Shot Pattern Intelligence Guide

Shot Pattern Intelligence connects your own recorded rounds to an AI assistant such as Claude or ChatGPT. This guide defines every number the connector can show: the data behind it, the windows and comparisons it uses, the statistical signal, the benchmarks, and each tool's tables and columns. It is written for players, for coaches, and for the assistant itself.

What Shot Pattern Intelligence is

Shot Pattern records every shot you hit on the course with GPS, from the Apple Watch, iPhone or Android app. Each shot is priced in strokes gained against a benchmark tier you choose. The connector gives an AI assistant read access to that record through a small set of tools, together with the analysis engine that already does the statistics. The assistant narrates; the engine computes.

What the assistant can see:

• Your complete rounds: date, course, tees, score, strokes gained by category.
• Every hole: par, yardage, score, putts, penalties, flags such as a three-putt or a two-chip.
• Every shot: where it started (lie and distance to the hole), the club, how far it went, where it finished relative to your aim point, and its strokes gained.
• The engine's analyses: a ranked list of where strokes go, evidence tables, club distances, a hole-by-hole course report and hole-sequence patterns.
• Your benchmark tier, bag and units.

What it cannot see or do:

• No lie quality, wind, elevation, slope, swing or ball-flight data. Nothing on this page infers any of them.
• No course maps, hole shapes or coordinates in any result.
• No raw export. Individual shots only appear in one capped list, as evidence.
• No writes. The connector cannot change your rounds, your bag or your benchmark tier.
• One player at a time: the signed-in player, or a player on a coach's roster.

How to get good answers

The assistant works best when it is pointed at a window of rounds and a question the engine was built to answer. A few habits make the difference.

1. Start with your profile. The first call in every conversation is get_player_profile. It lists your rounds, courses, benchmark tier and units, and it tells the assistant what windows are possible. If the assistant skips it, ask for it.
2. Name the window once. "My last 20 rounds" is the default. Other windows you can ask for: "since my lesson on 21 August", "the 2025 season against 2026", "tournament rounds only" (round types exist since 2025-09-26), "18-hole rounds only", or "the rounds at my home course". Every tool accepts the same filters, so one window keeps every number on the same footing.
3. Say where you play. Dates are labelled in your time zone when the assistant knows it. Tell it your home course or your zone once.
4. Ask for the ranking first, then the evidence. "Where am I losing strokes?" produces the leak report: the gap between your better and worse rounds, split by category, with a ranked list of specific situations. Follow with "show me the evidence for the top one" or "which holes were the doubles".
5. Ask for the benchmark beside every rate. A make rate or a GIR figure means little alone. The tables carry the benchmark column; the assistant should quote it.
6. Respect "not enough data". Every comparison carries a signal: clear, likely, no real difference or not enough data. A pattern resting on four shots is not a finding, and the assistant should say so.
7. Ask for a different comparison. The default split is your best 40 % of rounds against your worst 40 %. You can ask for rounds under 80 against 80 and over, one round against the rest, since a date, the last five against the earlier ones, or all rounds pooled against the benchmark.

Questions it answers well

• "Where am I losing strokes?" The leak report, then the doubles.
• "Give me a full review." Profile, stats summary, leak report with two drill packs, the double-bogey holes, club distances.
• "I have a lesson tomorrow." What changed since the last lesson, the one or two leaks worth the hour, and the numbers a coach will ask for.
• "How far do I hit my 7-iron?" or "Where are my gaps?" Club distances with dispersion and a confidence tier.
• "How do I compare to a 5-handicap?" The stats summary against that tier's benchmarks.
• "Walk me through Saturday's round." Hole by hole with the costliest shot on each hole.
• "Which holes cost me at my home course?" The course report.
• "What happens after a bogey?" or "Why is my back nine worse?" Hole sequences, par-matched.
• "Build me a winter practice plan." The leak report pooled over the season, prioritised by strokes per round, plus club distances.

Working with your coach

Shot Pattern Intelligence is built to support a coach, not to replace one. The engine shows what is happening and how much it costs; a coach sees why. Share the conversation, or the fact sheet and why pack the assistant produced, before a lesson. The numbers a coach will ask for are the ones the "I have a lesson tomorrow" recipe gathers: what changed since the last lesson, the one or two leaks worth the hour, club distances on the same window.

A coach who has a player on their Shot Pattern coaching roster can run every tool on that player's data from their own account, exactly as the app's Coach Center shows it, by naming the player; every result is then marked "Coach view". The roster is managed in the app, not through the connector.

What it will not do

• It will not tell you why a shot went wrong. The data has no swing, contact or ball-flight information, so cause is a coach's call. The assistant may name what the data can show, such as "a one-club gap between your 7-iron and 8-iron", and no more.
• It will not plan a course hole by hole. Hole-by-hole strategy is not part of this version; the course report shows which holes cost you there.
• It will not recommend products or equipment.
• It will not compute its own statistics from raw shots when an engine tool answers the question. If you want a number the tables lack, ask for a slice, and it will use slice_shots.
• It will not sweep courses or run bulk loops. The connector works on one player's rounds and is capped accordingly.

Eight rules for reading the numbers

1. Strokes gained is the distance to your target tier, not a grade on your current play. Every shot is compared with what a golfer at your benchmark tier would expect from the same spot. A 9-handicap whose tier is Scratch will show a large negative total. That measures the gap to the goal.
2. Positive is gained, negative is lost. Never read the sign the other way. Within about ±0.10 per round is neutral.
3. A rate without its benchmark is half a number. Tables carry the benchmark beside your figure; the tier is named once.
4. The signal column decides what is a finding. "Clear" and "likely" come from a real test. "No real difference" and "not enough data" are not findings, whatever the raw numbers look like.
5. Weigh frequency as well as size. Four bad shots in 25 rounds are rarely a priority. Every table carries its sample sizes.
6. On the greens, speed first. The speed-control table has its own verdicts. Only when speed is fine does start line or face come up, and that needs a coach's eyes.
7. Two-chips and three-putts are not an aggression problem by default. They may be skill. The data cannot show which.
8. The engine's thresholds are the definitions. "Well short" means a shot that travelled under 90 % of the distance to its target. A "big miss" off the tee is strokes gained of −0.7 or worse in the tee tables but under −0.3 in the fact sheet's stored counter. When two figures look alike, check which definition is being quoted.

The data

Rounds, holes and shots

A round is one scorecard: date, course, tees, score, number of holes played and the app's stored strokes gained by category. Only complete rounds are available to the connector: rounds the player finished in the app. A round in progress is invisible until it is finished.

A hole carries par, yardage from the tees played, the score (strokes including penalties, as the app recorded it, never re-derived from shot counts), the putts taken, penalty strokes and three flags the app sets when the hole is played: three-putt, two-chip and short-sided.

A shot carries the lie it started from, the distance to the hole at address, the club, how far it travelled, the lie it finished in, the distance to the hole afterwards, where it finished relative to the aim point (sideways and along the line), any penalty strokes, its category and its strokes gained. Start and finish coordinates exist in the app but never leave it through the connector. Shots recorded by the watch that the phone has not yet enriched are left out.

Per 18 holes. Every "per round" figure is normalised to 18 holes using the holes actually played, so a nine-hole round counts as half a round.

Hole flags and outcomes

Shot categories

Every shot belongs to one of four categories, which are also the four strokes-gained categories. The app assigns the category when the shot is recorded:

Par-3 tee shots are approaches everywhere in Shot Pattern, so driving figures never include them.

Lies

A lie is one of: tee, fairway, rough, deep rough, bunker, recovery, green, in the hole, penalty. "Recovery" is a lie the player marks when the next shot is about escaping trouble rather than advancing; "penalty" is a finish in a penalty area or out of bounds. The trouble lies used by the approach-miss buckets are penalty, recovery, deep rough and bunker.

When a tool asks for a lie group it uses four groups: fairway (the fairway and the tee), rough (the rough only), bunker, and other (deep rough, recovery and anything else, including a missing lie). any takes every lie. Deep rough sits with recovery, not with rough: the scoring and proximity expectations from deep rough are far closer to a recovery lie than to the rough, so the two are never pooled.

Aim point, offline and short or long

Sideways misses (offline) and short/long are measured against the line from the shot's start to the player's aim point, not the pin.

• On approaches the aim point is the target the player set in the app, which is often but not always the pin.
• On short-game shots and putts the target is set to the pin automatically, so proximity there is distance to the pin.
• Offline is negative to the left of the aim line and positive to the right, as the golfer sees it looking at the target. The data carries no handedness.
• Short and long are read from the shot ÷ target ratio: how far the ball travelled divided by the distance to the aim point. Under 0.90 is "well short", over 1.10 is "well long".
• Degrees offline is the standard deviation of the miss angle, atan(offline ÷ shot length), in degrees. It is the main accuracy figure because it does not depend on how far the club goes. Corridor width in yards is context.
• Proximity is the straight-line distance from where the ball finished to the aim point; on short-game shots and putts, to the pin.

Units, dates and time zones

Distances are shown in the player's units: yards or metres. Putts are shown in feet or metres. The engine computes in yards and converts for display. Round and shot dates are shown as YYYY-MM-DD in the time zone the assistant passes; without one, UTC. Date filters such as "since 2026-08-21" start at local midnight in that zone, so the label and the filter agree.

Round types (Casual, Tournament and so on) have been recorded since 2025-09-26. Earlier rounds show "–" and are never matched by a round-type filter.

Windows and splits

Windows

A window is the set of rounds a call analyses. Every tool that analyses rounds takes the same filters.

One call analyses at most 400 rounds. When a window is loading for the first time the tool answers "still loading" and asks to be called again in a few seconds; the assistant should say so rather than guess. Loaded windows are cached for ten minutes.

Splits

The analysis tools compare two groups of rounds. The split says how the groups are formed.

Reading "A → B"

In a two-group result, "A → B" reads first group → second group, and every figure is per 18 holes. "13.7/rd → 15.2/rd" means the first group averaged 13.7 per round and the second 15.2. The first group is the better or subject group: the best rounds, the rounds under the threshold, the single round, the recent rounds, the rounds since the date. A candidate's strokes lost is the first group's SG per round minus the second's for best-versus-worst and threshold splits, and the baseline's SG minus the subject's for the subject splits, so a positive figure is always strokes the worse or subject group lost. Pooled windows compare one group against the benchmark, and strokes lost is simply the strokes lost to the tier.

The default split needs 5 rounds. Under that, the window is the latest round against the others, or, with one round, pooled.

The signal column

Every comparison carries one of four labels. They come from a statistical test on the rounds, holes, shots or putts behind the two figures, not from the size of the difference.

Which test: rates (make %, GIR %, fairways) use a two-proportion test and need 5 observations in each group; per-round counts, strokes gained per round and per shot, and scores to par use Welch's t-test on the per-round (or per-shot) values and need 3 in each group. Medians, counts and "x of y" cells are never tested. In a pooled window, SG per round is tested against zero.

Other minimums: a row needs 5 observations before it appears in a why-pack chain or as a fine neighbour (8 for a neighbour); a hole-sequence cut needs 20 holes; a speed verdict needs 10 putts; a course-report hole needs 5 rounds. When a tool prints "not enough data (n)", n is the sample it had.

Benchmarks and strokes gained

The tiers

Strokes gained is priced against the player's benchmark tier: Pro, Female Pro, Scratch, 5, 10, 15 or 20. The tier is chosen in the Shot Pattern app (Account → "Strokes Gained Handicap") and the app re-prices past rounds when it changes.

• Benchmark rates in the tables (make %, GIR %, proximity, dispersion) are the tier's figures for the same band, averaged over that tier's own mix of distances, not read at the band's midpoint.
• The SG total is the distance to the goal. The sum of fixable weaknesses is the potential saving and may exceed it, because strengths offset weaknesses. The two are kept apart.
• "Plays like" phrasing uses the closest tier for one metric: "your driver dispersion plays like a 15".
• The benchmark tables themselves are not published here or through the connector. Only the benchmark figures a tool prints beside the player's own number are available.

Benchmarks are reference points, not standards.

Benchmark fit

When a player loses 10 or more strokes per round to the tier on average over the last 20 rounds, get_player_profile prints a Benchmark fit line. A tier that far away does not guide practice; benchmarks work best one step beyond the player's level. The fix is in the app: Account → "Strokes Gained Handicap", then the tier just beyond the handicap index (0–4.9 → Scratch, 5–9.9 → 5, 10–14.9 → 10, 15–19.9 → 15, 20+ → 20). The connector cannot change it.

Tool reference

Eleven read-only tools. Every analysis tool takes the window filters above; the comparison tools take the split arguments too. Coaches pass student_id to read a player on their roster.

get_player_profile

The first call in a conversation. Prints the player's name, access, units, benchmark tier, active bag, how many complete rounds are available with their date range, the courses played (course id, name, rounds), today's usage against the limits, and the Benchmark fit line when it applies.

list_rounds

Complete rounds, newest first, from the round records alone, so it is cheap at any size.

The note above the table says how many rounds matched and which offset fetches the next page. Page size: last_n when given; otherwise every matching round with a date filter, else the newest 20; at most 400 per call.

get_round

One round hole by hole. The title carries the course, the date (in the time zone the assistant passes, UTC without one) and the tees; the line under it the score, par, holes played, the app's stored strokes gained by category and the benchmark tier the round was priced against.

get_stats_summary

The report statistics for the window with the comparison to the benchmark tier: the same numbers, from the same code, as the Stats report in the Shot Pattern app and the coach report. Numbers only; for why, use the leak report. Distances are shown in the player's units (yards or metres, green-side distances in feet or metres); the band labels keep the report's own yards and feet. The header line says which. A closing note repeats the three caveats below that most often mislead: how approach bands are keyed, that driving rows are driver-only, and what insufficient_data means.

Header

"Rounds N" counts the rounds with a valid score; "holes H" sums the holes played over every round of six holes or more. A window with nine-hole rounds therefore shows more holes than 18 × N. "Per 18" figures divide by holes played, so a nine-hole round counts as half a round.

Scoring

Average score and average to par are over the window's 18-hole rounds (rounds of six holes or more when there is no 18-hole round); best and worst are the lowest and highest scores among them. Scores include penalty strokes. Birdies, pars, bogeys and doubles-or-worse per 18 are the share of each outcome among all scored holes, scaled to 18.

Strokes gained per round vs benchmark

Total, driving, approach, short game and putting: the app's stored strokes gained for each round, summed over the window and scaled to 18 holes (hole-weighted, not a plain mean of rounds). "Driving" is the Tee category, "short game" is Around the green. A category that nets to exactly zero prints "–".

Driving vs benchmark

Every row but one is driver only: the first club in the bag whose name contains "Driver". Players with no such club get no driving table.

Par-3 tee shots are approaches, not tee shots, everywhere in Shot Pattern.

Approach by distance

Approach shots that start on the tee or in the fairway, with the aim point within 30 yards of the pin (deliberate lay-ups are excluded). Approaches from the rough are not in this table. Bands are half-open, lower bound inclusive: 50-75, 75-100, then ten-yard steps to 200, then 200-225, 225-250, 250-275, 275-300 yards. A shot is placed in the band of the distance it travelled, while the benchmark is read at the band's midpoint as a starting distance. Benchmarks stop at 225 yards; longer bands show "–".

Short game 0–25 yds by lie

Around-the-green shots that start within 25 yards of the pin, by starting lie: fairway, rough and bunker (shots from deep rough, recovery lies or the tee are not in this table). A lie with no shots in the window is not listed. No benchmark columns.

Putting make rates vs benchmark

All putts, by distance to the hole at address. Bands own whole displayed feet, so the edges sit on the half foot: 0-3' is under 3.5 ft, 4-6' is 3.5 to 6.5 ft, 7-9', 10-15', 16-24', 25-40' and 41+' follow the same rule. make % = putts holed ÷ putts in the band. bench % is the tier's make rate for the whole band, weighted by how often putts of each length inside the band occur, not the rate at the band's midpoint.

Three-putt rates

For the four bands from 10 feet out. "of" counts first putts in the band; "3-putts" counts those holes where three or more putts followed, as the app flagged them; "rate %" is the one over the other. bench % is adjusted to the player's own mix of first-putt lengths within the band, so a player whose 25-40' first putts are mostly 38 feet is held to a 38-foot standard, not a 32-foot one.

Doubles

A double is a hole scored at two over par or worse. The five causes are evaluated independently, so one hole can carry several causes or none:

Closing line

Putts per 18 and SG putting per 18 are computed from the putt records over the rounds that contain at least one putt. The up-and-down percentage on this line is labelled "from the rough": it is the rough row of the 0–25 yard table, not an all-lie figure.

How "read" is decided

Every benchmark comparison carries a read. diff is always player minus benchmark, whatever the direction, so on a lower-is-better metric a negative diff is good.

get_club_distances

Per club, from the player's own full swings.

Only tee shots and approaches that start on the tee or in the fairway, with a club and a recorded distance, count; a club needs 5 qualifying shots to appear, and the clubs below that are listed by name with their counts. Shots marked "ignore for dispersion" in the app (a toggle for tree and cart-path bounces) are excluded. The lateral columns need 5 shots with a sideways miss recorded. sources limits the shots to tee shots, approaches or both. Clubs print in bag order with the app's labels; a club-less putt reads "Putter" wherever putts are listed.

build_leak_report

The analysis engine's answer to "where am I losing strokes": a fact sheet for a window under a split, then a why pack of evidence tables for each of the top candidates (drill_packs, 0 to 4, default 2). Ask for one pack later with candidate_id. The result opens with the split and its groups, the window note and any warnings (a group under 5 rounds is flagged as thin; shots without strokes gained are counted and left out of SG figures).

The fact sheet

FACT SHEET (every figure is per 18 holes; "A → B" reads first group → second group)
Groups: "Best 40%" = 8 rounds (144 holes), average +1.5 to par; "Worst 40%" = … Split: Best 40% vs worst 40%.
Gap: 10.5 strokes per round to par (Best 40% +1.5 vs Worst 40% +12.0)
By category (SG per round): Putting +0.27 → −3.14 (gap 3.4, 37% of it); Approach play …; Around the green …; Off the tee …
Round counters: Greens in regulation 11.8 → 8.2/rd (clear); Fairways hit 6.2 (49%) → 5.8 (41%)/rd (no real difference); …

CANDIDATES (ranked by strokes lost per round × how often the situation comes up):
1. [putting.byDistance#4-6ft] Putts from 4–6 ft — Best 40%: 2.6/rd, 76% made, +0.30 SG/rd · Worst 40%: 4.2/rd, 53% made, −0.64 SG/rd. Strokes lost: 0.94/rd (10% of the gap). Signal: SG per round: likely; made: likely. Split: more of them 0.03, played worse 0.91. Value per swing 0.22. Fine nearby: … Related: Three-putts 0.8 → 2.2/rd (clear); … Tag: highest frequency.

Candidates

A candidate is one row of one analysis table that clears the bar: the situation comes up at least once every other round, costs at least 0.05 strokes per attempt, and the gap is positive. Six tables feed candidates: every putting band, every short-game lie-and-distance cell, the four short-sided chip rows, four of the five approach-miss buckets (not "average"), approaches inside 175 yards from the rough as a whole, and off the tee as a whole. Candidates are ranked by strokes lost per round × attempts per round, keeping one per family (a short-sided chip row and the matching lie-and-distance row are the same family), at most two per section, up to six in all.

With a single-round split the detail reads as counts in that round ("4 shots, 50% made, −0.80 SG") against the other rounds' per-round figures, and strokes lost is the strokes lost in that round. With no candidate the sheet says so: "none met the bar".

Why packs

WHY PACK for "Putts from 4–6 ft" (Best 40%: 2.6/rd, 76% made, +0.30 SG/rd · Worst 40%: 4.2/rd, 53% made, −0.64 SG/rd)
Chain (computed; first group → second group; the word in brackets is how much to trust that line):
- 4–6 ft (Putting by distance): Attempts/rd 2.6 → 4.2, SG/rd +0.30 → −0.64, make % 76% → 53%, Scratch make % 68% [likely]
- …

Tables:
[table id: putting.byDistance]
### Putting by distance
Question: Which putts separate the groups?
| … |
n: Best 40% putts=202, Best 40% rounds=8, …
Note: …

The chain is every row of the pack's tables that has at least 5 observations, as one line each: up to three figures per row (attempts, SG per round, the rate, then SG per shot, leave, score to par or ratio), the benchmark figures beside them, and the row's signal in brackets. Lines are sorted by signal strength and then by the size of the difference, at most ten. Tables for a single-value column print one line for the whole table.

The tables are the candidate's own table first, then the section's set:

Every table prints as [table id: …], its title, its question, one column set per group of the split (the group name prefixes each column), reference columns for the benchmark where the table has them, an n: line with the sample behind each group, and its notes. Per-round figures are per 18 holes; SG figures use only shots with a strokes-gained value; counts include every shot. For metric players feet and yards in labels and notes are converted.

The analysis tables

Putting by distance putting.byDistance. Question: which putts separate the groups? One row per putting band (0–3, 4–6, 7–9, 10–15, 16–24, 25–40, 41+ ft, by distance to the hole at address; the half-foot edges mean 4–6 ft is 3.5 to 6.5 ft). Columns: Attempts/rd, Make % (the hole's last putt when it is also the hole's last shot and it finished in the hole, over all putts in the band), SG/rd; reference column make % for the tier, the benchmark make rate for the whole band as a band average, not the midpoint. Every row is a candidate ("Putts from 4–6 ft").

First putts by distance putting.firstPutt. Question: is it the first putt or the short ones? Same bands, first putts only. Columns: First putts/rd, 3-putt % (share of those holes with three or more putts), Avg leave when missed (mean distance left on first putts that were not holed), SG/rd.

Putting speed control putting.speedControl. Question: is it speed? Same bands, all putts. Columns: Putts; Short % (putts finishing short of the hole over all putts in the band, holed putts counting as not short); ≤3 ft % (putts finishing holed or within 3 ft, a tap-in next, over all putts, shown from 16 ft out); Avg leave (mean distance left after a missed putt); Verdict. Reference columns: short max, the share a golfer at the benchmark tier may leave short before speed is the issue, and the tier's ≤3 ft %, its expected rate at each putt's actual start distance averaged over the player's putts. Verdicts: "leaving short" when Short % exceeds the max by more than one standard error; "lag below benchmark" or "lag better than benchmark" when ≤3 ft % trails or beats its benchmark by more than one standard error; "speed fine" otherwise; "not enough data (n)" under 10 putts. Diagnose speed before anything about start line or face. The maxima and lag benchmarks are the tier's figures and print beside the player's own in the table; they are not listed here.

Short game by lie and distance shortGame.byLieAndBand. Question: which chips and pitches separate the groups? Around-the-green shots by distance to the hole (0–10, 10–25, 25–50, 50+ yds) and starting lie group (Fairway, Rough, Bunker, Other), rows with no shots dropped. Columns: n/rd, SG/rd, SG/shot, Median leave (feet), Missed green (shots that ended neither on the green nor in the hole). Every row is a candidate ("Short game: Rough 10–25 yds").

Chips from 10–25 yds: short-sided or not shortGame.shortSided. Question: am I facing harder chips, or hitting easy ones badly? Around-the-green shots from 10–25 yds in four rows: Rough, short-sided; Rough, not short-sided; Fairway; Bunker. Short-sided is the hole's stored flag. Columns: n/rd, SG/rd, SG/shot, Median leave. Every row is a candidate ("Chips 10–25 yds: rough, short-sided").

Short-sided and two-chip holes shortGame.shortSidedHoles. Question: do the hard positions cause the two-chips? One Value column: Short-sided holes/rd; Two-chip holes/rd; Two-chips that were short-sided (x of y).

Two-chip holes: first chip vs second shortGame.twoChipSequence. Question: is it the first chip or the second? Rows First chip and Second chip, the first two around-the-green shots on holes flagged two-chip. Columns: SG/shot, Median leave.

Two-chip holes: outcome shortGame.twoChipOutcomes. Question: how much does a two-chip cost on the card? One Value column: Ended bogey or worse; Ended double or worse, as shares of two-chip holes.

Approach miss profile approach.missProfile. Question: what kind of miss is it? One row per approach-miss bucket in the order Big lateral miss, Well short, Average shot bad result, Well long, Average / good. Columns: n/rd, Share (of all approaches), SG/rd, SG/shot, GIR % (bucket shots ending on the green or in the hole). Every row but "average" is a candidate ("Approaches: well short").

Approaches inside 175 yds from the rough approach.fromRough. Question: what does missing the fairway cost on the next shot? Approaches from the rough lie only, with a target distance under 175 yds (distance to the hole when no target was recorded). Deep rough, recovery and bunker are left out: their scoring and proximity expectations differ too much from the rough to be pooled with it. One Value column: n/rd; Well short %; GIR %; SG/rd. The whole table is one candidate, and it carries an Upstream line.

Short approaches by lie and distance approach.shortByLieDistance. Question: where do the short ones come from? Well-short approaches by distance to the hole (under 125, 125–175, 175+ yds) and lie group. Columns: Short/rd, Short rate (share of that cell's approaches), SG/rd (short).

Short rate by club approach.shortByClub. Question: which clubs come up short? Clubs with at least 5 approach attempts across all groups, sorted by pooled short rate; approaches with no club assigned are left out and counted in a note. Columns: Attempts, Short rate (per attempt with that club), Median shot ÷ target.

How short are the short ones? approach.howShort. Question: chunks or under-hit partials? Well-short approaches by shot distance as a share of target distance: under 60 %, 60–80 %, 80–90 % of target. Columns: Shots, Share of short.

The average shots approach.averageShots. Question: are my good shots worse too? Approaches in the average/good bucket only. One Value column: GIR %; Median leave; Inside 20 ft; SG/shot.

Off the tee driving.summary. Question: is the tee shot part of the gap? Par-4 and par-5 tee shots. One Value column: Tee shots/rd; SG/rd; Fairways hit % (the stored round stats when present); Big misses/rd (SG ≤ −0.7; the fact sheet's big-miss counter is the app's stored one at SG under −0.3); Penalty strokes/rd; Median tee shot (par 4/5); Ending in rough %; Ending in bunker/penalty/recovery % (a finish in deep rough counts here, with the recovery lies). The whole table is one candidate, with big misses per round as its frequency.

The next shot after the drive driving.downstream. Question: what does the rough cost me afterward? One Value column: Approach SG/shot from fairway; Approach SG/shot from rough; Rough approaches/rd; Fairway approaches/rd. Fairway means fairway and tee; rough means the rough lie only (deep rough is in Other).

Tee club on par 4s and 5s strategy.teeClub. Question: which club off the tee costs strokes? Clubs with at least 5 par-4/5 tee shots in a group, the six most used. Columns: n/rd, Fairways (finishing on the fairway, green or in the hole), Big misses (SG ≤ −0.7), Penalties (share of tee shots that incurred a penalty), SG/shot.

The engine computes further tables (birdies, double-bogey causes, par-5 decisions, recovery shots, round by round) that this version does not send; the fact sheet's overview rows and the stats summary carry the figures a conversation needs from them.

slice_shots

A small comparison table (at most 8 rows, both groups, per round) for one category of shots, filtered by lie group, distance band and par, grouped by lie, club, distance band, par or approach-miss bucket. For a number the fact sheet and the packs do not have, such as rough approaches inside 175 yards by club. Distances are in feet for putts and yards otherwise; min_distance is inclusive and max_distance exclusive. Shots without strokes gained are left out.

One row per group key, the busiest rows first. For each group of the split:

Distance bands for group_by: distanceBand: putts 0–3, 4–6, 7–9, 10–15, 16–24, 25–40, 41+ ft; around the green 0–10, 10–25, 25–50, 50+ yds; approaches under 50, 50–100, 100–125, 125–150, 150–175, 175–200, 200+ yds; tee shots under 300, 300–350, 350–400, 400–450, 450–500, 500+ yds. A lie grouping uses the four lie groups. Long tables are trimmed to fit, and the note says so.

list_holes

Holes with a given outcome, at most 12, each with the round, the score and the costliest shot.

Columns: Group, Round · hole (date and hole number), Par, Score, Costliest shot (kind, starting lie, distance, strokes gained, and "penalty" when the shot carried one). Worst score to par first (best first for birdieOrBetter), then the most recent round. par narrows to one par; limit defaults to 8.

list_shots

Individual shots matching a slice, worst strokes gained first, at most 20 rows (default 12). It is the only tool that returns individual shots. In a putts-only list a club-less putt reads "Putter".

The club filter takes the label as the tables show it ("7i", "Dr", "60"). The approach_miss filter takes one of the five approach-miss buckets, which are assigned in order, first match wins:

Shots with no target or shot distance skip the short and long tests.

get_course_report

One row per hole for one course across the player's own rounds there. Player data only: every comparison is the player against the player, and there are no benchmark baselines in this version.

Summary lines name the costliest holes, the holes played better than the player's norm on that course (by more than 0.1 SG per round), front nine against back nine, and the dominant loss category. The default window is every round in the cap; narrow it with the window filters or tee.

get_hole_sequences

Mental-game and sequencing patterns, par-matched. Three sets of cuts per group, each cut a subset of the group's scored holes:

With a two-group split the tool adds a group-difference table: per cut, each group's diff, the gap between them, and a signal from a test of one group's residuals against the other's. A final table lists the worst three-hole stretches in the window: round, holes, scores (±) on the three holes, their total to par, and running before, the round's score to par when the stretch began. Overlapping stretches in one round count once, the worst of them. Nine-hole rounds contribute to holes 1–9 only.

Limits

When a cap is reached the tool says so in plain words, and the limit text says when it resets. Caps keep one player's analysis fast and rule out bulk use.

Privacy and your account

• You connect with your Shot Pattern account. The connector reads your rounds, holes, shots, bag and benchmark tier, and nothing else from your account.
• Everything is read-only. There is no way to change, add or delete data through the connector.
• Your shot coordinates never leave Shot Pattern. Results carry distances, lies, clubs and strokes gained, not positions.
• Your data is sent to the AI assistant you connected, and that assistant's own privacy terms apply to what you discuss there.
• To stop access, disconnect the connector in your assistant's settings. The sign-in it holds lapses after 60 days without use and a year at most, and deleting or disabling your Shot Pattern account ends every connection at once. To revoke every connection sooner, ask support.
• Questions: shotpattern.app/support. Shot Pattern's privacy policy is at shotpattern.app/privacy-policy.

Technical notes

For hosts and developers. Shot Pattern Intelligence is a remote server speaking the Model Context Protocol over streamable HTTP, with OAuth 2.1 sign-in through the player's Shot Pattern account. Every tool is annotated read-only and non-destructive. Results are plain text (Markdown tables) so every host reads the same thing. The server's instructions, sent to the host on connection, point to this page for definitions.

Changes

• 2026-11: first version of this guide, for the eleven tools above.
• 2026-11-04: deep rough moved from the rough lie group to other; the from-the-rough and off-the-tee tables follow. get_stats_summary shows approach proximity in the player's units and as a percentage, drops short-game lies with no shots, labels the closing up-and-down figure, and converts distances for metric players. get_round takes a time zone.

This guide is generated from the connector's own source so that it tracks the server. When a tool's own note and this page disagree, the tool note is the newer of the two; tell us at shotpattern.app/support.