Serve Quality, WHOOP and where amateur tennis coaching is going

If you walked through the media area at the US Open in late August 2026, you would have seen something new sitting alongside the traditional match-stats board. It appeared on the screen after every serve: one number, called Serve Quality, updated in near real time, alongside four sub-dimensions with names like efficiency, accuracy, consistency and ball toss. IBM and the USTA launched it on 24 August, across all 254 singles matches. The system captures 21 joint positions from each player at fifty samples a second. It reads eight distinct phases of the serve motion. Across the fortnight, it generates somewhere in the region of 1.2 billion data points. The underlying formula is proprietary. The output is a score fans can read at a glance.

The tour has just been given a level of instrumentation the recreational game has never had. But what does any of this meansfor the club player? In the short term, not very much: a slightly richer television experience during the last week of the Slam. In the medium term, judging by the data arc every mature performance market has already crossed, quite a lot. And in the long term, it decides who wins the next decade of amateur tennis coaching.

What happens after measurement becomes abundant

Every mature performance market has crossed this same threshold before tennis. The pattern is remarkably consistent. Sensing becomes good enough. Measurement becomes expected. Then interpretation becomes valuable. Then memory becomes valuable. Then the whole thing settles into decision support.

Golf worked this out first. TrackMan turned ball-flight measurement into an entire coaching ecosystem: a launch monitor became a stored history of your session progress, which became 3D motion analysis of body plus club plus ball, which became the reference infrastructure of professional golf coaching. The measurement was the beginning. The compounding value has always been in the layers built on top.

Endurance and health followed. WHOOP began by measuring heart-rate variability and reading it back to you as a recovery number. The recovery number was useful. It has since become considerably less important than what WHOOP now calls My Memory: the store of your goals, routines, events and history that lets the coaching layer say what today’s recovery number means for you, given what you were trying to change. WHOOP’s product page describes it now as coaching that remembers.

Garmin does the same thing through a different aesthetic. Its Training Readiness score composites sleep, HRV, acute load and recent stress into one number that answers a decision: should I train hard today? The composite metric is valuable because it answers a decision the athlete is already trying to make.

Strava demonstrated a different variant of the same lesson. It has never manufactured the sensor. Its whole product sits above whatever device the athlete happened to buy, and its Athlete Intelligence layer turns the resulting statistics into plain-language interpretation, considered against the athlete’s history and declared long-term focus. The athlete relationship, it turns out, does not have to belong to whoever owns the wrist.

Hudl did the same for team sport. Video becomes defensible not when the recording is a bit better than the alternative, but when it becomes embedded in a weekly workflow: game film, tagged and analysed, connected to athlete monitoring, connected to scouting, connected to coach communication. Video that leaves the system after one watch has no compounding value. Video that stays becomes the operating layer of a programme.

Even Apple, which does not think of itself as a sports company, is quietly writing sports-technology capabilities into the operating system. Training Load, once a specialist metric that a runner needed a Garmin for, is now on the Apple Watch, calculated automatically from the last seven days against the previous twenty-eight. Whatever a specialist sports app can do today, Apple’s next release will likely make routine tomorrow.

The pattern across all of them is that after measurement becomes abundant, value migrates upward, into memory, decision and habit. Not into the sensor. Not into the raw metric. Into what the system knows about the athlete over time.

Tennis’s own path to this, from the tour down

The tennis arc has been running for eight years. Most people who read tennis technology news see the fan-facing product at the end of it and miss the progression that led there.

In 2018 IBM launched what it called AI Highlights at the US Open, initially as a media-side tool for automatically indexing points from match footage. Point-by-point tagging, previously requiring somewhere between 24 and 48 hours of manual analyst time per match, was compressed to about 30 minutes. The USTA player-development team, then led by Martin Blackman, took the same technology and pointed it at their coaching operation: coaches now had a searchable database of tagged match footage they could interrogate for their own purposes.

The more interesting move that same year was something called Red Steps. It began with two named tennis experts sitting in a room with IBM technology programme manager John Kent. Former world number one Ivan Lendl gave IBM an example of a tactical pattern in Novak Djokovic’s shot selection. Andre Agassi’s former strength coach Gil Reyes gave them an example from footwork. Reyes had noticed that when a player reverses lateral direction, whether they stay low and push off immediately or need an extra half-step tells you something about their fatigue. IBM took Reyes’s observation and translated it, via Hawk-Eye positional tracking data, into a machine-measurable metric: extra steps at moments of direction change. An expert had a hypothesis. The system was given a way to see it. The name Red Steps came out of that specific piece of coaching insight.

The 2019 launch was called Coach Advisor. It combined match video, tracking data and player-specific physical context, and its most substantive product was something called the Energy System, a composite of Physiological Load and Mechanical Intensity. It was tested with Sloane Stephens and Frances Tiafoe. The construct was not new speed or new distance readings. It was a coach-legible layer above them: given this player’s height, weight, movement profile and the demands of the match, is what we are asking them tactically also physically possible for them to execute? Raw tracking variables were not the product. The composite meaning above them was.

From 2020 onward the visible public arc shifted toward the fan-facing side. Watsonx AI Commentary launched in 2023 across all 17 US Open courts, generating tennis-language captions and narration from a foundation model trained on the sport’s specific vocabulary. Match Chat launched in 2025, an agentic system spanning all 254 singles matches, orchestrating data retrieval and quality checking and returning conversational answers to fan questions. Live Likelihood to Win updated point by point through the tournament. IBM’s own account of the 2025 event described the platform as delivering millions of AI-generated insights across the fortnight to more than 14 million digital users.

Serve Quality, launched on 24 August 2026, is the current end of the line. It is the deepest instrumentation the sport’s showcase tournament has ever exposed to the general fan and is a clear example of the pattern above. The evidence is dense (1.2 billion data points, 21 joint positions, 50 samples a second) but the user-facing product is compressed (one score, four sub-dimensions). The formula is not published. The system’s compression matters more than the size of its input.

So the eight-year arc looks like this: first automated recognition and indexing in 2018, then expert observations translated into machine-measurable metrics in the Red Steps case, then coach-legible composite constructs in Coach Advisor, then generative interpretation in the 2023 Commentary, then agentic reasoning and orchestration in Match Chat, then compressed real-time meaning delivered to fans this August in Serve Quality. The measurement was always the beginning. What compounded, and what people ultimately paid for, was the layers built on top.

The club is about to get the same

The ATP and WTA tour arc is the visible one because IBM and the USTA hold press conferences. The club arc is quieter but faster.

Phone-based capture applications have moved from recording only into recognition, statistics, longitudinal history and coach reviews, one release at a time, but none captures the full value chain. And the more recent releases across the category have begun accepting footage recorded outside the relevant apps themselves: that is, the apps are accepting footage from wherever it may have been filmed, be it on the phone, a mounted camera from another company and so on.

PlaySight has been doing it from a very different starting point: installed court cameras, deployed across USTA facilities, college tennis programmes and professional sports environments. Wingfield has built a similar installed-court model in Europe with persistent player accounts, drill scoring and coach access underneath it. Save My Play removes the capture-behaviour friction that has quietly held back every previous generation of amateur tennis technology: you book the court, you play, the video exists. Baseline Vision, ITF-approved in 2024, packs 3D ball tracking, bounce position, shot type and player position into a net-post-mounted dual-camera unit.

Then there is CourtReserve, which is more strategically interesting than most club-management software. It sits between the club, the court, the booking and the player identity. In April 2025 it added a live integration with Save My Play. In March 2026 it added one with PlaySight. On 19 August 2026 it added integration with the USTA Connect programme, which brings the player’s official recreational singles rating, doubles rating and United States performance level directly into the member’s profile. What CourtReserve is trying to stitch together inside one player experience is something tennis technology has never had at the club level: court, booking, player identity, official rating, installed camera and external analytics.

None of these are yet the amateur equivalent of what happened at the 2026 US Open. But the pieces are being assembled with unusual speed. In five years the club player will have a version of what the tour got this August: sophisticated capture, dense measurement, connected identity. What they will not automatically have is the interpretation and memory layer on top. That is what the next decade of amateur tennis coaching decides.

What the racket-sensor generation taught the sport

However, this is not an inevitable outcome, and tennis has already tried this once.

The 2013 to 2018 generation of racket-mounted sensors (Babolat Play, Sony Smart Tennis Sensor, Zepp Tennis, Babolat POP) attempted the same instrumentation project one wave earlier. They measured swing speed and impact position and spin type and shot count. What they produced was interesting data. Most of them have since been withdrawn, discontinued or lost support.

The metrics they produced were not the problem. The technology mostly worked. The problem was that a new number is not a product. A player who bought a sensor, attached it to their racket, charged it, synced it, opened the app, read the numbers and put the sensor back in the bag until next Tuesday was doing a lot of work for very little in return. There was no memory, no decision the system was making on the player’s behalf, no equivalent of Garmin’s Training Readiness saying: given this data and your history, this is the one thing that matters today. Sensing innovation without a recurring decision loop stays a gadget.

What the ATP and WTA tours are doing and the club instrumentation of the last few years does not necessarily lead to automatic value for the amatuer. The data measurement will be there, whether captured by a phone on the back fence or a mounted camera on the net post. The question is whether the layer built on top of the measurement is any better than what the racket-sensor generation offered a decade ago. If it is not, this wave will fail in the same way the last one did.

The three layers, at the amateur level

What every survivor from the previous decade of performance technology has in common is a three-layer stack.

The first layer is the data. Enough of it, from a reliable enough source, sampled often enough to see what happened.

The second layer is the interpretation. Not the number. The meaning of the number. What Reyes did for Ivan Lendl’s team in the Red Steps case study, or what WHOOP does when it takes today’s recovery number and reads it against your history, or what a touring pro does when they watch two minutes of an amateur’s footage and hand back one specific cue.

The third layer is memory across time. Not the last session as an isolated event, but the last session read against the last twelve sessions, so the system can tell you what has changed, what has not, and what to work on next.

Golf worked out that the launch monitor is useful, but that TrackMan Range’s stored history of your session progress is what makes you come back. Endurance sport learned the same thing about heart-rate straps: the wearable that remembers your history is worth more than the wearable that measures more precisely. Hudl worked out that game film is useful, but that game film embedded into a weekly team workflow is what makes it defensible. Tennis, at the amateur level, is arriving at exactly the same three-layer stack now.

At All Court we have been building the amateur version of the three-layer stack for the last several months. Phone footage on the back fence supplies the evidence. Touring pros supply the interpretation, and increasingly teach the system what to look for. A persistent record, what we call the All Court Rating, supplies the memory: what your tennis is doing right now, what evidence supports it, what to work on next, and whether last month’s priority moved. The tour player has had a version of this for a decade in a coaching team, a video archive and, increasingly, a stats stack. The amateur has never had it.

What keeps cropping up

As founder, I have been talking with the amateur players who have submitted footage to our reviewers over the past couple of years. Some things keep cropping up across the range of playing styles and levels. The player who is handed a diagnosis without a specific drill to work on next feels stuck. The player who is told what is wrong without being told what to try tomorrow disengages. The player who films their tennis, gets feedback, and then has to remember to film it all again the following month without a prompt from the system: most of them do not. Almost every serious amateur we hear from wants the same three things: a specific fix, a follow-up assignment, and a sense that the system remembers what they were working on. Without those three things, even excellent video analysis becomes a gadget.

We wrote about the specific pattern of what those touring pros keep seeing in a separate piece: the ten most common observations across a hundred amateur reviews this season.

The reckoning

In five years the amateur game will be as instrumented as the tour is now. The question the next five years decide is who turns that data into an understanding of the player that survives from one session to the next, and who tells them the true next thing to work on.

Measurement will be plural. Meaning will be scarce. That is where the decade is being won.