Data-Driven Course Strategy: Choosing Shots by Expected Score

What expected score measures

Expected score is a probability-weighted average of the strokes likely to be needed to complete a hole from a particular position. It answers a practical question: if this situation occurred many times, how many strokes would this choice cost on average?

For a shot with several possible outcomes, the basic calculation is:

Expected score = sum of (probability of each outcome × expected strokes to finish from that outcome).

The “outcomes” should describe useful landing areas, not just a binary distinction between success and failure. An approach might finish on the green, in a greenside bunker, in rough, or beyond the target. Each location has a different expected number of strokes remaining. The shot itself also counts: a player who hits an approach has used one stroke, whether it finishes on the green or in the rough.

This is related to strokes-gained analysis, which compares performance with a benchmark expectation for a given location. But a player does not need a tour-level model to use the underlying idea. A sensible, consistently collected personal dataset can help compare options—provided its uncertainty and limitations are kept in view.

A worked example: 150 yards to the green

Imagine a player facing a 150-yard approach. The flag sits near the right edge, with a bunker guarding that side. A broad section of green lies left of the pin. The player can aim at the flag with a club that reaches it comfortably, or choose a safer target toward the middle-left.

For a clear comparison, suppose the player’s past shots from similar lies and distances suggest the following illustrative outcome estimates. These numbers are an example, not a universal table: actual probabilities depend on the golfer, the lie, the conditions, and how the categories are defined.

For the aggressive target, the expected strokes remaining are (0.45 × 1.9) + (0.30 × 2.8) + (0.25 × 3.1), or about 2.53. Add the approach stroke and the expected total from the current position is about 3.53.

For the safer target, the calculation is (0.70 × 2.0) + (0.22 × 2.6) + (0.08 × 3.0), or about 2.23 strokes remaining. Including the approach, the expected total is about 3.23. Under these assumptions, the safer target is better by roughly 0.30 strokes.

The aggressive choice still has a plausible reward: its green outcomes include more chances to leave a short putt. But the bunker and other misses cost enough to outweigh that advantage in this example. The model does not say the aggressive shot is foolish or forbidden. It says that, for the assumed player and conditions, the average cost is higher.

Small differences should not be mistaken for certainty. If the estimates were based on a limited number of shots, a 0.30-stroke gap might be less convincing than the arithmetic makes it appear. The player should consider the quality of the data, the match situation, and whether the estimates truly describe the shot in front of them.

Why “hit the green” is not enough

A green-hit percentage can conceal important differences. Two targets may produce the same proportion of greens hit while leaving very different next shots. One might leave a manageable putt or chip; the other might produce a short-sided bunker shot and a narrow recovery margin.

Likewise, “missed green” is too broad to serve as a useful outcome category. A ball in fairway-height fringe is not equivalent to one in deep rough, and neither is equivalent to a penalty area. If the model treats them as the same event, it will misprice risk. The next-shot expectation is what gives a miss its strategic meaning.

It is also important not to count the approach twice—or omit it. In the worked example, the outcome values represent strokes remaining after the approach. The shot itself is then added once. Another valid system can calculate total strokes from the current position directly, but the convention must be consistent across every option being compared.

Build estimates from the player, not an imaginary average

A statistical model is only as relevant as the data behind it. A player’s dispersion pattern may be asymmetric: perhaps the common miss is short-right, while a left miss is rare but severe. A single “shot radius” around the target can hide that difference. Computer vision and launch-monitor systems can help record start direction, carry, and landing location, but a useful strategy model also needs context: club, distance, lie, target, and conditions.

For a coach or player collecting data, begin with a manageable set of categories. Record the club and approximate distance, the lie, the intended target, the landing zone, and the quality of the next-shot position. Add wind and surface conditions when they are likely to matter. Do not assume a sophisticated sensor automatically produces a strategically complete dataset; a precise carry measurement cannot, by itself, tell you whether the resulting bunker shot is difficult for that player.

Comparable samples matter more than a large pile of mismatched shots. A 150-yard shot from a clean fairway lie is not interchangeable with a 150-yard shot from a downhill lie in rough. If the player has only a few relevant observations, use broad ranges and label them as provisional rather than presenting a finely calculated result as fact.

For each outcome area, estimate the strokes needed to finish from there. This can be done from the player’s own records, a carefully selected benchmark, or both. Personal data may be noisy, especially for uncommon events; benchmark figures can provide context but may not reflect the player’s skill or course. The goal is not to manufacture precision. It is to make assumptions explicit enough that a coach can challenge them.

A practical workflow for comparing shots

1. Define the decision. Write down the actual alternatives: for example, aim at a right-side flag with a 9-iron, or target the broad left-center area with an 8-iron. “Be aggressive” and “play safe” are not precise enough.
2. Describe the landing areas. Mark green zones, bunkers, rough, penalty areas, and other meaningful destinations. Group areas only when their next-shot costs are reasonably similar.
3. Estimate probabilities. Use the player’s relevant shot pattern to estimate the chance of each outcome. Check that the probabilities for each option total 100%.
4. Estimate strokes to finish. Assign an expected number of strokes remaining to each landing area, including the likely putt, chip, bunker shot, or recovery. Be consistent about whether penalties are included in the outcome cost.
5. Calculate and compare. Multiply each outcome probability by its expected strokes remaining, add those products, then add the shot if the calculation starts before it. Compare like with like.
6. Test the assumptions. Change uncertain inputs within plausible limits. If a small change reverses the answer, treat the recommendation as a close call, not a command.

Wind, firm greens, and changing targets

Conditions change both the probability of an outcome and the cost of reaching it. A headwind may reduce carry and increase the chance of finishing short. A crosswind may shift the landing distribution toward a bunker or penalty area. Firm greens can make a front target less receptive and increase the chance that a shot landing safely on the surface releases into trouble.

That does not mean every weather detail needs a bespoke model. For a practical decision, ask which assumption is likely to move. If the wind is expected to push the player’s common miss toward the water, the probability assigned to that outcome should rise. If the green is firm and fast, the expected value of a shot that lands near the flag may change because the follow-up putt or chip becomes harder.

Target selection can also alter the distribution. The player might aim away from trouble, not because the swing suddenly becomes more accurate, but because a familiar miss now finishes in a more playable area. This is one reason course strategy is not just a club-selection problem. It is a choice about where to place the full pattern of likely outcomes.

From a single approach to a hole-by-hole plan

Expected-score thinking can be applied across a hole. From the tee, compare targets by estimating the next-shot positions they create: fairway, rough, bunkers, trees, or a penalty area. Then evaluate the approach from each likely position. A long drive into a narrow landing area may offer a shorter second shot but also carry a higher chance of a difficult recovery. A shorter club to a wider corridor may keep the expected score lower.

On a par five, the same framework helps assess whether to attempt a shot toward the green or lay up. The comparison is not simply “two shots to reach” against “lay up and wedge.” It is the expected number of strokes from each option, including the chance of a miss into trouble and the quality of the next position. A lay-up that reliably leaves a comfortable wedge distance can beat a low-probability attempt that often creates a recovery problem; a player with a strong long-game pattern may reach the opposite conclusion.

Plans should therefore be conditional rather than rigid. A useful hole plan might say: from the tee, favor the left-center landing area; if the drive finishes in the right rough, prioritize a playable approach rather than forcing the original target; if the lie and wind are favorable, reconsider the more aggressive line. Such a plan connects statistical expectations to decisions a player can actually make.

Risk is not the same as score

Expected score describes an average, not the full shape of the possible results. Two options can have equal expected scores while one produces mostly routine outcomes and the other alternates between birdie chances and costly mistakes. A player protecting a lead, playing a match, or pursuing a comeback may reasonably care about those different distributions.

In match play, for example, the value of a risk can depend on the opponent’s position. A player who needs to win the hole may accept more volatility than one who can secure a half with a conservative target. In stroke play, a tournament position or a late-round need may also influence the acceptable spread of outcomes. These considerations do not make expected score irrelevant. They show that the decision objective may include more than average strokes.

Coaches can make this distinction concrete by showing both the expected score and the outcome distribution: the chance of a penalty, the chance of a strong scoring opportunity, and the chance of a routine result. That gives players a better basis for choosing than an unexplained label such as “high risk.”

Where models go wrong

• Using an unsuitable comparison group. A tour player’s average outcome may not describe a club golfer’s bunker play or approach dispersion. Benchmarks are references, not automatic instructions.
• Ignoring selection effects. Players may choose aggressive targets only when conditions are favorable. Their historical results for those shots may not transfer to a different wind, lie, or pressure situation.
• Overlooking penalties and recovery costs. If a hazard outcome is recorded without the penalty stroke and subsequent position, the model understates its cost.
• Creating too many categories. A detailed map looks authoritative, but tiny samples in each zone produce unstable estimates. Start with categories that change the next-shot decision.
• Treating a recommendation as certainty. A model estimates from data and assumptions. It does not know whether the player is injured, uncomfortable over the ball, or dealing with a gust that appeared seconds ago.

Use the calculation as a coaching conversation

Numbers are most valuable when they sharpen a question. If the model favors the middle-left, ask whether the player can execute that target with the selected club. If the aggressive line appears better, ask whether its estimated miss pattern is based on a representative sample or a few memorable shots. If two options are nearly tied, ask what matters in the situation: the lowest average score, avoiding a penalty, or creating a chance that matches the player’s current objective.

After the round, compare the plan with the actual decision and result. Do not judge the choice only by whether one ball finished near the flag. A good decision can produce a poor outcome; a weak decision can get lucky. Review several comparable situations, then update the estimates when the evidence supports a change. This separates process from outcome and gives practice a clear purpose: improving a specific part of the player’s probability profile, such as reducing short-right misses or making bunker recoveries more reliable.

Expected-score models are not a substitute for judgment. They are a way to make the hidden trade-off visible. Instead of asking which shot looks boldest, longest, or safest, a player and coach can ask what each option is likely to leave behind—and what that position will cost next.

A quick checklist before choosing

• Have you defined the actual target and club for each option?
• Are the player’s shot-pattern estimates relevant to this lie, distance, and condition?
• Do the outcome categories distinguish playable misses from costly trouble?
• Does each outcome include a realistic estimate of strokes to finish?
• Have you included the approach stroke and any penalty strokes exactly once?
• Would a reasonable change in uncertain assumptions reverse the decision?
• Does the player’s situation call for a low average, a lower chance of disaster, or a more volatile result?

A sound strategy does not require perfect data. It requires transparent assumptions, estimates grounded in the player’s own game, and enough humility to recognize when the numbers cannot separate two choices with confidence.