ratio is low, leaving them underpriced.

How to Trade It

• Systematic favourite betting: Focus on contracts priced above $0.50 (>50% implied probability). These are statistically underpriced relative to their true probability across virtually all prediction market domains.

• Longshot fading via "Buy No": Most prediction market platforms, including Kalshi, do not allow traditional short selling. Instead, they offer "No" contracts---a separate instrument that pays out $1.00 if the event does not occur. When a Yes contract is overpriced at $0.15 (implying 15% probability for something that is truly a 10% event), you do not need to short the Yes. Instead, buy the corresponding No contract. On Kalshi, if Yes is at $0.15, No is typically priced around $0.85. If the event does not occur, your No contract pays $1.00, netting you $0.15 per contract. This is economically equivalent to shorting the Yes, but uses a long-only mechanic. Buying No on overpriced longshots is the single highest-edge strategy in the literature.

• Sweet spot: The bias is most exploitable when Yes contracts are in the $0.05--$0.15 range (buy the corresponding No at $0.85--$0.95) and in the $0.75--$0.92 range (buy Yes directly). At true extremes (below $0.03 or above $0.97), liquidity dries up and spreads consume the edge.

Platform Mechanics: Trading on Kalshi

Kalshi is a CFTC-regulated prediction market where every event has two contracts: Yes and No. Unlike traditional financial markets, Kalshi does not support short selling. Instead, every bearish thesis is expressed by buying a No contract. This is a critical distinction, because the favourite-longshot bias literature often describes the edge as "selling overpriced longshots," which sounds like it requires shorting. On Kalshi, you achieve the same economic exposure by buying No.

On Kalshi, contracts are priced in cents from $0.01 to $0.99. Yes and No prices for the same event should sum to approximately $1.00 (the small difference is the platform spread). Your maximum loss on any position is your purchase price, and your maximum gain is $1.00 minus your purchase price. There are no margin calls, no borrowing costs, and no risk of forced liquidation---your downside is always capped at what you paid.

Worked Example: Fading a Longshot on Kalshi

Suppose Kalshi lists a market: "Will the Fed cut rates by 75bps at the June meeting?" The Yes contract is trading at $0.12 and the No contract at $0.88. You believe the true probability of a 75bp cut is around 5%, meaning Yes is overpriced at 12 cents. You cannot short the Yes contract. Instead, you buy 100 No contracts at $0.88 each, spending $88. If the Fed does not cut by 75bps (which you believe is 95% likely), each No contract pays $1.00. You receive $100 and profit $12, a 13.6% return. If the Fed does cut by 75bps, your No contracts expire worthless and you lose the $88. Your expected value is (0.95 × $12) − (0.05 × $88) = $11.40 − $4.40 = +$7.00 expected profit per 100 contracts, a +8.0% expected return on capital deployed.

Worked Example: Buying the Favourite on Kalshi

Now consider: "Will GDP growth exceed 1% in Q2?" Yes is at $0.78 and No is at $0.22. You believe the true probability is closer to 85%. You buy 100 Yes contracts at $0.78, spending $78. If GDP exceeds 1%, you receive $100 and profit $22 (28.2% return). Expected value: (0.85 × $22) − (0.15 × $78) = $18.70 − $11.70 = +$7.00, a +9.0% expected return. Notice this is the other side of the favourite-longshot bias: the favourite (Yes at $0.78) is underpriced relative to its true 85% probability.

Key Differences from Polymarket

Polymarket uses an order-book model where Yes and No shares trade independently and can sometimes be sold back to the market before resolution. On Polymarket, you can sell a Yes position you hold (effectively going flat) but you cannot open a naked short. The practical result is similar to Kalshi: to bet against an outcome, buy No shares. However, Polymarket does not have CFTC regulation, contract pricing conventions differ (shares rather than cents), and fee structures vary. On both platforms, the favourite-longshot bias is exploited the same way: buy No on overpriced longshots, buy Yes on underpriced favourites.

Where It Works Best

The bias is strongest in markets with many casual participants (sports, entertainment, novelty markets) and weakest in markets dominated by sophisticated traders (financial derivatives, well-followed political races). Franck et al. found that betting exchanges (closer to prediction markets) showed less bias than bookmaker markets, suggesting the bias narrows with more informed participation but never fully disappears. On Kalshi specifically, economic and policy markets (Fed decisions, GDP, inflation) tend to have a mix of casual and sophisticated participants, making them productive hunting grounds for this strategy.

Strategy 2: Hedge Underpricing

Source Papers: Frederick et al. (2018); Chatterjee & Mookherjee (2018); Newall & Cortis (2019)

Expected Edge: 3-15% on paired positions

Confidence: High (consistent across experimental and field data)

The Inefficiency

Three independent research groups found that market participants systematically undervalue hedging instruments. Chatterjee and Mookherjee ran experiments where participants could buy a risky bet (e.g., win $10 if a coin lands heads) and a hedge (win $10 if tails). A rational agent would value the pair at $10, since it guarantees a payout. Participants consistently valued the pair at only $5-6, underpricing the hedge by 40-50%.

Frederick et al. confirmed this with a different methodology, finding that the risk attitudes implied by bet valuations are fundamentally inconsistent with those implied by hedge valuations. People use the price of the risky bet as an anchor for valuing the hedge, rather than pricing it independently.

Newall and Cortis applied this to real-world high-stakes betting. Using the famous Leicester City Premier League case, they showed that hedging would have yielded 9-15% annualized returns, yet the vast majority of bettors in similar positions fail to hedge.

How to Trade It

• Complement arbitrage: On platforms where YES and NO contracts trade independently (like Polymarket), check whether YES + NO prices sum to less than $1.00. When they do, buying both sides locks in a risk-free profit equal to the difference. This is the direct prediction market analog of hedge underpricing.

• Cross-platform hedging: If you hold a position on one platform, check the complementary contract price on another. Hedge underpricing means the hedge is often cheaper than it should be, making cross-platform hedging more attractive than most participants realize.

• Accumulate underpriced hedges: When a market moves significantly after you enter a position, the hedge contract becomes relatively cheap. Buy it to lock in guaranteed returns. Per Newall and Cortis, these hedged returns can reach 9-15% annualized.

Where It Works Best

This strategy works best on platforms with independent YES/NO order books, during periods of high volatility (when hedges get cheapest), and in markets where participants have strong directional conviction (making them least interested in hedging).

Strategy 3: Informed Trader Flow Signals

Source Papers: Ottaviani & Sorensen (2009); Flepp, Nuesch & Franck (2017)

Expected Edge: 2-5% following the direction of late informed flow

Confidence: High

The Inefficiency

Ottaviani and Sorensen demonstrated that in parimutuel markets, informed traders strategically time their bets to minimize information leakage. They place large positions as close to the deadline as possible, creating sharp probability shifts in the final minutes before an event closes. These last-minute movements are driven by participants with superior information and are strongly predictive of outcomes.

Flepp et al. confirmed a related finding: market microstructure matters. Quote-driven markets (where a market maker guarantees prices) process information differently than order-driven markets (where participants trade against each other). Informed traders preferentially enter order-driven markets where their superior information is most valuable, making flow signals in these venues most reliable.

How to Trade It

• Monitor final-hour flow: Track contract prices and volume in the last 30-60 minutes before event resolution. A sharp movement of more than 5 percentage points in this window is highly likely to reflect informed participants.

• Follow the direction: If a YES contract rises sharply near the deadline, buy YES. If it falls sharply, sell or buy NO. The direction of late money is the most reliable short-term signal in prediction markets.

• Volume confirmation: Price movements accompanied by above-average volume are more significant than price movements on thin volume. The combination of price shift + volume surge is the strongest informed-trader signal.

• Avoid thin markets: This strategy requires sufficient liquidity to enter and exit. In markets with very few participants, late movements may be noise rather than signal.

Where It Works Best

Most effective in markets with clear deadlines (election night, sporting events, binary economic data releases) and sufficient liquidity for late trading. Less effective in markets that resolve gradually or have ambiguous resolution criteria.

Strategy 4: Multi-Platform Arbitrage

Source Papers: Cortis (2015); Axen & Cortis (2019, 2020); Peel (2017)

Expected Edge: 1-5% per arbitrage opportunity

Confidence: Very High (mathematically guaranteed when conditions are met)

The Inefficiency

Cortis demonstrated that bookmaker profitability depends on the overround: implied probabilities across all outcomes summing to more than 100%. Axen and Cortis extended this by identifying no-arbitrage and no-dominance conditions that rational markets must satisfy. When these conditions are violated, risk-free profits are available.

In prediction markets, the overround manifests as the spread between YES and NO prices. If YES trades at $0.55 and NO at $0.50 on the same platform, the $0.05 gap is the platform margin. But across platforms, the same event may have YES at $0.52 on Platform A and NO at $0.44 on Platform B, creating a $0.04 arbitrage (buy both for $0.96, guaranteed to receive $1.00).

Peel showed that bettors actually exhibit a preference for multi-outcome wagering (dutching), and that this behavior is consistent with Rank Dependent Utility preferences. This means the demand for hedged positions can itself create mispricing in correlated contracts.

How to Trade It

• Cross-platform price scanning: Monitor the same event across Polymarket, Kalshi, PredictIt, Metaculus, and any other active platforms. When the sum of the cheapest YES across one platform and the cheapest NO across another falls below $1.00 (minus fees), execute both sides.

• Overround monitoring: Calculate the total implied probability across all mutually exclusive outcomes on a single platform. An overround above 5% signals heavy platform margin; look for the same event on a lower-overround platform.

• Back-lay spread exploitation: On exchange-style platforms with back (buy) and lay (sell) prices, identify when the back-lay spread on one contract is tighter than on its complement. This asymmetry often indicates directional mispricing.

• Dutching: In multi-outcome markets (e.g., who will win an election with 5+ candidates), construct a portfolio of positions that guarantees a positive return regardless of outcome. This requires all contracts to be available at prices where the sum of implied probabilities is below 100%.

Fee Considerations

Axen and Cortis (2020) specifically addressed the impact of exchange fees on hedging profitability. With a typical 5% fee, odds must shift by at least 15% from your entry price to create a profitable hedge after fees. The key formula: profitable hedging requires (entry odds - 1) / (current hedge odds - 1) > 1 / (1 - fee rate) squared.

Strategy 5: Long-Horizon Forecasting Edge

Source Papers: Berg, Nelson & Rietz (2008); Dreber et al. (2015)

Expected Edge: 5-15% forecast error reduction vs. polls and expert opinion

Confidence: High

The Inefficiency

Berg et al. analyzed prediction markets against 964 election polls across five election cycles. Prediction markets outperformed polls 74% of the time overall, but the advantage was most pronounced at long horizons: markets significantly outperformed polls in forecasts made 100 or more days before elections.

Dreber et al. extended this to science: prediction markets correctly identified 93% of published studies that would fail replication attempts, with accurate signals emerging 1-2 months before actual replication results were announced. Market prices converged to approximately 55% for studies that would replicate and significantly lower for those that would not.

How to Trade It

• Early positioning: For events 60+ days out, prediction market prices are more accurate than polls, expert surveys, or structural models. Enter positions early when you believe the market has not yet fully incorporated long-run fundamentals, and market prices diverge from the base rate implied by historical data.

• Fade poll-driven reversals: When a short-term poll causes a prediction market price to spike or crash, this represents an overreaction. The market price before the poll was likely closer to the true probability. Buy against poll-driven moves, especially for events more than 100 days away.

• Science replication markets: If prediction markets for scientific replication exist (as on platforms like Metaculus), contracts priced below 50% for study replication are strong signals of impending negative results. These markets are accurate enough to provide 60-90 day advance warning of failed replications.

Why It Works

Prediction markets aggregate diverse information sources and are continuously updated, while polls are episodic snapshots subject to sampling bias, question wording effects, and timing. At long horizons, these polling biases compound, while market prices self-correct through arbitrage.

Strategy 6: Price Interpretation at Extremes

Source Papers: Manski (2006); Wolfers & Zitzewitz (2006)

Expected Edge: 2-10% probability correction at contract price extremes

Confidence: Medium-High (theoretically well-grounded, empirically supported)

The Inefficiency

Manski demonstrated that prediction market prices approximate the mean beliefs of traders rather than being exact probability estimates. The divergence between price and true probability depends on belief dispersion among traders and their risk aversion. Wolfers and Zitzewitz showed that under log utility, prices exactly equal mean beliefs, but under more realistic utility assumptions (CRRA with risk aversion > 1), prices are biased toward extremes.

Concretely: a contract priced at $0.10 may reflect a true probability closer to $0.13-$0.15 because risk-averse traders are willing to sell at prices below fair value to offload risk. Conversely, a contract at $0.90 may reflect a true probability of $0.85-$0.87.

How to Trade It

• Mid-range confidence: Trust prediction market prices most when they fall in the $0.20-$0.80 range. In this zone, prices are reliable proxies for aggregate beliefs regardless of utility assumptions.

• Extreme price adjustment: For contracts below $0.15 or above $0.85, apply a 3-8% correction toward $0.50. The true probability of a $0.10 contract is likely $0.13-$0.15; the true probability of a $0.90 contract is likely $0.85-$0.87.

• Belief dispersion assessment: When participants on a market publicly disagree (visible in comment sections, social media, or prediction platform discussions), the market price is a less reliable probability estimate. High dispersion = larger correction needed.

• Use this defensively: Even if you are not actively trading on extreme price corrections, understanding this bias prevents you from over-weighting the signal from contracts near 0 or 1. A contract at $0.95 does not mean 95% probability; it may mean 88-92%.

Strategy 7: Political Connection Trading

Source Papers: Coulomb & Sangnier (2014)

Expected Edge: 1-2.5% abnormal returns per 10% shift in candidate probability

Confidence: Medium (single study, but strong methodology with French election data)

The Inefficiency

Coulomb and Sangnier used prediction market data from the 2007 French presidential election to study how political outcomes affect firm values. They found that firms with executive or shareholder connections to political candidates experienced abnormal stock returns of up to 25% based on changes in the prediction market probability of their connected candidate winning. Crucially, the stock market lagged the prediction market by 48-72 hours.

How to Trade It

• Monitor political prediction markets as leading indicators: When a candidate gains or loses 10+ percentage points in prediction markets, identify publicly traded firms with known connections to that candidate (through executive donations, policy platform alignment, regulatory exposure, or government contracts).

• Trade the lag: The prediction market moves first; the stock market follows 2-3 days later. This window allows you to position in equities based on the prediction market signal before the broader market prices in the political shift.

• Reverse application: If you observe unusual stock price movements in politically connected firms, this may signal that insiders expect a political outcome shift. Monitor the corresponding prediction market for confirmation.

Limitations

This strategy depends on correctly identifying political connections, which requires research beyond what prediction market data alone provides. It also works best around elections with clear binary outcomes and may be less effective in systems with coalition governments or complex policy landscapes.

Strategy 8: Polymarket Intra-Market & Combinatorial Arbitrage

Source Paper: Saguillo, Ghafouri, Kiffer & Suarez-Tangil (2025) --- "Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets"

Expected Edge: Risk-free returns of 2--60% per dollar on detected opportunities; approximately $40M extracted historically on Polymarket during the measurement period.

The Core Insight

Strategy 4 covered arbitrage between different platforms (e.g. Kalshi vs. Polymarket). Saguillo et al. show that substantial arbitrage also exists within Polymarket itself. They identify two distinct forms: Market Rebalancing Arbitrage (within a single market) and Combinatorial Arbitrage (across dependent markets). Empirically, they found $40 million of realized arbitrage profit extracted during their measurement period, with the median profit-per-dollar on detected intra-market opportunities running around $0.60.

Variant A: Market Rebalancing Arbitrage

Every Polymarket condition has a YES and NO token that should sum to $1.00. When the sum deviates, there is a guaranteed profit available. The paper identifies both directions:

• Long Rebalancing (YES + NO < $1.00): Buy one unit of YES and one unit of NO. When the market resolves, one token pays $1.00 and the other pays $0.00. Profit = $1.00 − (YES price + NO price).

• Short Rebalancing (YES + NO > $1.00): Use Polymarket's Split mechanism to mint a YES/NO pair from $1.00 of USDC, then sell both tokens immediately at the elevated prices. Profit = (YES price + NO price) − $1.00.

Of 17,200 Polymarket conditions in the study, 7,051 had at least one arbitrage opportunity meeting the authors' profit threshold. The sports category had the most frequent opportunities; politics markets (particularly 2024 U.S. presidential election conditions) had the largest absolute profits per opportunity.

Variant B: Combinatorial Arbitrage

Polymarket's NegRisk markets group related conditions that are mutually exclusive --- for example, "Who will win the 2024 Democratic nomination?" lists Biden, Harris, Newsom, and so on. The sum of all condition prices in such a market should equal $1.00. When two related NegRisk markets have dependent subsets whose prices are inconsistent, a combinatorial arbitrage exists.

The execution pattern: identify a subset of conditions in Market 1 that must logically map to a subset in Market 2 (e.g. "Democratic nominee" conditions and "2024 presidential winner is a Democrat" conditions). When the aggregated prices diverge, buy the cheaper side and sell the more expensive side. The paper shows these opportunities were more lucrative and longer-lived than intra-market rebalancing arbitrage, particularly in politics markets.

How to Execute

• Scan Polymarket's on-chain orderbook for conditions where the VWAP sum of YES + NO deviates from $1.00 by more than 2--5 cents.

• For opportunities where the sum is below $1.00, buy both tokens. For sums above $1.00, call the Split function on the USDC conditional-token contract to mint a matched pair, then sell both at market.

• For combinatorial opportunities, use a dependency graph (topical similarity + mutual exclusivity) to identify related NegRisk markets. The paper used an LLM-assisted heuristic to reduce the search space from exponential to tractable.

• Budget for gas and slippage --- the authors limit their analysis to opportunities yielding at least $0.05 per dollar because executing multiple orders is non-atomic and partial fills create risk.

Where It Works Best

The study found arbitrage concentrations differ by category. Sports markets had the highest frequency of opportunities (median profit per dollar around $0.60), driven by fast-moving odds and many thin conditions. Politics had fewer but larger opportunities, especially in NegRisk markets around the 2024 election cycle. Crypto markets showed the largest outliers in individual opportunity size. The authors note that arbitrage tends to appear during periods of volatility --- when prices are moving quickly, market makers lag and price inconsistencies persist long enough to capture.

Caveats

On-chain execution requires gas and atomic-execution tooling; multi-leg trades where only part of the book fills can leave naked exposure. The market rebalancing strategy is asymmetric: long arbitrage requires holding positions until resolution (capital locked for days or weeks), while short arbitrage via Split/Merge can be closed immediately. Combinatorial arbitrage requires more sophisticated monitoring infrastructure and careful verification that the market structures actually map to one another --- subtle differences in resolution criteria can turn an apparent arbitrage into a directional bet.

Strategy 9: Liquidity Provision -- The Underwriting Edge on Kalshi

Source Paper: Palumbo (2026) --- "A Microstructure Perspective on Prediction Markets"

Expected Edge: Aggregate passive-liquidity profit of approximately $29M across one NFL regular season; profitable on a season basis but with significant weekly drawdowns.

The Core Insight

Palumbo analyzes Kalshi's NFL orderbook at the trade level and finds that passive liquidity providers --- traders who post resting limit orders rather than crossing the spread --- systematically end up with directional terminal exposure. In most markets, LPs finished the season holding an asset on one outcome and a liability on the other, rather than flattening to zero as classical market makers do. Aggregate profits were positive, but profitability depends on managing imbalance rather than eliminating it. The key takeaway: Kalshi LPs are functionally closer to underwriters than market makers.

Why This Edge Exists

Event contracts have four structural features that distinguish them from equity market making: (1) dynamic supply creation --- contracts are minted and retired inside the venue; (2) direct retail distribution --- end users trade against posted liquidity rather than through intermediaries; (3) binary resolution at $0 or $1; and (4) no underlying spot market to hedge against. When retail demand concentrates on one side of a contract, a classical market maker has no way to lay off the risk. Instead, the LP must hold the directional exposure through settlement and get paid for bearing it.

Palumbo's regression results suggest profitability is correlated with flow imbalance and pricing discipline, not passive spread capture. In other words, LPs who price wider when flow skews heavily one-directional earn more than LPs who quote tight on both sides of every market.

How to Execute

• Target markets with persistent one-sided retail flow --- sports markets where public bias is visible (favourites, popular teams) are natural candidates. Quote passively on the side the public is buying into, and widen the spread on that side as flow concentrates.

• Accept terminal exposure as the profit engine, not the risk. Treat the position like underwriting: size it such that adverse outcomes are survivable at the portfolio level even when a single market moves against you.

• Use Kalshi's published liquidity incentive programs as a rebate, not the primary edge. The paper notes that penny-wide spreads with only rebate compensation likely underprice the inventory risk being absorbed --- meaning there is room for a disciplined LP to quote slightly wider and still get filled.

• Diversify across many uncorrelated events. Weekly P&L showed significant drawdowns when multiple favourites lost simultaneously; volume of independent bets is what converts the edge from noisy to reliable.

Worked Example

Consider an NFL game where the Chiefs are favourites priced at $0.72 Yes / $0.28 No, with heavy retail demand for Chiefs Yes. A classical market maker quoting $0.71/$0.73 for 100 contracts gets filled repeatedly on the $0.73 offer and accumulates short exposure to Chiefs. If the Chiefs win, the LP loses $0.27 per contract on that inventory. To compensate, the LP under Palumbo's framework would price the offer side wider (say $0.74) when flow is persistently one-directional, treating the spread as underwriting premium. Over a season of hundreds of such games, the premium compounds into the $29M aggregate figure the paper documents --- but only if position sizing survives the weeks when multiple favourites lose.

Where It Works Best

This strategy works best in markets where (a) retail flow is persistent and one-directional, (b) there is no natural hedging instrument (true for most Kalshi events), and (c) the LP can diversify across many independent contracts. Sports markets meet all three conditions, as do broad economic and policy markets. It works worst in markets with professional two-sided flow (e.g. high-volume political contracts where informed participants quote against you) and in markets where outcomes are highly correlated (e.g. multiple contracts tied to the same macro event).

Caveats

Unlike arbitrage strategies, this is not risk-free. Weekly drawdowns in the paper were significant, and a poorly diversified book can blow up on a single correlated event. The strategy also requires either direct exchange connectivity or API-driven order management to reprice quotes quickly as flow develops. Finally, Palumbo's results are aggregate across all LPs; the marginal LP entering this strategy today competes with incumbents and may see compressed margins. The conceptual edge --- that event-contract liquidity provision is economically underwriting, not market making --- is the durable takeaway even if the specific $29M figure is not directly replicable.

Strategy 10: Dynamic Hedging -- Lock-In Profit on Odds Movement

Source Paper: Axén & Cortis (2020) --- "Hedging on Betting Markets"

Expected Edge: Variable; depends on magnitude of odds movement during holding period. Paper derives closed-form profit expressions under several market structures.

The Core Insight

Most academic work on prediction markets assumes participants bet on outcomes. Axén and Cortis formalize a different --- and practically dominant --- approach: speculate on the direction of odds movement, not the outcome itself. Take an initial position when you believe the price is away from fair value, wait for the market to correct, then close the position by taking the opposite side. If the odds move your way, you lock in a guaranteed profit regardless of how the event resolves. This is the same logic that governs closing a winning options position before expiry.

Unlike Strategy 1 (outcome betting) and Strategy 2 (hedge underpricing at a point in time), this strategy requires no view on the underlying event. It only requires a view on whether current prices are displaced from where they will settle over the next hours, days, or weeks.

The Basic Result

On a two-sided market with both back and lay options available (e.g. Polymarket, Betfair, Smarkets), the paper's Proposition 1 gives the profit from opening a back position at odds αᵢ and later closing it with a lay bet at odds β′ᵢ. The return is αᵢ/β′ᵢ − 1, with profit when αᵢ > β′ᵢ. The reverse holds for laying first and closing with a back bet: profit when α′ᵢ > βᵢ.

Intuitively: if you bought at $0.20 implied probability and now the market is quoting $0.35 on the same outcome, closing the position at $0.35 locks in a 75% return on your original stake, risk-free. The maximum theoretical return is bounded by the original odds (as lay odds approach 1, the profit approaches αᵢ − 1), meaning the biggest lock-in profits come from opening positions at longer odds.

Fees on a Betting Exchange

Kalshi, Polymarket, Betfair, and similar venues charge fees. The paper's Proposition 3 shows that when the fee τ is charged on the winnings of each bet separately, the break-even condition becomes (aᵢ − 1)/(b′ᵢ − 1) > 1/(1−τ)². In other words, the fee is applied twice (once on each leg), so the required odds movement scales inversely with (1−τ)².

However, Proposition 4 shows that if the exchange charges the fee only on the net P&L of paired positions, the break-even condition reverts to aᵢ > b′ᵢ --- the fee rate drops out of the break-even expression entirely. This is a meaningful distinction: a trader running a dynamic hedging book should identify which fee structure the platform uses, since it materially affects the minimum odds movement required to be profitable.

Markets Without a Lay Option

Traditional sportsbooks (Pinnacle, William Hill) do not allow direct lay bets. Kalshi technically does not offer a symmetric lay either --- it routes "buy No" orders through a different contract. The paper shows you can synthetically replicate a lay on outcome i by Dutching (splitting stake proportionally) across all other outcomes in N. But this only works when the market's overround permits it: the condition 1/αᵢ > π(α) must hold. When the implied probability of the target outcome is smaller than the total excess overround across the book, the synthetic lay is unprofitable by construction.

Practically: in a two-outcome binary market (most Kalshi and Polymarket contracts), this reduces to buying the opposite contract at the current price. In a multi-outcome market, you need to Dutch across all remaining outcomes, weighted such that each pays off the same amount regardless of which one hits.

How to Execute on Prediction Markets

• Identify contracts where you believe the current price is displaced from where the market will settle --- overreactions to news, thin liquidity pushing prices too far, or early-cycle mispricing that you expect to correct.

• Open the initial position in the direction of the expected correction. Size based on your confidence in the move and the liquidity available to exit.

• Set a target exit price in advance, based on the minimum odds movement required to cover fees and clear your cost-of-capital hurdle. Using the paper's framework, this is αᵢ/β′ᵢ − 1 ≥ your minimum acceptable return.

• Close by selling the position (on Polymarket/Betfair) or buying the opposite contract (on Kalshi for binary markets, or Dutching across other outcomes for multi-outcome markets).

• Monitor liquidity constraints. The paper's Leicester 2015/16 example illustrates the extreme case: closing a bet that moved from 5000-to-1 down to 4-to-1 would require 3,750× the original stake in fresh capital to cover the closing position. Position sizing must account for the possibility that the odds move far enough that you cannot fully close the hedge.

Worked Example on Polymarket

Suppose a political candidate is trading at $0.15 Yes on a Polymarket condition (implied probability 15%), and you believe a market-moving event is likely to push the price toward $0.30 within a week. Buy 1,000 Yes tokens at $0.15 for $150 total. A week later the event occurs and Yes is trading at $0.32. Sell all 1,000 Yes tokens at $0.32 for $320. Locked-in profit: $170 (113% return on the initial $150 stake), regardless of whether the candidate ultimately wins or loses. The trade had outcome risk during the holding period, but at exit the position is fully flat and the profit is deterministic.

Where It Works Best

This strategy thrives in markets where prices move meaningfully over time and where your edge is in identifying temporary displacements rather than predicting final outcomes. It works particularly well when combined with the signals from earlier strategies: Strategy 3 (informed-trader flow) can identify entry points; Strategy 5 (long-horizon markets beating polls) can identify situations where the market is ahead of public perception; Strategy 6 (price-interpretation corrections) gives you an exit target. Betting exchanges with the "fee on net P&L" structure (per Proposition 4) are structurally more favourable than fee-on-winnings venues.

Caveats

The strategy is not risk-free during the holding period --- the profit only becomes deterministic at the moment the hedging leg is placed. If the odds move against you before you can close, you may be forced to exit at a loss. Liquidity risk is material, especially in markets where your opening trade consumed much of the visible order book: closing may require walking down the book and giving up some of the apparent edge. Finally, the strategy compounds fees (two round trips instead of one), so on high-fee platforms the required odds movement to clear the break-even threshold may be larger than casual traders realize.

Strategy Comparison

Strategy Edge Confidence Complexity Best Domain

---

Favourite-Longshot Bias 2-5% Very High Low Sports, entertainment, novelty markets Hedge Underpricing 3-15% Medium Medium Platforms with independent YES/NO books Informed Trader Flow 2-5% Medium Medium Markets near resolution deadlines Multi-Platform Arbitrage 1-5% Very High High Events listed on 2+ platforms Long-Horizon Forecasting 5-15% High Low Elections, policy, science replication Price Extremes Correction 2-10% Med-High Low All markets at price extremes Political Connection 1-2.5% Medium High Election cycles, connected firms Intra-Market Arb (Polymarket) 5-60% Medium High Polymarket YES/NO & NegRisk mkts LP / Underwriting (Kalshi) Variable High High One-sided flow sports & events Dynamic Hedging Variable Medium Medium Volatile markets, news events

Practical Implementation Notes

Platform Considerations

Different prediction market platforms have different fee structures, liquidity profiles, and contract types. Strategies 1, 2, and 6 work on any platform. Strategies 3 and 5 require platforms with sufficient liquidity for real-time trading. Strategy 4 requires accounts on multiple platforms. Strategy 7 requires access to both prediction markets and equity markets. Strategy 8 is Polymarket-specific and requires on-chain execution tooling to capture intra-market and combinatorial arbitrage. Strategy 9 is Kalshi-specific and suits participants with exchange connectivity and the capital base to absorb directional LP exposure. Strategy 10 works on any platform that allows closing positions before resolution and is particularly efficient on venues where fees are charged on net P&L rather than per-leg winnings.

Bankroll Management

The favorite-longshot bias and price-extreme correction strategies generate small, consistent edges that compound over many trades. They are best suited to a high-frequency, small-position approach. Hedge underpricing and arbitrage strategies offer larger per-trade returns but require more capital per position. Informed trader flow is a selective strategy best used sparingly when clear signals emerge.

Combining Strategies

The strategies are largely complementary. A systematic approach might combine favourite betting as a baseline (Strategy 1), with hedge monitoring for risk management (Strategy 2), informed-flow overlays near resolution (Strategy 3), cross-platform arbitrage scanning for risk-free opportunities (Strategy 4), intra-market and combinatorial arbitrage scanners running on Polymarket (Strategy 8), and extreme-price correction as a valuation adjustment (Strategy 6). The long-horizon and political connection strategies operate on different timeframes and can run in parallel. For participants with the capital base and execution infrastructure to support it, passive liquidity provision on Kalshi (Strategy 9) offers a separate revenue stream with low correlation to the directional strategies above --- effectively letting an operator act as both taker (via Strategies 1, 3, 5, 6, 7) and underwriter (via Strategy 9) across different books.

Key Risks

• Liquidity risk: Many prediction markets have thin order books. Large positions move prices, and exiting can be costly. Size positions relative to visible liquidity.

• Resolution risk: Ambiguous resolution criteria can turn winning positions into losses. Read contract terms carefully, especially on newer platforms.

• Regulatory risk: The legal landscape for prediction markets varies by jurisdiction and is evolving. Ensure compliance with local regulations before deploying capital.

• Platform risk: Prediction market platforms can fail, freeze withdrawals, or change rules. Diversify across platforms and avoid concentrating capital.

• Edge decay: As prediction markets mature and attract more sophisticated participants, some of these edges will narrow. The favourite-longshot bias has persisted for decades, but arbitrage opportunities close more quickly as markets grow.

Source Papers

Core Strategy Papers

Buhagiar, R., Cortis, D. & Newall, P. (2018). Why Do Some Soccer Bettors Lose More Money Than Others? Journal of Behavioral and Experimental Finance.

Ottaviani, M. & Sorensen, P. (2009). Surprised by the Parimutuel Odds? American Economic Review.

Franck, E., Verbeek, E. & Nuesch, S. (2010). Prediction Accuracy of Different Market Structures. International Journal of Forecasting.

Frederick, S. et al. (2018). Valuing Bets and Hedges: Implications for the Construct of Risk Preference. Judgment and Decision Making.

Chatterjee, K. & Mookherjee, D. (2018). Valuing Bets and Hedges.

Newall, P. & Cortis, D. (2019). High-Stakes Hedges are Misunderstood.

Cortis, D. (2015). Expected Values and Variances in Bookmaker Payouts. Journal of Quantitative Analysis in Sports.

Axen, G. & Cortis, D. (2019). Extending the Price Constraints of Betting Markets.

Axen, G. & Cortis, D. (2020). Hedging on Betting Markets.

Peel, D. (2017). Wagering on More Than One Outcome.

Market Accuracy and Interpretation Papers

Berg, J., Nelson, F. & Rietz, T. (2008). Prediction Market Accuracy in the Long Run. International Journal of Forecasting.

Dreber, A. et al. (2015). Using Prediction Markets to Estimate the Reproducibility of Scientific Research. PNAS.

Manski, C. (2006). Interpreting the Predictions of Prediction Markets.

Wolfers, J. & Zitzewitz, E. (2006). Interpreting Prediction Market Prices as Probabilities.

Market Structure and Context Papers

Agrawal, S. et al. (2010). Equilibrium in Prediction Markets with Buyers and Sellers.

Flepp, R., Nuesch, S. & Franck, E. (2017). The Liquidity Advantage of the Quote-Driven Market. Quarterly Review of Economics and Finance.

Coulomb, R. & Sangnier, M. (2014). The Impact of Political Majorities on Firm Value. Journal of Public Economics.

Kovalchik, S. (2016). Searching for the GOAT of Tennis Win Prediction. Journal of Quantitative Analysis in Sports.

Platform Microstructure & On-Chain Arbitrage Papers

Palumbo, N. (2026). A Microstructure Perspective on Prediction Markets. SSRN Working Paper 6325658.

Saguillo, O., Ghafouri, V., Kiffer, L. & Suarez-Tangil, G. (2025). Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets. arXiv:2508.03474.

Disclaimer: This document is for informational and educational purposes only. It does not constitute trading or investment recommendations. Trading involves substantial risk of loss. Past performance and academic findings are not indicative of future results.