The Favourite-Longshot Bias: Empirical Evidence Across 2,490 Matches & Mathematical Mechanics

1. Introduction: The Fundamental Anomaly of Sports Betting Markets

In classical financial economics, the Efficient Market Hypothesis (EMH) posits that asset prices reflect all available information, and expected returns across assets should equalize once adjusted for systematic risk. If a betting exchange or bookmaking market were perfectly efficient in an information-theoretic sense, the expected return of placing a wager would be identical across all price points. A bettor wagering on a heavy favorite priced at 1.25 decimal odds should face the exact same negative mathematical expectancy—dictated solely by the bookmaker's transaction fee or overround—as a bettor backing a 15.00 longshot outsider.

Empirical reality across every major sports betting jurisdiction in the world violently contradicts this assumption. For nearly a century, empirical researchers, quantitative syndicates, and financial econometricians have documented a pervasive, resilient, and statistically indisputable pricing distortion: the Favourite-Longshot Bias (FLB).

First formally documented by Griffith (1949) in United States pari-mutuel thoroughbred horse racing and later validated in European sports markets by Ali (1977), Thaler and Ziemba (1988), and Shin (1991, 1993), this phenomenon is not a temporary statistical aberration. It is a structural feature of betting market microstructure. Whether analyzing horse racing, professional tennis, NFL point spreads, or European association football, the empirical return curve slopes downward: as odds lengthen, financial drag typically increases.

This article presents empirical evidence of the favourite-longshot bias derived from our benchmark dataset of 2,490 English Premier League (EPL) closing lines (August 2019 to January 2026, comprising 7,470 1X2 selections, source: football-data.co.uk, dataset: public/papers/epl-closing-lines-2019-2026.csv). We explore the behavioral, psychological, and market microstructure mechanisms driving this asymmetry, compare how multiplicative and Shin devigging price longshots, and establish concrete quantitative rules for navigating this structural phenomenon.

2. Empirical Evidence: Analysis of 2,490 Premier League Closing Lines

To evaluate the empirical presence and magnitude of the favourite-longshot bias in modern liquid soccer betting markets, the Quantitative Research Division analyzed 2,490 consecutive match closing lines across 1X2 match-odds markets (August 9, 2019 – January 8, 2026, across 7 seasons) from Pinnacle and market-average consensus closing lines, compiled in our benchmark repository (source: football-data.co.uk, dataset: public/papers/epl-closing-lines-2019-2026.csv).

The 2,490 matches comprise exactly 7,470 individual 1X2 betting selections. We categorized selections into six discrete odds buckets based on their decimal closing line price, then evaluated the cumulative financial performance under a standardized flat-staking simulation of 1.00 unit per wager across both Pinnacle closing odds and the broader market average closing odds.

Summary of 2,490 EPL Matches by Decimal Odds Tier (7,470 Selections)

Nevertheless, the empirical gradient shows that realized returns generally decline as decimal odds increase. On Pinnacle closing lines, flat staking on moderate favourites (1.50–2.00) yielded a mild -0.32% ROI, whereas longshots priced above 5.00 experienced returns of -5.44% and -5.51%. Across market-average closing lines—where retail bookmakers incorporate larger overall margins (averaging 4.16% vs. 2.70% at Pinnacle)—the penalty on longshots is even more pronounced, with losses deepening to -9.70% in the 5.00–10.00 bracket.

3. Behavioral Economics: Why Bettors Irrationality Inflates Longshot Prices

Why does this pricing inefficiency persist in liquid, multimillion-dollar markets? In traditional equity markets, an asset that is systematically overpriced by 20% would be ruthlessly shorted by arbitrageurs until its price collapsed to fundamental value. In sports betting markets, however, shorting mechanisms are restricted (traditional retail bookmakers do not allow you to lay an outcome), transaction costs are non-trivial, and retail capital continuously replenishes the distortion.

The persistence of the favourite-longshot bias is rooted in human cognitive architecture and decision-making under risk, as explained by three foundational behavioral economics theories.

1. Cumulative Prospect Theory & Probability Weighting (Kahneman & Tversky, 1992)

In standard Expected Utility Theory (EUT), an agent evaluates a gamble by weighting the utility of each financial payout by its objective mathematical probability $p$. In contrast, Daniel Kahneman and Amos Tversky demonstrated through Cumulative Prospect Theory (CPT) that human decision-makers do not process objective probabilities linearly. Instead, cognitive processing applies an inverted S-shaped probability weighting function $w(p)$:

Where empirical parameter estimations typically find $\gamma \approx 0.65$ for monetary gains. The mathematical implications of this weighting function in sports betting are profound:

• Overweighting of Low Probabilities ($p<0.10$): A recreational bettor evaluates a 2.0% objective probability as though it were a 6.0% or 7.0% chance. The prospect of an underdog winning appears cognitively far more plausible than it actually is.
• Underweighting of Moderate to High Probabilities ($p>0.40$): A bettor evaluates a 75% favourite as if it were a 68% chance, feeling that a heavy favourite carries "too much risk for too little return."

Because retail bettors systematically overweight low probabilities, their subjective willingness to pay for a 10.00 or 15.00 ticket far exceeds the fair mathematical value. Bookmakers respond to this unyielding consumer demand by lowering the payout (raising the overround) on longshots.

2. Skewness Preference & The Lottery Ticket Effect

Modern portfolio theory measures risk primarily through variance (${\sigma }^2$). However, behavioral finance has demonstrated that retail gamblers and retail options traders exhibit a pronounced preference for positive skewness (the third standardized moment of the probability distribution, ${\mu }_3=𝔼[{(X-\mu )}^3]/{\sigma }^3$).

A wager on a heavy favourite ($O=1.20$) has negative skewness: you win a small amount frequently, but occasionally suffer a catastrophic 100% loss of your stake. A wager on a longshot ($O=12.00$) has positive skewness: you lose small amounts frequently, but hold the possibility of an explosive, multi-unit windfall. Bettors derive non-monetary psychological utility from positive skewness—the thrill of a life-changing payout for a negligible upfront cost. Much like national lottery tickets, bettors willingly accept a negative expected return (-20% EV) in exchange for positive skewness.

3. Cognitive Anchoring on Absolute Payouts

Recreational bettors rarely perform Bernoulli trial calculations or calculate logarithmic bankroll growth. Instead, they anchor on absolute nominal payouts. When presented with two choices:

• Bet A: Risk $100 to win $20 on a 1.20 favourite (82% win rate).
• Bet B: Risk $10 to win $110 on a 12.00 outsider (7% win rate).

The recreational mind perceives Bet A as unattractive ("Why risk $100 just to make $20?") and Bet B as an asymmetric bargain ("It's only $10, and I could win $110!"). This cognitive anchoring renders the longshot price completely inelastic to changes in the bookmaker's margin. If a bookmaker widens the margin on a 12.00 outsider to 10.50, retail wagering volume barely fluctuates. If the same bookmaker cuts a 1.20 favourite to 1.15, sharp bettors and volume players instantly flee. Bookmakers optimize their revenue by placing the bulk of their tax where demand is least elastic.

4. Market Microstructure: Bookmaker Risk Management & The Shin Model

While consumer psychology explains retail bettor behavior, it does not explain why bookmakers actively set asymmetrical prices. Why don't bookmakers price longshots fairly and pocket a uniform margin across all outcomes? The answer lies in market microstructure and risk management against informed traders.

In 1991 and 1993, econometrician Hyun Song Shin developed a groundbreaking structural model of odds compilation. Shin proved that the presence of informed traders (bettors possessing non-public private information, such as match-fixing syndicates, insider team news, or superior quantitative models) compels bookmakers to distort longshot odds to protect their solvency.

The Mathematics of the Shin Model

Shin modeled a market consisting of two distinct classes of market participants: 1. Noise Bettors (Proportion $1-z$): Uninformed recreational bettors who wager randomly or according to subjective biases. 2. Informed Insiders (Proportion $z$): Sharp bettors who possess certain knowledge of the true outcome.

If an insider possesses private knowledge that a heavy favourite will win, their profit potential is capped by the low odds (e.g., betting $1,000 at 1.25 yields $250 profit). However, if an insider discovers private information regarding a massive upset (e.g., key starters rested, goalkeeper bribed, extreme sickness in the camp) and wagers on a 12.00 longshot, the financial liability inflicted on the bookmaker is substantial (betting $1,000 at 12.00 drains $11,000 from the bookmaker's reserves).

To insulate their books against catastrophic adverse selection by informed traders on longshots, the bookmaker must shade the published odds of low-probability outcomes downward. Shin showed that the bookmaker's published probability ${\pi }_i=1/O_i$ relates to the true probability $p_i$ and the insider parameter $z$ via:

Solving this quadratic equation for the true probability $p_i$ yields the celebrated Shin Inversion Formula:

Where the insider trading parameter $z$ is solved iteratively (typically via the Newton-Raphson method or fixed-point iteration) such that the sum of true probabilities equals unity: ${\sum }_{i=1}^n{p}_i=1.000$.

Margin Allocation Under Shin's Model: An Illustration

Shin's model provides mathematical reasoning that rational, profit-maximizing bookmakers must assign a higher percentage margin to longshots.

Illustrative example (Shin model), not an observation from the dataset:

The ratio $(q-p)/p$ in this hypothetical example is relative price loading, not expected loss per unit stake or realized hold. Expected loss is $1-pO=1-p/q$ when $q=1/O$. The displayed 10.40% and 2.45% ratios compare loading relative to the assumed p values; they do not mean the bettor loses those percentages of turnover.

5. The Devigging Trap: Multiplicative vs. Shin in the Presence of Bias

The existence of the favourite-longshot bias creates a challenge for quantitative bettors who rely on a single devigging method. By far the most common no-vig calculation used across retail odds comparison websites and simple EV calculators is the Multiplicative (Proportional) Method:

The limitation of the multiplicative method is its underlying assumption: it assumes that the bookmaker distributed their overround exactly in proportion to the raw implied probability of each outcome. On 2,490 Premier League matches (football-data.co.uk) all four methods gave the same accuracy on Pinnacle closing odds (Brier 0.5676). Shin moves probability toward favourites: +0.79 pp for odds below 1.70 and −0.28 pp for odds above 3.50. The choice of method matters most for longshots and for markets with high margins or many outcomes; on liquid EPL 1X2 closing lines it did not improve accuracy.

A Practical Case Study: +EV Under One Method Only

Consider a hypothetical 1X2 match where a bookmaker offers the following closing prices with a total overround of 5.63% (S≈1.0563):

• Home Favourite: $O_1=1.30$
• Draw: $O_2=5.50$
• Away Underdog: $O_3=9.50$

Let us calculate the implied fair probabilities using both Multiplicative devigging and the Shin Model ($z=0.029$):

Now observe the consequence if a bettor finds another bookmaker offering the underdog at 10.50:

1. The Multiplicative Bettor: Compares 10.50 to the multiplicative fair probability of $9.97\%$ (fair odds 10.035).
   $${\text{EV}}_{\text{mult}}=(0.0997\times 10.50)-1=+4.69\%\text{(+EV under multiplicative)}$$
   The bettor believes they have found a lucrative positive EV wager.
2. The Same Bet Under Shin: Under Shin's model ($z=0.029$), the fair probability is $9.01\%$ (fair odds 11.104).
   $${\text{EV}}_{\text{Shin}}=(0.0901\times 10.50)-1=-5.40\%\text{(-EV)}$$
   Under Shin the bet is −EV (−5.40%). The edge exists under only one method, so it is not a confirmed edge.

6. Quantitative Exploitation: How to Systematically Profit from the Bias

Understanding the favourite-longshot bias is not merely a defensive exercise in avoiding bad bets; it forms the foundation of four profitable quantitative betting strategies.

Strategy 1: Laying Longshots on Betting Exchanges

An exchange lay position has liability, commission and matching requirements distinct from a sportsbook back bet. Use the account’s current commission and actual accepted prices. A price discrepancy against a model can imply estimated value under that model, but it does not establish the true probability or a typical exchange fee.

By systematically laying longshots with odds between 8.00 and 30.00 where the implied probability exceeds the Shin/Dixon-Coles fair probability by a margin wider than the exchange commission, quantitative layers extract positive mathematical expectancy from recreational optimism.

Strategy 2: Synthetic Favourites via Asian Handicaps

When an elite team (such as Real Madrid or Bayern Munich) plays a weak opponent, the retail 1X2 price on the favourite might be compressed to 1.18 (-1.5% EV due to vig). However, instead of taking the compressed moneyline, sharp bettors transition to the Asian Handicap market (e.g., -1.5, -2.0, or -2.5 goals).

A two-way handicap board need not have true 50/50 probabilities. Its two reciprocal prices can contain asymmetric loading. Compare the complete current quotes and use the win/push/loss settlement model appropriate to the line. The handicap label neither removes favorite-longshot effects by theorem nor establishes a universal margin range.

Strategy 3: Avoiding the Multi-Leg Accumulator (Parlay) Mirage

No market-wide revenue share is established here. For independent selections with a chosen expected loss 15% each, the next formula gives the ticket’s expected loss. This is a mathematical scenario; correlated tickets require their joint probability.

If you combine four longshots, each carrying an embedded negative expectancy of -15% due to the bias, the total ticket expectancy is:

The bettor is surrendering nearly half their stake to the house before a single match kicks off. Conversely, constructing accumulators out of high-probability outcomes (odds 1.20 to 1.40) where individual margin is minimal limits compounded overround to negligible levels.

Strategy 4: Calibrating Staking Sizes with Fractional Kelly

Because the variance of longshot wagers is extreme, any error in probability estimation has catastrophic consequences under Kelly Criterion staking. The Kelly staking fraction is given by:

In the stated hypothetical p=0.12,O=10 example, full Kelly is about 2.22%; if the actual p were 0.085 the true EV would be negative. A smaller chosen fraction reduces exposure but does not turn a false probability into a valid edge. Quarter and Eighth are possible policy inputs, not universally optimal or guaranteed protective fractions.

7. Conclusion & Mathematical Summary

The Favourite-Longshot Bias is the single most enduring empirical distortion in sports wagering markets. Driven by human cognitive vulnerability to low-probability high-reward payoffs (Cumulative Prospect Theory) and sustained by bookmaker risk management imperatives (The Shin Model), it systematically penalizes those who chase big payouts and rewards those who grind disciplined edges.

To succeed as a quantitative sports bettor, remember the four cardinal tenets of the bias: 1. Longshot returns face increased drag (-5.44% to -5.51% ROI on odds > 5.00 at Pinnacle close, and down to -9.70% on market-average close across 2,490 Premier League matches, source: football-data.co.uk). 2. Moderate favourites display minimal drag (-0.32% ROI on odds 1.50–2.00 at Pinnacle close, though 95% confidence intervals cross zero). 3. Relative to Shin, multiplicative devigging gives longshots slightly more probability (0.28 pp more on average above odds 3.50), which can make a longshot look +EV when it is not; check both methods. 4. On 2,490 Premier League matches all four devigging methods had the same accuracy (Brier 0.5676); the method matters for single longshot bets, not for overall calibration on liquid 1X2 lines.