At the beginning of last year, I took a "Closer Look at Performances" focusing on top times faster than a 1990 benchmark to identify any trends in the EPO-era.
Here is another, longer, and deeper look at all time best performances: a quantitative and qualitative summary of performances going as far back as 1960 to allow a richer identification of any trends. This was motivated by criticisms that a 1990 benchmark was the wrong one.
Events:
Men's events: 1500m, 5000m, and the Marathon
Source of Data:
I used all the data available at "alltime-athletics": OUTDOOR, INDOOR, DOPING, DQ, and OVERSIZE, only excluding times where the course was short. In effect, this means the performance cutoffs are:
1500m - 3:38 (8780 performances until 29-Mar-2019)
5000m - 13:30 (7457 performances until 30-Jan-2019)
Marathon - 2:10:30 (3376 performances until 30-Jan-2019)
Time periods:
4-year periods corresponding to Olympic cycles.
Note it does not include most of 2019 (and the effects of Vaporflys) nor 2020.
Measures:
Quantity: Total number of athletes, not performances
Quality: Average of the best times for the top-5 athletes; if there were less than 5, the average of all best times
Groups - I applied the same principle of creating groups by ethnic origin rather than citizenship:
1) WORLD - All athletes appearing in the all-time lists
2) EAST - Athletes originating from Kenya, Ethiopia, Uganda, Eritrea, Sudan, Djibouti, Tanzania, Somalia, Burundi, Congo, and Rwanda
3) NORTH - Athletes originating from Morocco, Algeria, and Tunisia
4) REST - Any and all athletes which are not included in EAST or NORTH groups: includes North and South America, Europe, Russia, Asia, Oceania, and West and South Africa
For each event and for each GROUP, I calculated:
1) The Quantity of athletes in each period (in parentheses)
2) The Quality of the best athletes
3) Total number of unique athletes in the list, removing duplicates