Industries / Consumer goods / Sporting goods
Testing and standards
Testing and standards is the work of writing equipment rules and building the lab tests that check whether a ball, club, bat or exercise machine meets them, and of tracking recalls when products hurt people.
6
Milestones achieved, from 1921 to 2023.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1921
Golf caps the ball
The rubber-cored ball that took over golf from about 1902 flew much farther than the old solid ball, and courses began to look short. The R&A, golf's rule-maker outside the United States, ruled that from 1 May 1921 a ball could weigh no more than 1.62 ounces and measure no less than 1.62 inches across, and reserved the right to limit how far balls fly. Sources disagree on the date: Golf Australia's archive gives the R&A rule of 1 May 1921, while golf historian Peter Trenham says the USGA set the same limits in 1920.
Credit
The R&A; the United States Golf Association (USGA)
1931
The balloon ball fails
The USGA adopted a larger, lighter ball, at least 1.68 inches across and no more than 1.55 ounces, which became compulsory at the start of 1931. Golfers found it hard to control in wind and nicknamed it the "balloon ball". On 15 September 1931 the USGA allowed the weight back up to 1.62 ounces and kept the larger size, a rule Golf Australia's archive dates to 1932. The R&A kept the smaller ball until its rules of 1 January 1990.
LessonA rule that changed how the ball flew was reversed within a season, and the US and British balls then stayed different sizes for nearly 60 years.
Credit
USGA
2004
A pendulum measures spring
A thin driver face can flex like a trampoline and add distance, which golf's rules call "spring-like effect". The R&A and the USGA jointly developed a pendulum test in which a steel mass swings into the clubface at low speed and the contact time is measured in microseconds. From 1 January 2004 every driver submitted was ruled by this test, with a limit of 239 microseconds plus a tolerance of 18, replacing the coefficient of restitution (COR) test.
LessonA club can pass when new and fail after use; in December 2023 the two bodies said drivers tested in the field had crept over the limit and that they may hit sample drivers 150 times at 125 mph before ruling.
Credit
The R&A; USGA
2011
College bats made wood-like
From 1 January 2011 US college baseball allowed only aluminum and composite bats certified to BBCOR, the ball-bat coefficient of restitution, a lab measure of how much bounce a bat gives the ball. The aim was for these bats to hit like wood, after home runs and runs per game had climbed. By the NCAA's midseason count, Division I scoring fell from 6.98 to 5.63 runs per team per game and home runs from 0.85 to 0.47, compared with the same point in 2010.
ncaanewsarchive.s3.amazonaws.comncaanewsarchive.s3.amazonaws.com
Credit
NCAA Baseball Research Panel; NCAA Baseball Rules Committee
2021
A fatal treadmill hazard
On 5 May 2021 Peloton and the US Consumer Product Safety Commission (CPSC) recalled about 125,000 Tread+ treadmills. A 6-year-old child had died after being pulled under the rear of one, and Peloton had received 72 reports of adults, children, pets or objects pulled under, including 29 injuries to children. Peloton later recalled about 833,000 Bike+ exercise bikes, in November 2025, after seat posts broke.
LessonMachines used at home meet children and pets, so a gap that is safe in a gym can be fatal in a house.
Credit
CPSC; Peloton
2023
Golf balls rolled back
On 6 December 2023 the R&A and the USGA decided that balls will be tested at a clubhead speed of 125 mph instead of 120 mph, and must still fly no farther than 317 yards plus a 3-yard tolerance, so a ball must fly less far at a given swing speed to pass. The new test applies to the List of Conforming Golf Balls from January 2028, and balls listed earlier may be used until January 2030. In March 2023 they had proposed it as an optional local rule for competitions; after comments, they changed the test for all balls.
Credit
The R&A; USGA
Watch for
A 125 mark on balls
The R&A and the USGA proposed that every ball listed as conforming under the 2028 test must carry the number "125" inside an outline, at least 4.45 mm wide and 3.175 mm high, so golfers and officials can tell it apart by eye. They took comments until 22 September 2025.
Achieved whenthe January 2028 List of Conforming Golf Balls includes marked balls and they are on sale.
Claimed by
The R&A and USGA, August 2025
Faster injury alerts
The CPSC says its National Electronic Injury Surveillance System (NEISS), which samples emergency room visits and was designed in 1972, will be rebuilt as NEISS-R. It is to cover all 50 states, spot rare and fast-emerging product hazards sooner, and be fully running by early 2027, as reported by News4JAX.
Achieved whenNEISS-R publishes data from hospitals in all 50 states and an outside analysis shows it flagged a product hazard earlier than the old system would have.
Claimed by
CPSC, July 2026
Open
01
A history of conforming balls
An open dataset of every monthly List of Conforming Golf Balls, one row per ball model per month, with maker, country, cover color, construction, spin rating, dimple count and markings, plus a file of the models added and removed each month. When the 2028 lists start, add a column for the "125" mark.
The briefCredit · unclaimed02
Backtest a treadmill hazard alarm
A classifier that reads the short injury descriptions in NEISS emergency-room records and marks those where a person or pet was pulled under a treadmill. Run it month by month over 2015 to 2021 and report the first month a simple alarm rule would have fired, compared with the Tread+ recall of 5 May 2021.
The briefCredit · unclaimed
Open 01 · the brief
A history of conforming balls
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open dataset of every monthly List of Conforming Golf Balls, one row per ball model per month, with maker, country, cover color, construction, spin rating, dimple count and markings, plus a file of the models added and removed each month. When the 2028 lists start, add a column for the "125" mark.
Why it is still open
The R&A and the USGA publish the list as a monthly PDF, 88 pages for February 2025, that replaces the previous one; the list itself says balls are "added to and deleted from the list each month". A web search and a GitHub search found no compiled history, only single copies of old lists.
Built with what exists
Python with PyMuPDF or pdfplumber to read the PDFs; a language model to split the marking descriptions into fields, with a person checking samples; archived copies of past lists; Zenodo or Hugging Face to publish the CSV files.
First step
Download the current list from the R&A, turn it into a CSV and check that the row count matches the entries in the PDF.
Done looks like
At least 24 monthly lists parsed, 300 randomly chosen rows checked by hand against the PDFs with the error rate published, and a monthly count of listed models by maker and country.
Open 02 · the brief
Backtest a treadmill hazard alarm
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A classifier that reads the short injury descriptions in NEISS emergency-room records and marks those where a person or pet was pulled under a treadmill. Run it month by month over 2015 to 2021 and report the first month a simple alarm rule would have fired, compared with the Tread+ recall of 5 May 2021.
Why it is still open
Machine learning on NEISS narratives is established; Vichare and colleagues reported 90.47% accuracy in 2024 at predicting injury codes from them. The CPSC says its rebuilt system will spot emerging hazards sooner. Recalls have been backtested against other text: Bhat and Culotta trained a classifier on complaints to the CPSC and, run on Amazon reviews of recalled products, it found reviews reporting a hazard before the recall date for 45% of them (2017). A web search found no published backtest of a sporting-goods or fitness-equipment recall against NEISS narratives.
Built with what exists
NEISS public data, also packaged for R as hadley/neiss on GitHub; scikit-learn for a baseline; a small open language model fine-tuned on one GPU, such as Liquid AI's LFM3 (see the Note "One GPU to fine-tune"); the CPSC recall API at SaferProducts.gov for recall dates.
First step
Download NEISS records for 2015 to 2021, keep the exercise-equipment narratives that mention a treadmill, and read the first 100.
Done looks like
500 treadmill narratives labeled by two people with their agreement reported, the classifier's precision and recall on 100 held-out narratives, and a published monthly chart with code showing when the alarm would have fired, or that NEISS held too few cases for it to fire at all.