An athletics mark is only credible when wind, altitude, shoes and splits tell the same story

On June 29, 2026, at the U.S. Olympic Trials in Eugene, Oregon, Tyson Gay ran the 100 metres in 9.68 seconds. The clock at Hayward Field produced the fastest time ever recorded over the distance. No world record was set that evening, because the anemometer read +4.1 metres per second, twice the legal limit. I rewound the footage at quarter speed. Walking one beat behind, I saw the race begin at the twelfth frame. The 9.68 told the truth about Tyson Gay's legs, and one more thing: those legs were pushed by a wind no other sprinter in the field was given at the same moment. Track and field is the only sport that prints its conditions next to the result. World Athletics sets the legal wind threshold at +2.0 m/s for the 100m, 200m, women's 100m hurdles and men's 110m hurdles. Above the threshold, a mark stays in the competition record but leaves the record books. In the long jump and triple jump, the gauge only averages the final five seconds of the approach, so two jumps minutes apart can carry completely different wind readings. Altitude is the second variable. The 2026 Mexico City Olympics took place at roughly 2,240 metres above sea level, where the air is nearly a quarter thinner. Jim Hines ran 9.9 hand-timed, equivalent to 9.95 electronic. Bob Beamon long-jumped 8.90 metres, a mark that stood 23 years until Barcelona 2026. Shoes are the third and youngest variable. Since 2026, World Athletics has capped sole thickness and required that a shoe model be available on the open market for at least four months before competitive use. The rule arrived after two seasons in which records fell simultaneously across sprints and distance events. Splits are the fourth variable, and the most ignored. A 100m result is only the last number of four differently structured phases: the start, the acceleration, the attainment of top speed, and the maintenance of it. Those four phases can be distributed in many ways, and the same final number does not mean the same way of running. Wind converts into time. Modelling published in sprint biomechanics puts each 1 m/s of tailwind at roughly 0.05 seconds over 100m, with the margin shifting by an athlete's height and mass. The gap between a run with +1.9 and a run with -0.5 therefore exceeds a tenth of a second. On August 4, 2026, in the Olympic 100m final in Paris, Noah Lyles and Kishane Thompson both finished in 9.79 seconds; gold was decided by 0.005 seconds. Half a metre per second of wind is enough to reverse the podium. Altitude converts too. Thinner air reduces drag, and the advantage concentrates in short events where drag makes up a large share of the total force the body must overcome. Rewatching meets held above 1,500 metres, in Nairobi or Colorado Springs, I have logged the same athlete running one to two tenths slower at sea level inside a single season, with training volume unchanged. That is observational data, and I label it as such. Shoes are the most contested variable. On August 3, 2026, in the Olympic 400m hurdles final in Tokyo, Karsten Warholm finished in 45.94 seconds and Rai Benjamin took silver in 46.17. Kevin Young's previous world record, set in 2026, was 46.70. A final that broke the old mark twice in one evening had never happened. On August 8, 2026, Sydney McLaughlin-Levrone ran 50.37 in Paris. The carbon-plated spike generation arrived in exactly that window. I have no laboratory to separate the shoe's contribution from nutrition, recovery and coaching method. One sentence I can state with confidence: a mark set in 2026 cannot be compared directly with a mark set in 2026 unless the result sheet names the equipment generation. Splits are where the race actually happens. On August 16, 2026, Usain Bolt ran 9.58 in Berlin into a +0.9 m/s wind. Published splits put him through 60m in about 6.31 seconds, with the 60-80m segment covered in roughly 1.61 seconds, the fastest such segment ever recorded by a human being. Read 9.58 and you know Bolt was fast. Read the 60-80m segment and you know how he won: he did not have the quickest start, he held top speed longer than anyone else. If I could keep only one data point about Bolt, I would keep the second one. On August 4, 2026, Lyles and Thompson both ran 9.79. The splits show Thompson leading at 60m and Lyles crossing first over the final ten metres. Two athletes, one result, two entirely different race structures. On August 8, 2026, Letsile Tebogo won the Olympic 200m in 19.46 seconds, becoming the first African athlete to take the event at the Games. The old reading called it a shock. The split reading shows he ran the best bend in the field, and had done the same thing weeks earlier. Then there is a category of data I file outside every comparison table: training marks. Every summer a phone video appears with a short caption about an unprecedented time. No anemometer, no certified track, no officials, no electronic timing. A training mark is data about who has been practising, not about who will deliver in a final with a crowd in the stands. The most reliable indicator of an athlete is not a season's best but the spread across starts. A runner who clocks 9.79 in a final after 9.83 in a semifinal and 9.86 in the heats carries a different profile from one who clocks 9.79 once and then goes over 10.10 three times. Same peak mark, two levels of confidence. In 2026 I got Luka Modrić wrong. That was the most honest analysis of my life. I mispronounced his name three times in one half of football, and the price was a month of rewatching footage at quarter speed to learn the phonetics of an entire tournament. My habit of cross-checking data before speaking started there, and it transfers intact to athletics. A misstep is another footprint on the same trajectory. I only trace it again. The story told most often over the past two years is that shoes and tracks are devaluing records. That fact holds. But the biggest reason athletics data is hard to use sits somewhere else: citation. A mark separated from its wind reading, altitude, equipment generation and splits becomes decorative, usable in a headline and useless in analysis. In print, a 9.86 run with a +1.8 wind is reported exactly like a 9.86 run with a -0.3 wind. The second blind spot is how events are framed. Fans judge athletes by one final; coaching staffs judge by an entire sequence of rounds. A peak on a June evening at home is not equivalent to a peak repeated three times in eight days at a championship. News desks routinely mix the two datasets together. Next time the clock shows something memorable, I will ask three questions before I memorise it: what did the anemometer read, at what altitude was the track, and which generation of shoe was on the feet. When the stadium is empty, I can hear the mark roll along every metre of the track. I started with the frame. Then I learned that the real game sits between the frames.