Medinah After Renovation: The Tee Shot Becomes a Probability Problem

Medinah After Renovation: The Tee Shot Becomes a Probability Problem

Hook

On my hard drive I hold shot-by-shot data from 14 rounds at Medinah Country Club, Course No. 3, recorded before the renovation. After the renovation, the matching number is zero. Not because I stopped logging — my daily work in Nagoya is still dissecting individual shots — but because no official round at this course has released shot-level data clean enough to feed a model.

I sat with that gap for a while. A gap in a table speaks, if you are willing to listen. It told me two things. First, we are about to see Medinah No. 3 after renovation under the pressure of the biggest team event in men's golf. Second, most of what is being written about this course still sits in the realm of assumption, not evidence.

I do not mind assumptions. I mind when someone sells an assumption at the price of evidence.

Context: What Changed at Medinah

Medinah Country Club sits outside Chicago, Illinois, and opened in 2026. Of its three courses, No. 3 is the championship layout: the 2026 U.S. Open won by Lou Graham, the 2026 U.S. Open won by Hale Irwin, the 2026 and 2026 PGA Championships, both won by Tiger Woods, and the 2026 Ryder Cup.

The 2026 Ryder Cup left a specific memory. Europe entered Sunday trailing 10–6 and won 14.5–13.5. Ian Poulter's putts on Saturday, Martin Kaymer's decisive putt on the 18th on Sunday, and Tiger Woods' miss on the final hole are the three fragments anyone who watched remembers. I remember them for a different reason: that was the last time in more than a decade that analysts had enough shot-level data to talk about Medinah No. 3 with numbers instead of memory.

Between those two markers, the course changed. The design group OCM Golf — Geoff Ogilvy, the 2026 U.S. Open champion, alongside Mike Clayton and Mike Cocking — handled the renovation of Course No. 3. Their philosophy has been consistent across projects: remove trees, restore width to the driving corridors, make the turf firmer and faster, and create multiple routes instead of a single correct line.

With the renovation complete, the published schedule brings the 2026 Presidents Cup here. That is the moment every prediction model for Medinah will have to be rebuilt from scratch — mine included.

Core: Four Variables Behind a Single Tee Shot

The tee-shot problem on a wide course is different in kind from the problem on a tight, tree-lined course. On a tight course, there is essentially one question: how do I keep the ball in the fairway. On a wide Medinah-style course, the question splits into two layers: where do I hit it, and from which angle.

Tour players solve the first layer with four variables they understand better than anyone outside the ropes.

Carry distance. On a stock shot, most tour players know exactly how far their ball flies, within roughly two to three yards. That number determines whether they dare carry a bunker.

Dispersion pattern. This is the most undervalued variable. Dispersion is not a circle around the target. It is a skewed ellipse, usually leaning hard to one side. A player who misses right should never aim at the left edge of the fairway, however much longer that line looks.

Distance to trouble. The same bunker at the same distance carries a completely different trade-off depending on whether the player trusts their carry number.

Club consistency. Driver and long irons disperse wider than short irons. A decision to hit 3-wood on a 450-yard par-4 can be tactically sound yet wrong in expectation if that club's dispersion exceeds the width of the safe zone.

Laid out as a table, the four-variable frame looks like this:

| Variable | What a tour player knows | Accepted error | Cost of ignoring it | |---|---|---|---| | Carry distance | Stock carry yardage | ±2–3 yards | Wrong target, ball in bunker | | Dispersion pattern | Shape of skewed ellipse | Not symmetrical | Line chosen against miss bias | | Distance to trouble | Distance to bunker or water they will challenge | Club-dependent | Risk mispriced | | Club consistency | Stability of each club | Wider on long clubs | Expectation misjudged |

These four are the physics. What remains, and what is more interesting, is the information layer.

When a hole is redesigned to offer multiple lines, the expected values of those lines converge. This is simple mathematics: if the aggressive line yields an expected 3.85 strokes and the safe line yields 3.87, the gap is 0.02 strokes — inside the error margin of almost any amateur model. At that point the edge stops living in strength or technique. It lives in the quality of information a player holds about themselves.

I call this the decision premium. On a narrow course it is close to zero, because only one correct option exists. On a wide course it grows, because at least two options are legitimate and only personal data can adjudicate.

The chosen line does not only affect the tee shot. It affects the second shot, and that is where strokes gained separates.

A drive from the left half of the fairway on a hole with a green tilting right produces a completely different attack angle than a drive from the right half. Same remaining distance, same club in hand, different expected-stroke distribution on the approach. If I look only at Strokes Gained: Off the Tee — the metric that values a drive against the field baseline — the two shots look identical. Only when I add Strokes Gained: Approach by angle do I see the real cost of choosing the wrong line.

That is a modelling error I have made before. In 2026, doing data analysis at Nagoya Grampus, I built a manual xG model from video and omitted the home-venue variable. I ended up wrong on 6 of the final 10 matchweeks. After sitting through the entire tape again, I understood something I now apply to golf: raw data is not enough, context is what decides.

A driving model that only counts fairways hit will repeat exactly that mistake. It will treat two drives that both find the fairway as equal, when in reality one leaves a good attack angle and the other does not.

There is a second layer of variables that public models almost always skip: turf conditions and wind by time of day.

Post-renovation Medinah is built firm and fast. On firm turf the ball runs further, meaning the same carry can produce different run-out in the morning and the afternoon. Chicago-area wind also tends to differ between morning and afternoon in both direction and speed. Which means an aggressive line at 8 a.m. can become a bad line at 3 p.m., on the same hole, for the same player.

I have paid for ignoring the time variable. In 2026, in the Japan versus Belgium round-of-16 match at the World Cup, I collected PPDA figures and concluded Japan was pressing effectively. I ignored the distance covered by Belgian players after the 70th minute. Belgium came back to win 3–2 through vast space in midfield. I publicly criticised myself on my own page and admitted the model lacked a real-time fitness variable.

Since then, every analysis I write must include an intensity chart split into time bands. In golf, that chart is wind and turf moisture by hour. Without it, I do not conclude on club selection.

So how do you test a hypothesis about driving lines when shot-level data does not yet exist? I use three steps.

Step one, rebuild the hazard map from satellite imagery and public design drawings, identifying the zones where a ball can come to rest.

Step two, assign each zone an expected-stroke distribution based on historical data from holes similar in length, width and green orientation.

Step three, run a Monte Carlo simulation across dispersion profiles of different player groups, and see which line wins at which dispersion threshold.

This method does not give me the right answer. It tells me what the answer depends on — and that is usually more valuable information.

I do not have an independently confirmed completion date for the renovation, and I do not have official technical drawings from OCM Golf. Those are two limits I must state before anyone cites this piece. Data is never wrong, only my question was — and in this case, the right question cannot yet be answered with numbers.

Contrarian: The Fairway-Hit Trap

Most commentary on the Medinah renovation revolves around one idea: wider fairways mean more options, therefore better golf. I have no aesthetic objection. I object to turning it into a technical conclusion before there are numbers.

What did NOT happen often tells more truth than what did. Here, what did not happen is a fall in fairway-hit percentage on a wide course. That sounds paradoxical, but it can absolutely happen — and if it does, it is a signal that players are deliberately choosing braver lines rather than playing safe. On a wide course, keeping the ball in the fairway becomes easier in probability terms but less valuable in scoring terms, so a rational player accepts more risk.

The result is that fairway-hit rate can fall while the hole's average score improves. That is why I never use fairway-hit rate as a dependent variable in any model. It is an intermediate variable, not an outcome variable.

A second trap is assuming width always benefits the long hitter. In theory, longer players gain more when space opens up. But if the whole field hits it long and takes the same line, that edge is flattened. The decision premium only appears when there is variation in personal data — that is, when a player knows their dispersion well enough to choose a line others will not.

This is where the football comparison must be bounded. I have written about gegenpressing as an analytical lens, but I only keep it when the corresponding metrics agree. In golf, the closest equivalent to gegenpressing is the ability to recover position after missing the fairway. If data showed that par-save rate from thick rough at Medinah matched the rate from the fairway — entirely possible on a treeless course — then the whole notion of a safe driving line would lose much of its meaning. I do not have that data. So I do not conclude. Gegenpressing does not break the data, it breaks my assumptions — but only when there is data to break.

Finally, one group worries me more than any other when we talk about wide courses: young players. An open driving corridor is an invitation to pull driver and swing out of your shoes. For a 17-year-old with unstable dispersion and an unfinished body, that is a recipe for injury and for bad technical habits repeated thousands of times. The width of the course creates no problem. Width plus missing personal data creates the problem.

Takeaway: Three Signals I Will Track

The 2026 Presidents Cup at Medinah will be my first chance to test these assumptions against shot-level data in team-match conditions.

Three things I will track, and I recommend anyone doing analysis track them too.

First, the line chosen against each player's historical dispersion. If a player picks a line outside their 80 percent dispersion envelope, that is an emotional decision, not a data decision.

Second, distance to the green from the left and right halves of the fairway on holes with tilted greens. The gap here is a direct measure of the value of attack angle.

Third, the correlation between pairings and driving profiles. If captains pair players by driving profile rather than world ranking, that is a signal they have read the course correctly.

I do not believe in luck; I believe in cultivated probability. Post-renovation Medinah, if it is what OCM Golf claims, is a course designed so that the reward goes to whoever prepared better rather than whoever hits it harder. That is a testable claim. And I will be there, with the spreadsheet already open.