Data-Driven Horse Racing Handicapping: How Analytics Beat Traditional Methods

The Analytics Revolution in Horse Racing

Horse racing handicapping has been transformed by data. What was once a craft built almost entirely on subjective observation and intuition has become an increasingly quantitative discipline. The best handicappers in 2026 combine traditional horse sense with statistical models, large-scale data analysis, and real-time prediction market signals to produce probability estimates that consistently beat the crowd.

This shift mirrors what happened in baseball with the Moneyball revolution, in basketball with the analytics movement, and in poker with the rise of solver-based strategies. In each case, practitioners who embraced data gained a decisive edge over those who relied solely on traditional methods. Horse racing is no different. The information is there for anyone willing to do the work -- and the reward for that work shows up directly in prediction market profits on platforms like predict.horse.

The fundamental principle of data-driven handicapping is simple: measure everything that matters, ignore everything that does not, and let the numbers guide your decisions. A horse's name, color, saddle cloth number, and connections' interview quotes are noise. Speed figures, pace data, track bias measurements, trainer statistics in specific situations, and jockey performance metrics are signal. Separating the two is where analytics earns its edge.

Speed Figures: The Foundation of Data-Driven Handicapping

Speed figures are the bedrock of modern handicapping. Every serious data-driven handicapper uses some form of speed rating to normalize performance across different tracks, distances, and conditions. The concept is straightforward: convert a horse's final time into a number that represents its performance level independent of where and when it raced.

How Speed Figures Work

Raw race times are meaningless in isolation. A 1:10 flat for six furlongs at Belmont Park on a fast day might represent a very different level of performance than 1:10 at Aqueduct on a slow day. Speed figure methodologies account for three primary variables: the inherent speed of the track surface on that specific day (the "track variant"), the distance of the race, and the wind and weather conditions.

The most widely used speed figure systems in 2026 include:

• Beyer Speed Figures: Published in the Daily Racing Form, Beyer figures are the industry standard. They use a scale where 100 represents a top-class performance and figures below 60 indicate lower-level competition. Beyer figures are accessible and widely understood, making them a useful baseline.
• Ragozin Sheets (The Sheets): A proprietary system that uses a lower-is-better scale. Ragozin figures are favored by many professional handicappers because they incorporate more granular adjustments for track conditions and because the sheet format allows visual pattern recognition of a horse's form cycle.
• Thoro-Graph: Similar to Ragozin in methodology (lower numbers are better), Thoro-Graph figures include adjustments for ground loss on turns and are particularly useful for identifying horses that ran better than their finishing positions suggest due to wide trips.
• TimeformUS: A modern speed and pace figure system that provides both overall speed ratings and individual fractional pace ratings. The pace component makes TimeformUS particularly valuable for pace analysis, which we cover in depth below.

Using Speed Figures in Prediction Markets

The key to using speed figures in prediction market trading is not simply picking the horse with the highest figure. It is understanding figure patterns and what they predict about future performance. Here are the three most important figure-based signals.

Ascending figures: A horse whose speed figures have improved in each of its last three races (for example, 82-87-93) is on an upward trajectory. Projection models suggest the next figure will continue the trend, often landing in the 96-100 range. If the prediction market on predict.horse prices this horse based on its most recent figure (93) rather than the projected figure (97+), it is undervalued.

Bounce patterns: Horses that run a career-best speed figure often regress in their next start -- a phenomenon known as the "bounce." If a horse just ran a 105 Beyer, the highest figure of its career by 8 points, the probability of it repeating or exceeding that figure in its next race is low. If prediction market prices assume the horse will maintain that level, the horse is overpriced. Sell or avoid.

Figure context: A speed figure must be evaluated in the context of the race. A figure of 95 earned in a wire-to-wire romp on a slow day might be less impressive than a 90 earned while closing from 15 lengths back in a fast-paced race. When two horses have similar top figures, the one whose figure was earned in more demanding circumstances is typically the better prospect.

Track Bias Analysis: The Hidden Edge

Track bias is one of the most powerful and most underutilized edges in horse racing handicapping. A track bias exists when the racing surface favors certain running styles or positions on the track in a way that is not accounted for by speed figures or standard handicapping methods.

Types of Track Bias

Speed bias: The most common bias. When the inside portion of the track is fast and the outside is deep, horses on the lead with inside position gain a significant advantage. On days with a strong speed bias, front-runners win at rates far above their expected probability. Speed biases are most common on dirt tracks after maintenance work that leaves the inside rail freshly harrowed.

Closer bias: The opposite of speed bias. When the rail is deep or cuppy and the outside path is faster, closers who swing wide on the turns have the advantage. Closer biases are common on dirt tracks after heavy rain when the inside becomes waterlogged and horses that stay off the rail have better footing.

Inside/outside path bias: On turf courses, the ground condition can vary dramatically across the width of the course. If the inside rail has been used for many days of racing, the ground may be chewed up and slower, creating an outside bias. Conversely, a course with fresh rail placement may have faster ground along the inside.

Measuring Track Bias

Quantifying track bias requires tracking the finishing positions and running styles of every horse on a given day or meet. The basic method is to calculate the win rate and average finishing position for each running style category (front-runners, stalkers, closers) and compare it to the expected baseline rates. If front-runners are winning at 40% on a track where the baseline rate is 22%, there is a strong speed bias present.

More sophisticated methods use statistical models that control for field quality, distance, and other variables to isolate the track effect. Several data providers now offer daily track bias ratings, and platforms like TimeformUS include bias indicators in their past performance data.

Trading Track Bias in Prediction Markets

Track bias creates prediction market opportunities because most casual participants do not account for it. If you observe a strong speed bias at Churchill Downs during Derby week (as occurred in 2019 when front-runners dominated the undercard), you know that the Derby itself is likely to favor speed horses. If the prediction market prices closers and speed horses without accounting for the bias, speed horses are undervalued. Act on this information before other participants recognize the pattern.

Bias Type	Favored Style	Detection Signal	Market Opportunity
Speed Bias	Front-runners	E/EP types winning >35%	Buy front-runners, sell closers
Closer Bias	Deep closers	S/C types winning >30%	Buy closers, sell speed
Rail Bias	Inside posts	Posts 1-4 winning >30%	Buy inside-drawn horses
Golden Rail	Stalkers on rail	2nd-3rd path dominance	Buy tactical stalker types

Pace Scenario Modeling: Predicting Race Shape

Pace is the most important predictive variable in horse racing that most handicappers underweight. The pace of a race -- how fast the field runs in the early stages -- determines which running styles are favored and which horses are most likely to be advantaged or disadvantaged by the race dynamics.

The Pace Spectrum

Every race falls somewhere on a pace spectrum from very slow (a single front-runner controls the pace unchallenged) to very fast (multiple speed horses engaged in a duel for the lead). Where a race falls on this spectrum has a direct and measurable impact on outcomes:

• Slow pace: Benefits front-runners and early pressers. The leader conserves energy and is difficult to catch. Closers rarely have enough runway to make up ground. Historical data shows front-runners win 35-40% of slow-pace races, compared to their 22% baseline.
• Moderate pace: The most common scenario. Benefits stalkers who sit just behind the pace and have the tactical speed to engage at the top of the stretch. No dramatic running style advantage.
• Fast pace: Benefits closers dramatically. Front-runners expend too much energy early and fade in the stretch. Deep closers who save ground and energy through the early stages can produce explosive late runs. In fast-pace scenarios, closers win at roughly 30-35%, compared to their 18% baseline.
• Suicidal pace: An extreme fast pace where multiple horses engage in a speed duel from the start. Benefits the deepest closers. In suicidal pace scenarios, horses that were last or near-last at the half-mile pole have won at rates exceeding 25%.

Building a Pace Projection

To project the pace of an upcoming race, classify each entrant's running style using their recent fractional times and running positions. Count the number of Early (E) and Early Presser (EP) types. Consider their post positions -- inside posts are more likely to engage early because they have the shortest path to the first turn.

A simple but effective pace model uses this framework: If there are zero or one E types, project a slow to moderate pace. If there are two E types, project a moderate to honest pace. If there are three or more E types, project a fast pace. If three or more E types are drawn within four stalls of each other, project a very fast to suicidal pace.

Once you have a pace projection, you can evaluate which horses in the field are most likely to benefit. Cross-reference this with prediction market prices on predict.horse. If the market has not priced in the projected pace scenario, you have an edge.

Weather and Surface Impact Quantified

Weather affects horse racing outcomes in ways that are measurable, predictable, and frequently mispriced by prediction markets. The three primary weather variables that impact racing are precipitation, temperature, and wind.

Precipitation and Track Condition

Rain transforms a racing surface. On dirt, even moderate rainfall can change a track from "fast" to "good" to "muddy" to "sloppy," and each gradation affects different horses differently. Data analysis of over 100,000 dirt races shows that wet-track performance is a persistent trait: horses that run well on wet tracks consistently outperform on off-going, and horses that struggle on wet tracks consistently underperform.

The key metric is the horse's "wet track index" -- the difference between its average speed figure on wet tracks versus dry tracks. A horse with a wet track index of +5 (meaning it runs 5 Beyer points better on wet tracks than dry) is a significant edge on an off-track day. A horse with a -8 wet track index is a significant negative on the same day. This data is readily available through modern past performance services and should be integrated into every prediction market analysis when rain is in the forecast.

Temperature Effects

Temperature affects both the racing surface and the horses themselves. Extremely hot days (above 90F) produce slower track conditions as the sun bakes the surface, and some horses handle heat better than others. Cold weather (below 40F) firms up the ground and can produce faster times but affects horses from warm-weather training centers who are unaccustomed to the cold. These effects are subtle but measurable, and they create prediction market edges in the 1-2 percentage point range -- small individually but significant in aggregate over a season of trading.

Wind Impact

Wind is the most overlooked weather variable in horse racing. A strong headwind on the backstretch slows the pace, benefiting front-runners who save energy in the final turn and stretch when the wind is at their backs. A strong tailwind on the backstretch creates faster early fractions, tiring speed horses and benefiting closers. Wind speed and direction data is freely available from weather services but is almost never integrated into handicapping models or prediction market pricing, making it a genuine hidden edge.

For weather-related prediction market opportunities across the Predict Network, check predict.surf for weather and surf prediction markets, and use that data to complement your horse racing analysis.

Jockey-Trainer Combo Analytics

The jockey-trainer combination is one of the most statistically significant predictive factors in horse racing. Certain jockey-trainer pairings produce win rates that far exceed what either individual produces on average, suggesting a synergy effect that data analysis can quantify and exploit.

Measuring Combo Strength

The basic combo metric is the win rate and ROI of a specific jockey-trainer pair over the past two to three years with a minimum sample size of 50 starters. A combo that wins at 28% when the trainer's overall rate is 18% and the jockey's overall rate is 20% is producing a synergy effect worth approximately 8-10 percentage points of excess performance. This is a massive edge.

Advanced combo analysis goes further. It segments the data by race type (maiden, claiming, allowance, stakes), distance, surface, and situational factors (first off layoff, class drop, etc.). A trainer might have a 15% overall win rate but a 35% rate with a specific jockey in maiden races on turf. That specific cell in the data matrix is extremely valuable because it tells you exactly when this combo is most dangerous.

Jockey Switch Analysis

When a trainer switches jockeys on a horse, it is a powerful signal. An upgrade to a top jockey suggests the connections believe the horse is ready to run a big race. A downgrade or switch to an apprentice may indicate lower expectations. Data analysis shows that jockey upgrades (moving from a jockey ranked outside the top 10 at the meet to one ranked in the top 5) produce a win rate approximately 5 percentage points higher than the baseline, controlling for other factors.

In prediction markets, jockey switches are often underpriced in the first 24-48 hours after announcement because not all participants monitor jockey assignments closely. A trader who tracks daily entries and jockey changes can identify underpriced horses before the market adjusts.

Trainer Situational Statistics

Trainer statistics in specific situations are among the most persistent and exploitable edges in the game. Key situational categories include:

• First-time starters: Some trainers excel at having horses ready to run well on debut. A trainer who wins at 30% with first-time starters (versus the 10% base rate) is providing an enormous edge.
• Second off layoff: Many horses improve dramatically in their second start after a break. Trainers who specialize in this pattern win at rates 15-20 percentage points above baseline.
• Surface switches: Moving a horse from dirt to turf (or vice versa) is a common angle. Trainers with high win rates in surface-switch situations are signaling that they have identified an aptitude the horse has not yet demonstrated publicly.
• Class drops: A trainer dropping a horse in class while maintaining a top jockey is often setting up a "win bet." Data shows this combination produces win rates above 30% for top barns.
• Distance changes: Stretching out from sprints to routes or cutting back from routes to sprints. The best trainers have a clear record of success with specific distance change patterns.

Class Analysis and Company Line Methods

Class -- the quality of competition a horse has faced -- is one of the hardest variables to quantify but one of the most important. A horse that finished fifth in a Grade 1 stakes may have more ability than a horse that won a maiden claiming race. The challenge is measuring the gap precisely enough to identify mispriced horses in prediction markets.

The Company Line Method

Company line analysis tracks subsequent performance of horses that competed in the same race. If five horses from a specific race go on to win their next start, that race was clearly a strong field. Horses from that race that have not yet had a subsequent start are likely undervalued because the market has not yet seen evidence of the field quality.

This method is particularly powerful for evaluating maiden race graduates. A horse that won a maiden race where the second, third, and fourth-place finishers all came back to win is obviously more impressive than one that won a race where the other finishers have continued to lose. Company lines take two to four weeks to develop fully, so early-acting traders who track field strength indicators can identify strong company before the data becomes widely available.

Class Par Speed Figures

Each class level has an expected average speed figure, known as the "class par." The class par for a Grade 1 stakes on dirt at one mile might be 108 Beyer, while the par for an allowance race at the same track might be 94. Horses that have consistently met or exceeded the par for higher class levels are likely to dominate at lower levels. When a class dropper's prediction market price does not fully reflect the class advantage, it is a value buy.

Prediction Market Signals vs. Traditional Handicapping

Prediction markets and traditional handicapping are not opposing approaches -- they are complementary tools that work best when combined. Understanding how prediction market prices relate to handicapping analysis helps you identify where the crowd is right, where it is wrong, and where the real opportunities lie.

Where Prediction Markets Excel

Prediction markets are excellent at aggregating diverse information. The price on predict.horse reflects the combined knowledge of speed figure analysts, pace handicappers, pedigree experts, clockers, insiders, and data modelers -- all trading on the same market. This makes prediction market prices extremely difficult to beat on average. In races with high liquidity and broad participation, the market price is a formidable estimate of each horse's true chances.

Where Traditional Handicapping Has an Edge

Traditional handicapping maintains an edge in specific situations. In lower-profile races with less market participation, prices are less efficient and your individual analysis carries more weight. In situations requiring subjective judgment -- reading a horse's body language in the paddock, evaluating a workout against the clock when the track was playing unusually fast or slow, assessing whether a horse's troubled trip was truly excusable -- experienced handicappers can identify factors that the market has not priced in.

The Optimal Combination

The most profitable approach uses prediction market prices as a baseline and handicapping analysis as the edge. Start by accepting the prediction market price as a reasonable estimate of each horse's chances. Then, apply your data-driven analysis to determine whether specific horses are overpriced or underpriced relative to your model's output. When your model and the prediction market agree, stand aside -- there is no edge. When they disagree by a meaningful amount (typically 3+ percentage points), trade aggressively in the direction your model indicates.

Building Your Own Handicapping Model

Building a data-driven handicapping model does not require a degree in statistics or advanced programming skills. It requires a systematic approach, the right data, and the discipline to trust the numbers over your gut.

Step 1: Define Your Variables

Start with the variables that research shows are most predictive: last three speed figures (and their trajectory), pace projection for the race, track bias indicators, trainer situational win rate, jockey win rate at the track, days since last race, and class level relative to competition. Each variable should be measurable and available before the race.

Step 2: Assign Weights

Each variable needs a weight reflecting its predictive importance. Speed figures and pace projection typically carry the most weight (combined 40-50% of the model). Trainer and jockey statistics account for 20-25%. Class, fitness, and track bias make up the remainder. You can start with these approximate weights and refine them based on your results over time.

Step 3: Generate Probability Estimates

For each horse in a race, run it through your model to produce a raw score. Convert the raw scores into probability estimates by dividing each horse's score by the total of all scores in the field. This gives you a probability for each horse that sums to 100%. Compare these probabilities to prediction market prices to identify value.

Step 4: Backtest and Refine

Before trading real money, backtest your model against historical results. Run the model on past races where you know the outcomes and measure its performance. The key metric is calibration: do horses your model prices at 20% win approximately 20% of the time? If your model is well-calibrated, you can trust its output. If it consistently overestimates or underestimates certain categories of horses, adjust the weights.

Essential Data Sources and Tools

Data-driven handicapping is only as good as the data it is built on. Here are the essential sources and tools for modern horse racing analytics.

Source	Data Provided	Best For
Equibase	Official past performances, results, entries	Foundation data for all analysis
Daily Racing Form	Past performances, Beyer figures, workouts	Traditional handicapping baseline
TimeformUS	Speed/pace figures, pace projections, bias	Pace analysis and bias detection
Thoro-Graph	Proprietary speed figures with trip adjustments	Identifying visually impressive but statistically modest runs
Brisnet	Data files, speed/class ratings, pedigree data	Model building and data analysis
TrackMaster	Trainer/jockey stats, workout data, bias	Situational trainer/jockey angles
predict.horse	Prediction market prices and volume	Market consensus and value identification

Combining Analytics with Prediction Markets

The ultimate goal of data-driven handicapping is not just to pick winners -- it is to identify situations where the probability of an outcome is different from what the market believes. This is the definition of "edge," and it is what turns analytical skill into prediction market profit.

Here is the workflow that top analytical traders use on predict.horse:

1. Run your model to generate probability estimates for each horse in the race.
2. Check prediction market prices on predict.horse and compare them to your model's output.
3. Identify discrepancies where your model and the market disagree by 3+ percentage points.
4. Validate the discrepancy by checking whether there is a specific reason (track bias, pace scenario, trainer pattern) that explains why the market may be wrong.
5. Execute the trade by buying YES shares on underpriced horses and (if available) selling or avoiding overpriced horses.
6. Track results over time to measure whether your model produces a positive ROI and to identify areas for refinement.

This systematic process removes emotion from the equation. You are not betting on hunches. You are trading on quantified edges that have been backtested against historical data. Over hundreds of trades, this approach compounds small per-trade edges into significant profit.

Putting It All Together on predict.horse

Data-driven handicapping and prediction markets are a natural combination. The analytical tools give you the edge. The prediction market gives you the mechanism to profit from that edge with continuous trading, real-time pricing, and the ability to enter and exit positions at any time.

Start by applying the methods in this guide to a small number of races. Focus on tracks and race types where you have the most data and experience. Build your model, test it against prediction market prices on predict.horse, and refine based on results. As your confidence and track record grow, increase your activity and expand to more race types and tracks.

The tools available to horse racing handicappers in 2026 are unprecedented. Speed figures, pace models, bias data, trainer statistics, weather analytics, and prediction market signals provide more information than any handicapper in history has ever had access to. The only question is whether you will use it.

For a broader perspective on using prediction markets effectively, read How Prediction Markets Work: The Complete 2026 Guide. For Kentucky Derby-specific strategies, see our Kentucky Derby 2026 Predictions: Complete Betting Guide. And for the historical patterns that repeat across Triple Crown racing, explore Triple Crown Predictions: Historical Analysis and Market Patterns Since 1919.