On July 26, 1855 in London, Charles Westhall set a new record for the fastest time in the one mile run with a 4:28 time. In 1991, Dr. Michael Joyner published a paper proposing a mathematical model for marathon performance:
A variety of combinations of these variables from elite runners results in estimated running times that are significantly faster than the current world record (2:06:50). The fastest time for the marathon predicted by this model is 1:57:58 in a hypothetical subject...with exceptional running economy.
Dr. Joyner's theoretical max for the marathon requires a pace of 4:30 per mile, nearly the same time as the world record mile in 1855. In 2017, Nike sponsored an attempt to run a two-hour marathon in controlled conditions. Eliud Kipchoge just missed the goal, with a time of 2:00:25. In advance of that attempt, Wired Magazine produced a video showing how difficult this feat really is.
Today, in a similarly controlled event in Vienna, Mr. Kipchoge ran the marathon in 1:59:40. His time does not count as a world record, because it was not run in a real race, which seems fair to me. There were no other competitors in the field, forty-one professional runners rotated on and off the course as pacers and formed an aerodynamic pocket around Kipchoge, and an electric timing car projected the ideal path on the road with green lasers.
Mr. Kipchoge holds the world record in the marathon: 2:01:39 set in 2018 in Berlin. It seems inevitable that the two-hour barrier falls in a competitive race.