Start: the question
Nobody outside the BAA knows how the standards are set
To apply for Boston you need a qualifying time; charity places are the main other way in (BAA entry rules). Other Majors take time qualifiers too, Chicago among them, but they also hold a drawing open to everyone. The BAA publishes a standard for 11 age groups in each of two genders. Its announcements cite "careful analysis of results data" and stop there.
So "are they fair?" can only be answered by proxy: pick a reference for how hard each bracket should be, then check whether every standard sits the same distance from it. Any single reference is arguable, so I used three.
- 1World-record multiplier. The BQ time divided by the bracket's world record. Transparent, and exposed to one exceptional record.
- 2Top-three records. The same ratio against an averaged top three, to see whether a single performance drives the result. The second and third marks are estimated from fixed depth factors.
- 3WMA age grading. The share of age-specific potential each standard demands, from the WMA 2023 factor tables.
The board
Every bracket, men and women side by side
Each row is an age group; the blue square is the men's standard and the rust dot the women's, placed by how lenient the standard is against its reference. If the table were uniform, every mark would sit on the dashed median. The men's squares nearly do. The women's dots wander, and W80+ sits alone on the strict side. Switch the measure to see that the pattern holds under age grading and under the old 2020 to 2025 standards.
Hover or tab to a mark for the bracket's numbers.
outputs/analysis_results.csv. Age-graded
percentages are flipped so that right means more lenient on every view.
"Minutes off uniform" is the current standard minus the median-multiplier
standard.First result
The gender-bias claim does not hold
Welch t-test on the gender means, world-record framework. t = −0.25.
The men's multipliers average 1.488 and the women's 1.495. Age grading gives the same answer: the men's standards ask for an average of 69.1% of age-graded potential and the women's for 68.3% (t = 0.42, p = 0.68). The usual forum complaint is about the mean, and on the mean the two genders are treated alike.
The result that holds
Men's standards are consistent. Women's are not.
Every men's bracket falls between 1.43× and 1.53× its world record, a coefficient of variation of 1.9%. The women's run from 1.27× to 1.62×, a CV of 6.6%. Levene's test for equal variance gives W = 5.04, p = 0.036. Two women who both meet their standard can be running at very different fractions of what their age group has shown is possible.
The gap survives a change of reference. The ratio is about 3.5 to 1 under the two record-based frameworks and about 2 to 1 under age grading.
Coefficient of variation (sample standard deviation over mean) of each framework's measure across the 11 brackets, computed on this page from outputs/analysis_results.csv.
The outlier
W80+ is too strict under every framework
The W80+ standard is 5:20:00. The median multiplier would put it at 6:17:20, 57 minutes slower, and age grading puts the gap at 56 minutes. The two frameworks start from different premises and land a minute apart.
The reference explains it. Yoko Nakano's 4:11:45, run in Tokyo at 81, is an exceptional record, and the current standard asks women over 80 to finish within 27% of it. Every other bracket sits between 1.42× and 1.62× its record. The table below sorts by any column; it opens on minutes off the uniform standard, strictest first.
Uniform BQ applies the median multiplier (1.50×) to the bracket's record. Negative minutes mean the current standard is stricter than uniform. Click a column heading to sort; click again to reverse.
The 2026 tightening
Five minutes off every standard under 60
For 2026 the BAA cut five minutes from every standard below age 60, its largest tightening since 1990, after 33,249 applications for about 24,000 qualifier places. The 60+ standards did not move.
A flat cut leaves the shape of the table in place. Against the same world records, the women's CV was 7.3% under the 2020–2025 standards and is 6.6% now; the men's moved from 1.8% to 1.9%. (The board above has the old standards as one of its views.)
It did make one step steeper. The jump from 55–59 to 60–64 was 15 minutes and is now 20, for men and women alike, so a qualifier gains 20 minutes of allowance on their 60th birthday.
data/bq_standards_2026.csv. The shaded band is the
55–59 to 60–64 step, 20 minutes in both lines since the 2026
cut (15 before it).
Stress test
Is the variance gap an artifact of the records chosen?
I re-ran the world-record framework three ways: without W80+, with Sinead Diver's 2:21:34 as an alternative W40–44 reference, and with Tigst Assefa's women-only record of 2:15:41 in place of the mixed-race mark for W18–34.
The men's CV stays at 1.9% in every case. The women's is 6.6% with Diver and 6.4% with Assefa, and drops to 4.5% only when W80+ is removed. That one bracket accounts for about a third of the women's spread, and even without it the women's CV is more than twice the men's.
Recomputed on this page from outputs/analysis_results.csv with the three substitutions in sensitivity_analysis() of src/analysis.py: drop W80+, set the W40–44 record to 2:21:34 (8,494 s), set the W18–34 record to 2:15:41 (8,141 s).
Your bracket
What your time means in your bracket
Pick your bracket and enter a marathon time. The card compares it with the 2026 standard, with the 4:34 cushion the BAA needed to fill the 2026 field, and with the uniform standards the two main frameworks would set.
h:mm:ss or h:mm
Limitations
What this does not show
There are 11 brackets per gender, so the tests are underpowered. Levene's p = 0.036 clears 0.05 by a small margin, and it was run on the world-record framework only. A p of 0.81 means no difference was detected, which is weaker than evidence of no difference.
The top-three framework is an approximation. The second and third marks are set at 3% and 6% slower than the record for brackets with depth, and 5% and 10% for 70+; none of them comes from actual results.
Below 40 there is no separate masters record, so 18–34 and 35–39 share the open record. That makes 35–39 look more lenient than it may be.
Non-binary athletes (110 accepted in 2026) are left out because the BAA says it does not yet have the data to set standards for the category. Finally, "fair" here means equal proportional difficulty against a reference. The BAA may be aiming at field size, continuity or something else; a different definition could reach a different verdict on the same table.
Finish: sources and method
Where the numbers come from
- Standards
- 2026 BAA qualifying times from baa.org
- Open records
- World Athletics: men 1:59:30 (Sawe, London 2026, ratified July 2026), women's mixed 2:09:56 (Chepngetich, Chicago 2024). Chepngetich is serving a three-year ban, announced by the Athletics Integrity Unit on 23 October 2025, after a positive test for the diuretic hydrochlorothiazide on 14 March 2025; the AIU said her records from before that test stand while it looks into other material from her phone, and the World Athletics all-time list still puts 2:09:56 first (checked 28 October 2026).
- Masters records
- M40 to W80+, from the Wikipedia masters road-running list, checked against WMA records
- Age factors
- WMA 2023 Appendix B, the factor for each bracket's midpoint age
- Field size
- BAA releases: 33,249 applied, 24,362 accepted, 4:34 cutoff under BQ
- Tests
- Welch t-test on means, Levene's test on variance, coefficient of variation, z-score deviation heatmaps, sensitivity re-runs
- Stack
- Python: pandas, NumPy, SciPy, matplotlib, seaborn; Jupyter notebook. The board, table and card on this page are plain SVG and HTML drawn from
outputs/analysis_results.csv - Reproduce
pip install -r requirements.txtthenpython src/analysis.py- Sibling
- Auditing the scoring tables, the same fairness question asked of World Athletics points