The Model
How the Tour Lens model works
Every week, Tour Lens runs 10,000 Monte Carlo simulations of the upcoming PGA Tour event. Here's exactly what goes into the model, what comes out, and how to use the results to find real edges in golf betting and DFS.
Model track record — 2026 season
Ranked Matt Fitzpatrick #3 at the Valspar Championship at +1800 odds. He entered Sunday three shots back, birdied 18, and won by one stroke. The model identified him as the best course-fit value in the field days before the tournament began.
Full record →What is Monte Carlo simulation?
Monte Carlo simulation is a mathematical technique that runs a scenario thousands of times with randomized inputs to produce a probability distribution of outcomes. It's used in finance, engineering, weather forecasting — and golf analytics.
For Tour Lens, "running the tournament 10,000 times" means this: for each simulation, every player in the field draws a performance score based on their skill profile plus a random variance factor. The 72-hole scores are tallied, the lowest wins, and we record the result. After 10,000 runs, the percentage of times each player wins becomes their win probability.
Why 10,000 simulations?
Golf has enormous variance. A single simulation would just pick whoever has the best skill rating — which isn't realistic. Running 10,000 simulations lets variance play out the way it does in real life. A player rated 5th best might win only 8% of simulations because the other 92% of the time, variance works against them. That 8% figure is much more useful than a simple ranking.
The output isn't "who will win" — it's "what is each player's probability of winning given their skill, the course, and the inherent randomness of professional golf." That distinction matters enormously when you're trying to find value.
The four inputs
Each player's base skill rating is built from four components, weighted differently depending on the specific course being played.
Official World Golf Ranking
Foundation layerOWGR provides a global baseline of each player's ability across all conditions. It's the most stable input — less volatile than week-to-week stats but less precise for specific course types. It anchors the model so that elite players don't disappear from contention based on a few bad weeks.
Strokes Gained metrics
Primary skill signalSG data is the backbone of the model. We weight four SG categories differently for each course: SG: Off the Tee, SG: Approach, SG: Around the Green, and SG: Putting. A course that rewards precision iron play (like TPC San Antonio) weights SG: Approach at 38% while penalizing raw distance. Augusta National weights SG: Approach and SG: Putting near equally because both are essential to survive the back nine on Sunday.
Course-fit multiplier
Differentiating factorThis is where the model finds real edges. Course-fit is a multiplier applied to a player's base score based on how well their specific skill profile matches what the course demands. It's not just "has this player done well here before" — it's a structural analysis of whether their SG profile fits the course requirements. A player who ranks top-5 in SG: Approach playing a course where that's the dominant skill gets a significant boost regardless of their overall ranking.
Recent form
Momentum signalThe last 6–8 weeks of results carry meaningful weight, particularly missed cuts vs. top-25 finishes. A player who has made 7 straight cuts with two top-10s is treated differently from a player with the same season-long stats but a missed cut two weeks ago. Form matters most for players on the edge of contention — it rarely overrides elite ball-striking but often separates similar-ranked players.
Course-fit multipliers — where edges come from
The course-fit multiplier is the most important part of the model for finding value. It's also the hardest to explain simply, so here's a concrete example.
Copperhead Course at Innisbrook (the Valspar) is a precision placement course. The rough is severe, the greens are elevated and firm, and the winning profile historically skews toward elite ball-strikers who can hit fairways and stop the ball on fast greens. It's not a course where bombers dominate — it's where mid-iron excellence wins.
The Fitzpatrick example
Matt Fitzpatrick ranks consistently in the top 15 globally for SG: Approach and SG: Around the Green. He's not the longest off the tee but Copperhead doesn't penalize that — it rewards what he does best. His course-fit multiplier at the Valspar was 1.24 — meaning his performance score was multiplied by 1.24 in every simulation. At +1800 odds (implying ~5.3% probability), our model had him at 13% win probability. That gap between market price and model output is the edge.
Course-fit multipliers range from roughly 0.85 (poor fit) to 1.30 (exceptional fit). The biggest boosts go to players whose top SG skills directly match the course's highest-weighted requirements. The biggest penalties go to players whose style works against them — a pure bomber at an accuracy course, for example.
This is why the model sometimes disagrees with the betting market significantly. The market sets odds based on overall player quality. The model adjusts for the specific course. When those diverge, that's where the picks come from.
What the model outputs
After 10,000 simulations, Tour Lens produces four probability figures for every player in the field:
Percentage of simulations the player wins outright. The headline number — use this for outright bets and Kalshi win markets.
How often the player finishes in the top 5. Most useful for each-way betting, DFS captain slots, and Kalshi top-5 markets.
The most predictive metric for DFS lineup building. Players with high top-10 rates provide consistent points without needing to win.
Useful for identifying safe floor plays in DFS and for spotting players likely to make the cut and finish respectably.
The model also outputs an average finish — the mean finishing position across all 10,000 simulations. This is the cleanest single number for DFS salary value analysis: a player with an average simulated finish of 12 but a DraftKings salary priced like a top-30 finisher is an obvious value play.
Case study: Fitzpatrick at the 2026 Valspar
Here's how Fitzpatrick ended up ranked #3 despite being a relatively obscure pick at the time. Copperhead weights SG: Approach at 35% and SG: Around the Green at 23%. Fitzpatrick ranked in the top 10 globally in both categories entering the week. His course-fit multiplier was 1.24 — among the highest in the field.
The betting market had him at +1800 because it was pricing his overall world ranking (~22nd at the time) rather than his specific fit for this course. The model saw a player whose best skills aligned perfectly with what Copperhead demands.
He entered Sunday three shots off the lead, shot 68, birdied the 18th hole, and won by one stroke over David Lipsky. The model didn't predict he'd birdie 18 — it predicted he was much more likely to contend than the market implied. That's the job.
"The model doesn't need to be right about who wins. It needs to find players whose probability of winning is meaningfully higher than what the market is pricing. Do that consistently and you have an edge."
What the model can't do — and why that's honest
No model can predict winners consistently in golf. The sport has too much variance — the best player in the field wins roughly 20% of the time, and elite players miss cuts regularly. Anyone claiming to pick winners at a high rate is either lying or on a short lucky streak.
The model's value is in identifying structural edges — players whose skill profile fits a specific course better than the market acknowledges. That's a repeatable, systematic process. Predicting what happens on any given Sunday is not.
How to use the picks
Different use cases call for different parts of the output. Here's how to apply the model depending on what you're doing:
Kalshi / prediction markets
Compare our win % to the Kalshi market price. If our model gives a player 8% win probability but Kalshi is pricing them at 4%, that's a structural edge worth considering. The biggest edges come from players ranked 3rd–10th in our sim who are priced as longshots in the market.
DraftKings / FanDuel DFS
Sort by Top 10 % rather than Win %. For DFS you need players who will score points consistently — not necessarily win. Look for players in ranks 4–12 in the sim with high top-10 rates but mid-tier DFS salaries. That's where the salary value lives.
Outright betting
Focus on the gap between model win % and market-implied win %. Our sim's #1 pick isn't always the value — sometimes the market agrees with us. Look for our top-15 players priced at +3000 or longer where our model has them above 4% win probability.
Each-way / top-5 betting
Use the Top 5 % column. Players with 15%+ top-5 probability but long odds (implying sub-8% from the market) are consistent each-way value plays. The model's top-5 rate is historically its most accurate metric.
See the model in action
This week's full simulation is live — every player in the field ranked by win probability, top-5, top-10, and top-20 rates.