FootballGPT Coaching Intelligence · Refreshed weekly
Every chart on this page comes from real use of FootballGPT: the practices coaches create, when they plan, who they coach and where on the pitch the work sits. A generated practice is not proof it was delivered, so these numbers describe what coaches build rather than what happened at training.
All charts are aggregate-only; no coach, club, or session is identifiable. Last refresh: 07 Sept 2026.
2026 data report
An 18-page report of findings, definitions and limitations from anonymised FootballGPT activity.
Frozen page release refreshed weekly. The PDF records its 6 September 2026 snapshot.
20,706
coach-mode user turns
8,040
requested practices
7,422
automatic practices
3,597
coach-mode accounts
Strict publication cohort: known test, automation and opted-out accounts removed. Practice figures include coach-profile chat generations only; requested and automatic output are reported separately. 15,311 of 15,463 practices have both a reportable age and category. This is product evidence, not a census of grassroots football.
Public product evidence
These figures count the whole of FootballGPT and are rounded down, so the true numbers are at least this high. They are not estimates of grassroots football as a whole. Each chart below uses a narrower group of coaches and states its own sample size.
Evidence checked 2026-09-02. Read the full methodology.
This briefing / three signals
These are direct readings of aggregate FootballGPT activity. Each card names the measured unit and the limitation that matters most.
Observed product behaviour
57.9%
/ 26.8%
technical / game-based
57.9% of categorised output is technical, compared with 26.8% game-based.
This measures generated output. It does not prove what coaches selected, delivered or found effective.
Observed product behaviour
18%
Monday
Monday accounts for the largest share of practice generation.
Generation time is a product activity signal, not proof that the session was planned or delivered then.
Observed product behaviour
27%
Junior (U10-U12)
Junior (U10-U12) is the largest age-labelled share of generated practices.
The denominator includes generated practices with an assignable age band, not a representative census of grassroots football.
The coaching loop / evidence coverage
Questions and stated problems exist, but the clean real-world coaching cohort is still being repaired.
Generated practices and analysis activity are recorded directly.
Saved practices and persisted sessions show intent, not confirmed delivery.
Post-session evidence is too sparse for a population claim.
The link from reflection to the next coaching decision is still being built.
Evidence coverage reviewed 1 August 2026. Aggregate stages are not necessarily the same coach journey.
Chart 1 / generated practice mix
57.9% of categorised output is technical, compared with 26.8% game-based.
This chart describes FootballGPT's categorised output. It can show how the generated mix changes by age band, but it does not show what a coach selected, delivered or found effective. The chart is practice-weighted; the table below also reports coach-weighted shares so prolific accounts do not dominate.
| Age band | Technical, practice-weighted | Technical, coach-weighted | Contributing accounts |
|---|---|---|---|
| Mini (U6-U9) | 57.9% | 67.5% | 196 |
| Junior (U10-U12) | 49% | 60.3% | 394 |
| Youth (U13-U15) | 41.4% | 46.1% | 362 |
| Senior Youth (U16-U18) | 42.3% | 52.2% | 420 |
| Adult (U19+) | 30.4% | 32.6% | 415 |
| Mixed | 31.1% | 43.6% | 130 |
n = 15,311 categorised animated practices with age band. Each visible category has at least 50 contributing accounts. methodology
Published practice example
Watch the players, ball and coaching sequence move.
A U9-U10 practice for teaching through passes.
Chart 2 / planning rhythm
Monday accounts for the largest share of practice generation.
The heatmap records when FootballGPT generated a practice. It is a useful rhythm signal, but it does not prove when the coach made the final plan or delivered the session.
| 0 | 3 | 6 | 9 | 12 | 15 | 18 | 21 | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Sun | ||||||||||||||||||||||||
| Mon | ||||||||||||||||||||||||
| Tue | ||||||||||||||||||||||||
| Wed | ||||||||||||||||||||||||
| Thu | ||||||||||||||||||||||||
| Fri | ||||||||||||||||||||||||
| Sat |
n = 15,463 animated practices. methodology
Published practice example
Watch the players, ball and coaching sequence move.
A U14-U15 practice combining possession, counter-pressing and finishing.
Chart 3 / audience mix
44.4% of mode-tagged queries come from Football Manager video-game mode.
Coach-mode activity makes up the majority of mode-tagged queries, but Football Manager video-game players are the second-largest audience — sharing the same tool with very different intent.
n = 43,816 mode-tagged queries. Modes with fewer than 50 hidden. methodology
Published practice example
Watch the players, ball and coaching sequence move.
A defensive practice built around recognising pressing triggers.
Chart 4 / age band distribution
Junior (U10-U12) is the largest age-labelled share of generated practices.
Percentages show each age band's share of generated practices with an assignable age. This is a view of FootballGPT activity, not a representative census of grassroots football.
n = 15,311 animated practices with assignable age band. methodology
Published practice example
Watch the players, ball and coaching sequence move.
A U15+ practice for defending wide overloads from a mid-block.
Cut 5 / pitch concentration
89%
Of generated practice diagrams, this share centres the action in the middle third.
This is computed from player positions in AI-generated diagrams. It is mostly a product-output signal, not proof that coaches prefer the middle third. Historical intent flags span detector revisions, so we do not publish one blended coach-request percentage.
n = 14,954 generated practices with at least one player. Pitch thirds are computed from each practice's average player y-coordinate (0-100, where 0 is the defending goal line). 'Middle' covers y=33-66. methodology.
Cut 6 / player counts
100%
Of generated Mini-soccer practices use a 5v5-or-smaller format.
Generated formats get larger across the older age bands. That describes FootballGPT output, not a measured preference or learning outcome. Player-count intent is not reported as one historical percentage because the stored flags are not yet versioned and calibrated as a complete series.
| 1v1 | 2v2-3v3 | 4v4-5v5 | 6v6-7v7 | 8v8+ | total | |
|---|---|---|---|---|---|---|
| Mini (U6-U9) | 152 | 830 | 617 | 1,599 | ||
| Junior (U10-U12) | 200 | 1,404 | 1,517 | 754 | 230 | 4,105 |
| Youth (U13-U15) | 787 | 973 | 1,059 | 470 | 3,289 | |
| Senior Youth (U16-U18) | 723 | 870 | 749 | 550 | 2,892 | |
| Adult (U19+) | 535 | 636 | 636 | 554 | 2,361 | |
| Mixed | 157 | 155 | 312 |
n = 14,558 practices with both age band and player count. Player count is derived from drill_data.players[]; bands are 1v1, 2v2-3v3, 4v4-5v5, 6v6-7v7, 8v8+.
Cut 7 / cohort profile
58%
Mini-soccer's category profile is the most technical-dominant of any age band.
Each polygon is one age band's share of generated practices in six categories. Older bands open out into tactical, game-based and set-piece work, but technical still dominates. This measures classified output; it does not cleanly separate a stated request from the AI category fallback.
Categories: technical, tactical, game-based, set-piece, warm-up, physical. Each axis is the band's share of practices in that category. Bands plotted: Mini, Junior, Youth, Senior Youth (top 4 by volume). methodology
Cut 8 / weekly trends
6,965
Practices generated across the last 12 weeks. Peak week: 1,230.
Weekly counts show recorded platform activity over time. They do not isolate seasonality, acquisition changes or product-release effects. The mode-mix stack underneath describes how recorded activity changed across product modes.
Animated practices per week
Peak week: 10 Aug (1,230 practices)
Mode mix per week (queries)
Cut 9 / animation complexity
3.2 → 3.6steps
Practice complexity barely scales with age.
Average sequence step count per practice (each 'step' is one phase of the animation). Mini practices average ~3 steps; Adult barely reaches 4. Either coaches genuinely want short practices regardless of age, or the AI tends to produce a similar number of steps regardless of prompt.
Cross-product
The charts above come from FootballGPT. Separate 360TFT product cohorts provide supporting context below. They are not joined coach journeys and are not directly comparable unless stated.
Chart 5 / CoachPage profile cohort
This separate cohort covers directory-visible CoachPage profiles. Each person who builds a public CoachPage can state their licence, country, years coaching, the age groups they teach, and the specialities they list.
Licence band
Years coaching
Country
Age groups taught
Stated specialities
n = 73 directory-visible coaches, at or above the k≥50 anonymity floor used elsewhere on this page. Source: coachpa.ge. Coaches teaching multiple age groups appear in each band.
Chart 6 / what coaches reflect on
After a session, what do coaches think about? CoachReflect users tag each reflection, rate their mood and energy, log session type, and self-classify their coaching level. Free-text reflection content is never published — only the structured fields below. This is a small, early cohort — see the sample size below rather than reading the shares as population-level.
Top reflection tags
Coaching level (self-stated)
Post-session mood (1-5)
Post-session energy (1-5)
Session type
n = 121 reflections from 28 coaches. Most profiles do not specify a coaching level (onboarding does not force one). Source: coachreflection.com.
FAQ
Each answer below is grounded in the live numbers shown above, refreshed weekly. Where the underlying cohort is small, the answer says so.
In the current categorised output, 57.9% of practices generated for U6-U9 are technical and 26.8% are game-based. This describes FootballGPT output; it does not show which practices coaches selected, delivered or found effective.
see the chart →Monday accounts for the largest daily share of practice generation (18%). The busiest single hour is Wednesday at 16:00 UTC. These are generation timestamps, not proof of when a coach finalised or delivered a session.
see the chart →The largest age band in the current generated-practice dataset is Junior (U10-U12), with 4,105 generated practices. This measures generation volume, not delivery or effectiveness.
see the chart →No. Of all mode-tagged queries, 47.3% come from coach mode and 44.4% from Football Manager video-game mode. They share the same AI tooling but have different intent, so the report keeps the two lanes separate.
see the chart →Among directory-visible CoachPage coaches, the qualification mix includes UEFA B, FA Level 1/2, Coerver, S&C and others. See the CoachPage panel for the current breakdown; that cohort clears the k≥50 anonymity floor used elsewhere on this page.
see the chart →From CoachReflect, the most common structured reflection tag in the current snapshot is player_development. This is a small, early cohort — see the CoachReflect panel for the current sample size. Free-text reflection content is never published — only structured tags and ratings.
see the chart →The dataset refreshes weekly and every published chart applies a minimum cohort size before a figure is shown. See the methodology page for the exclusions, thresholds, and what each source does and does not cover.
see the chart →All charts are aggregate. No row-level data, no PII, no club or coach identifiers — ever. For interviews, additional cuts, or a press-ready summary, get in touch.
FootballGPT offers
For coaches · COACHOS50
50% off your first CoachOS payment
New customers only. Enter the code at checkout. Free to try without a card. View FootballGPT pricing.