Golf Tournament Picks: Course Fit Meets Strokes Gained

The Thesis

Golf is the ultimate skill-decomposition sport. Every shot belongs to a measurable category — driving, approach, short game, putting — and every course demands a different mix of those skills. A bomber who overpowers TPC Scottsdale might struggle at Augusta's precision-demanding layout.

Most golf bettors look at world rankings and recent form. We look at whether a player's specific skill profile matches what the course actually demands.

What We're Predicting

We generate signals across five prop markets for PGA Tour events:

Each prop has a market-implied probability and our model probability. When they diverge, we have a signal.

The Model: Two Pillars

Our golf model rests on two pillars, blended together:

1. Strokes Gained (SG:Total) — Overall player quality (70% weight by default)
2. Course Fit — How well a player's skill profile matches the course (30% weight)

This 70/30 blend captures a key insight: the best golfers tend to perform well everywhere (SG:Total), but course-specific demands create meaningful deviations from baseline (Course Fit).

Pillar 1: Strokes Gained

Strokes Gained is the most important statistical revolution in golf analytics. Instead of counting birdies or looking at scoring averages, SG measures how many strokes a player gains or loses relative to the field on every single shot.

We break SG into four components:

SG: Total is the sum of all four. A player with SG:Total of +2.0 gains two strokes per round relative to the average PGA Tour player. Over four rounds, that's an 8-stroke advantage — the difference between winning and missing the cut.

We use SG:Total as our primary skill ranking because it's the single best predictor of future performance in golf. World rankings lag. Recent form is noisy. SG:Total captures true underlying skill.

Pillar 2: Course Fit

Here's where it gets interesting. Not all courses are created equal.

Augusta National demands precision — narrow fairways, dramatic elevation changes, lightning-fast greens with extreme slopes. SG:Approach and SG:Putting dominate.

TPC Scottsdale rewards bombers — wide fairways, reachable par 5s, less punishing rough. SG:Off the Tee is king.

Harbour Town is a short game test — tight fairways, small greens, demanding recovery shots. SG:Around the Green and SG:Approach carry the most weight.

We maintain course profiles for every PGA Tour venue, each with four skill weights that sum to 1.0:

course_profile = {
  sg_off_tee_weight: 0.15 - 0.45,
  sg_approach_weight: 0.20 - 0.45,
  sg_around_green_weight: 0.15 - 0.35,
  sg_putting_weight: 0.15 - 0.30,
}

The Course Fit Calculation

For each player at each tournament, we calculate a course fit score (0-100):

1. Normalize each SG component from its raw range (-2 to +2) to a 0-100 scale
2. Multiply by the course weight for that component
3. Sum the weighted scores and divide by total weight

For each SG component:
  normalized = ((sg_value + 2) / 4) × 100

fit_score = Σ(normalized_component × course_weight) / Σ(course_weights)

A fit score of 75+ means the player excels in exactly the skills the course demands. A score below 40 means there's a meaningful mismatch.

Real Example: Bomber at a Precision Course

Player: Big-hitting player with these SG components:

• SG:Off the Tee: +1.5 (excellent)
• SG:Approach: +0.2 (slightly above average)
• SG:Around the Green: -0.4 (below average)
• SG:Putting: +0.1 (average)

Course: Precision layout demanding short game:

• Off Tee weight: 0.15
• Approach weight: 0.25
• Around Green weight: 0.35
• Putting weight: 0.25

Fit calculation:

• Off Tee: ((1.5+2)/4 × 100) × 0.15 = 87.5 × 0.15 = 13.1
• Approach: ((0.2+2)/4 × 100) × 0.25 = 55.0 × 0.25 = 13.8
• Around Green: ((-0.4+2)/4 × 100) × 0.35 = 40.0 × 0.35 = 14.0
• Putting: ((0.1+2)/4 × 100) × 0.25 = 52.5 × 0.25 = 13.1

Course Fit: (13.1 + 13.8 + 14.0 + 13.1) / 1.0 = 54.0

Despite being a strong overall player (SG:Total +1.4), this player gets a mediocre course fit because his best skill (driving) is the least important this week, and his worst skill (short game) is the most important.

This is exactly the kind of mismatch the market tends to overlook. Bettors see a top-20 world ranking and assume a top-20 probability. Our model sees a skill-course mismatch and adjusts downward.

Probability Calibration

We convert our blended skill + fit assessment into probabilities for each prop type using calibrated base rates and z-score factors:

The base rate is the unconditional probability for a random field player. The z-factor controls how much our skill assessment shifts that probability.

Higher z-factors mean skill matters more. This makes intuitive sense — winning a golf tournament requires exceptional play (high z-factor), while making the cut is more about avoiding disaster (lower z-factor). The cream rises to the top for winner and Top 5, but the cut line is more of a noise filter.

Edge and Signal Generation

We compare our model probability to the market probability for each player-prop combination:

edge = model_prob - market_prob

Signals are generated at three confidence levels:

We generate both BUY signals (model probability higher than market — we think the player is underpriced) and SELL signals (model probability lower — we think the player is overpriced).

SELL signals are particularly valuable in golf because the market systematically overprices popular players and underprices less well-known names. The public bets on name recognition. We bet on skill-course alignment.

Configuration Presets

Our model supports multiple configuration presets for different strategies:

These presets let users tailor the model to their conviction about what matters at a given tournament. If you believe Augusta is purely a ball-striking test, switch to the Iron Players preset. If you think course history is overrated, try SG Purist.

The Dashboard

Our golf dashboard shows for each signal:

• Player name and ranking (OWGR, DataGolf, SG rank)
• Prop type and line (Top 10 at -120, etc.)
• Model vs. market probability — the core comparison
• Edge and confidence level — how strong the signal is
• Course fit score — 0-100% with best/weak skills highlighted
• SG component breakdown — OTT, APP, ARG, PUTT values
• Additional stats — Driving distance, GIR%, rounds played

We also maintain a course profiles page where you can browse every venue's skill weights, and a player stats page with sortable SG data across all components.

Performance Tracking

Signals are tracked and settled after each tournament:

• By prop type: Win rate and P&L for Top 5, Top 10, Top 20, Make Cut, and Winner signals separately
• Overall: Combined win rate, P&L, and average edge
• Historical: Daily performance snapshots for trend analysis

This matters because different prop types have fundamentally different characteristics. Winner signals are low-hit-rate but high-payout. Make Cut signals are high-hit-rate but lower edge per signal. Understanding these dynamics is key to sizing positions correctly.

What Makes This Work

1. Skill Decomposition

Most bettors think "he's the 5th ranked player, so he probably has a Top 10 chance." We think "his SG:Approach is elite and this course demands approach play, so his Top 10 probability is higher than his ranking suggests." Decomposing skill into components and matching them to course demands creates real edge.

2. Course Profiles

Maintaining detailed skill-weight profiles for every venue lets us identify mismatches the market misses. The public knows Augusta is hard. We know specifically which skills Augusta rewards, and we know which players have those skills.

3. Multi-Prop Coverage

Five prop types per player means we can find edge at different probability levels. A player might not be mispriced to win, but their Top 20 line might be way off. Covering the full range increases our signal volume without lowering quality.

Current Limitations

1. Weather and course conditions: We don't yet adjust for wind, rain, or pin positions. These can significantly affect which skills matter on a given day.
2. Field strength: Our probabilities don't explicitly account for the overall strength of the field, which varies week to week.
3. Form and momentum: SG data is based on longer time windows. A player who's suddenly putting lights-out this month isn't fully captured.
4. Course history: Some players consistently outperform at specific venues for reasons our SG model doesn't capture (comfort, experience, course management).

What's Next

• Weather integration: Adjusting skill weights based on forecasted conditions (windy = more driving emphasis)
• Field strength normalization: Scaling probabilities based on the quality of the week's field
• Recency weighting: Blending recent SG data more heavily for form detection
• Course history overlay: Incorporating venue-specific player performance as a model input

The Bottom Line

Golf tournament betting is a skill-decomposition problem. The market prices players based on name, ranking, and narrative. We price them based on whether their specific skill profile matches what the course demands this week.

SG:Total tells you who's good. Course fit tells you who's good here. The blend tells you where the market is wrong.

---

See live golf signals on the Golf Dashboard. Browse course profiles at Courses. Explore player stats at Players.