Testing the World Athletics points tables against the elite

The best 5000 m runners outscore the best javelin throwers by 68.5 points

World Athletics points are meant to make a 1200-point shot put worth the same as a 1200-point 100 m. If that held, the top 30 marks in every event would earn about the same points. In the 2023 and 2024 top lists they don't: the median ranges from 1203.5 in the javelin to 1272 in the 5000 m, and running events sit high while jumps and throws sit low.

Read the short version Code and data on GitHub

68.5 pts

Spread of the top-30 median WA score between the highest event (5000 m) and the lowest (javelin).

+24.5

5000 m above the cross-event norm of 1247.5, the median of the 16 event medians.

−44

Javelin below the norm. Hammer (−39), high jump (−28.5) and triple jump (−27.5) follow.

H = 236

Kruskal-Wallis statistic across the 16 events' top-30 scores, p < 10⁻⁴⁰.

At a glance

QuestionDoes a WA score mean the same elite standing in every event?
DataWorld Athletics top lists, 2023 and 2024: the top 100 marks per event for men and for women, 6,400 marks in 16 events. Each event's elite slice is its 30 highest scores.
MethodEach event's top-30 median set against the norm, the median of the 16 event medians (1247.5). A Kruskal-Wallis test across the 16 slices.
Compared withNo holdout. The baseline is the cross-event norm, and the null that all 16 slices share one distribution: rejected, H = 236, p < 10⁻⁴⁰.
ResultA 68.5-point spread: 5000 m 1272 (+24.5), javelin 1203.5 (−44). Seven of the eight events above the norm are running events.
LimitationTable bias and field depth are mixed together. A thin event scores lower even under a fair table, and this design can't tell the two apart.

From tables/fairness_by_event.csv; H and p from analyze.py on the git-ignored raw scrape.

Where does one score land?

Pick a WA score. Each event's dot is the median of its 30 best marks from 2023 and 2024; the line shows how far your score sits above or below it.

WA score 1250

Your score clears the event's elite medianFalls short of itDashed line: norm, 1247.5

Event medians from tables/fairness_by_event.csv, drawn in the browser. Men and women are pooled in each event's top 30, as in the analysis.

The question

A claim the sport's own data can test

The scoring tables sit under every cross-event comparison: a decathlete's javelin against their 1500 m, a shot putter against a miler for an award, or a country's strength across events, as in my national strength profiles. Any bias in the tables carries into all of them.

The World Athletics top lists publish the points each mark was awarded, so the check needs no recomputing. I scraped the top 100 performances per event for men and women in 2023 and 2024, 6,400 marks across 16 events, and took the 30 highest scores in each event as its elite slice. A Kruskal-Wallis test rejects the idea that the 16 slices come from one distribution (H = 236, p < 10⁻⁴⁰).

The result sheet

Sixteen events against the norm

Seven of the eight events above the norm are running events; the shot put is the exception. Seven of the eight below it are jumps and throws; the 800 m is the exception. Sort by any column. The last column follows the score you picked above.

				Dev
1	5000 m	distance	1272	+24.5
2	1500 m	middle	1260	+12.5
3	100 m	sprint	1259	+11.5
4	200 m	sprint	1258	+10.5
5	Marathon	distance	1254	+6.5
6	10,000 m	distance	1253.5	+6
7	400 m	sprint	1253	+5.5
8	Shot put	throw	1250	+2.5
9	Pole vault	jump	1245	−2.5
10	800 m	middle	1241.5	−6
11	Discus	throw	1236	−11.5
12	Long jump	jump	1224	−23.5
13	Triple jump	jump	1220	−27.5
14	High jump	jump	1219	−28.5
15	Hammer	throw	1208.5	−39
16	Javelin	throw	1203.5	−44

From tables/fairness_by_event.csv. The heavy rule marks where the events cross the norm.

Horizontal box plots of the top-30 WA scores for each of 16 events, sorted by median and colored by discipline group, with distance and sprint events to the right of the dashed norm line and javelin and hammer furthest left. — The spread behind each median: all 30 top scores per event as box plots, drawn by `analyze.py` from the raw scrape. The dashed line is the norm.

Limitations

What this doesn't prove

An event's elite scores depend on two things at once: how the table converts marks to points, and how deep the event's current field is. A thinly contested event shows a lower ceiling even under a perfectly fair table, because fewer athletes are pushing its front. Javelin and hammer may well be thinner fields, and this study can't separate the two causes.

So the supported reading is that equal points don't currently mean equal elite standing across events. Whether any one table is miscoded is a separate question. Answering it would need each table checked against a fixed frontier, such as the world record, and that is not done here.

The top 30 per event pools men and women, so one sex can fill most of an event's elite slice. Two seasons is a short window, and a different pair of years could reorder events near the norm. H and p come from the raw scrape, which is git-ignored, so only the per-event table is reproducible from the committed files.

In practice

What it changes in practice

A points-based shortlist for a cross-event award favors distance runners over throwers who stand just as high in their own events. The decathlon and heptathlon use the same kind of conversion, so a similar tilt there is worth testing, though this study only covers the individual-event tables. Any national-strength measure built on WA points, including mine, inherits whatever tilt the tables have.

How it was built

Where the numbers come from

Data: World Athletics outdoor top lists, top 100 per event, men and women, 2023 and 2024
Sample: 6,400 performances across 16 events
Points: The WA score each mark was awarded ("Results Score"), not recomputed
Elite slice: Top 30 scores per event; event median compared with the median of medians
Test: Kruskal-Wallis H test (SciPy) on the 16 top-30 distributions
Built with: Python, pandas, lxml, SciPy, matplotlib; the score strip and sortable sheet are plain SVG and JavaScript
Companion: National strength profiles, which uses these points