La Défense 2026: When the Stands Returned, the Pool Still Answered in Data

On the night of the men's 400-metre individual medley final at La Défense Arena, the stands were so loud that reporters in row twelve had to shout into each other's ears. Underwater, Léon Marchand touched the wall. The scoreboard read 4:02.95. An Olympic record. A gold medal on home soil. And sitting right beside it, a fact almost no broadcast mentioned: a year earlier, in Fukuoka, Marchand had swum the 400 IM faster — 4:02.50, the world record. On the night all of France sang, its hero swam 0.45 seconds slower than himself. I write this not to diminish Marchand. Four gold medals and four Olympic records at a single Games is one of the greatest individual achievements in swimming history. But for someone who reads scoreboards for a living, 4:02.95 needs to be placed correctly. It does not deny the power of a crowd. It asks one question: how strong is the crowd, and by what mechanism? I have watched sports for more than twenty years, and I have learned something uncomfortable: most of what spectators believe about crowds has never been tested. We believe in the roar because we hear it. Belief is not a measurement. In a noisy arena, I still choose to sit with the numbers. My route into this question started in 2026, when the pandemic forced European football to play in empty stadiums. When the Bundesliga returned in May 2026, I had something sports analysts rarely get: a large natural experiment. I compared nine previous seasons against 93 matches without spectators. Home win rate fell from 41.3% to 34.7%. Average goals fell from 3.1 to 2.7. I wrote a twenty-page report, then delayed it two months trying to perfect the model, and nearly missed the news window. It eventually ran in an academic football journal. It taught me two things: always append a data-limitations section, and always set a deadline two days early. That study taught me something else, and it is the thing that haunted me all the way to Paris 2026: crowds do matter, but far less than the stories told about them. A 6.6 percentage-point shift sounds large. Broken down, most of it came from how referees handled contact, from match tempo, from how away teams chose risk — not from the noise itself. Noise is a surface variable. The real variables sit deeper. Before the pool, an older story. In 2026, aged 28 and a mid-level analyst in Miami, I was assigned an MLS season forecast. I spent two weeks building an xG model and found that Gerardo Martino's Atlanta United generated 0.21 xG per shot, the best in the league, yet the team was discounted as an expansion side. I wrote it up with charts. My editor rejected it, fearing readers would not follow. I posted it on my personal blog. A Belgian analyst shared it, and it drew more than 2,000 reads in 48 hours. When an editor says no, I learn to listen to the data. And I learned that a correct finding can still be rejected — not because it is wrong, but because it arrives before readers are ready. A year later, at the 2026 World Cup in Russia, an online football magazine brought me into its data team. While the newsroom fixated on Brazil and Germany, I analysed tracking data and found Croatia averaging a PPDA of 8.2, with Luka Modrić covering 10.6 km per match and showing almost no second-half decay. I wrote that Croatia would reach the final. Colleagues mocked it. When Croatia beat England in the semi-final, the desk apologised and republished my piece. Croatia reached the final before the media had read the numbers. Being right too early is a form of rejection too — but it is the kind I accept. I tell both stories to make my stance clear: I do not trust conclusions drawn from a single observation. And swimming, unfortunately, gives us roughly one observation every four years. Swimming at the Paris 2026 Olympic Games ran from 27 July to 4 August 2026 at La Défense Arena, a Racing 92 rugby stadium converted into a competition pool. Organisers quoted a capacity above 15,000 for finals sessions — the largest crowd swimming has ever competed in front of at an Olympics, at least by official figures. Thirty-five medal events: 17 men's, 17 women's, one mixed relay. Each final has eight lanes. But in any given event, the host nation typically has somewhere between zero and three genuine medal contenders. Multiply across 35 events and you get roughly 20 to 40 host-nation finals swims across an entire Games. That is the entire sample for measuring home advantage in Olympic swimming. And you get one Games per host nation per generation. n equals one. No confidence interval is narrow enough to support certainty. In other words, the home-advantage story in swimming is told with very little data and a great deal of emotion. So I want to redo the count from scratch, using three concrete numbers. Number one: 4:02.95 against 4:02.50. Marchand won the men's 400 IM in Paris in 4:02.95, an Olympic record. But his own world record, set at the 2026 World Championships in Fukuoka, was 4:02.50. He swam 0.45 seconds slower at home than he had a year earlier on neutral ground. In Paris, roughly 15,000 people sang his name. In Fukuoka, the arena was far quieter. The clock does not care who is singing. Number two: 1:54.06. That was Marchand's winning time in the 200 IM final in Paris, another Olympic record. The world record is 1:54.00, set by Ryan Lochte in 2026 in Shanghai. Marchand was 0.06 seconds away. Six hundredths of a second. A record that had stood thirteen years stood through the loudest night in swimming history. Number three: three. That is the number of world records set in the La Défense pool across the Games — Pan Zhanle's 46.40 in the men's 100 m freestyle, Bobby Finke's 14:30.67 in the men's 1500 m freestyle, and the United States women's 4x100 m medley relay in 3:49.63. None of them belonged to a host-nation swimmer. Three numbers, one direction: a packed pool did not produce the fastest swims. The fastest swims came from people who had flown to France. I want to pause here, because stopping at this point would make the conclusion too easy and too fast. A home hero swimming slower than his own record, and three world records going to foreigners, is not enough to say crowds do nothing. An athlete swimming slower than his own record at an Olympics is entirely normal: world records are usually set under near-perfect conditions — rivals pushing the pace, peak form, and a training cycle landing exactly on the right day. The Olympics is not always that day. Sometimes it is the day of maximum pressure. But if that is true, we need a cleaner comparison. And the cleanest comparison in recent swimming history comes from an Olympics with no crowd at all. Tokyo 2026, held in 2026, took place with essentially no spectators because of the pandemic. The arena was silent. If the roar hypothesis were right — if crowd noise genuinely pushes swimmers faster — Tokyo should have been a slow Games. It was not. Tokyo produced a cluster of world records. Tatjana Schoenmaker of South Africa swam 2:18.95 in the women's 200 m breaststroke. Caeleb Dressel of the United States swam 49.45 in the men's 100 m butterfly. Australia's women's 4x100 m freestyle relay set a world record of 3:29.69. The United States men's 4x100 m medley relay set a world record of 3:26.78. Great Britain's mixed 4x100 m medley relay set a world record of 3:37.58. And the host nation, Japan, in a silent pool, still won gold through Yui Ohashi in both the women's 400 IM and the women's 200 IM. This is the point I most want to stress. A silent pool produced a cluster of world records, and the host nation still won two individual titles. If the mechanism of home advantage were noise, both of those things could not happen at once. The arena was empty, but the numbers still knew how to score. The conclusion is not that crowds do nothing. The conclusion is that home advantage in swimming operates somewhere other than the ear. So where? I break it into five channels, ranked by how much I trust them. Channel one, and the one I trust most: host-cycle funding. A nation that knows it will host invests four, eight, even twelve years ahead. More full-time athletes. More dedicated coaches. More overseas training camps. This channel is measurable if you have federation budget data, and it explains both Tokyo 2026 and Paris 2026 without needing a crowd at all. Host nations do not win because crowds show up. Host nations win because they spent money before the crowds showed up. Channel two: venue access. Host athletes typically train in the actual competition pool before the Games — sometimes months before. They know the depth, whether the wall is hard or soft, how the water swirls in a given lane, how the light hits their eyes on a turn, how the starting block feels underfoot. These are real, measurable physical variables that directly affect times. In a sport decided by hundredths, knowing the pool is a concrete advantage, not a feeling. Channel three: selection. Host nations often hold more entries, and more importantly produce more A-standard qualifiers thanks to the investment in channel one. More entries means more chances to make a final, and more chances that one athlete swims the race of their life on the right night. This is pure statistics; no psychology required. Channel four: travel and circadian rhythm. Host athletes sleep at home, eat familiar food, train in their own time zone. Visiting athletes fly, shift time zones, adapt to hotels and communal dining. In a sport where peaking is planned to the hour, a six-hour shift is not trivial. But this channel is also the most easily exaggerated, because every major team now has nutritionists and sports-medicine staff. I rate it medium confidence. Channel five: media pressure. And this is the one I believe can carry a negative sign rather than a positive one. Marchand carried a nation in Paris. He handled it. But sports history is full of host favourites who collapsed under expectation. Home advantage and home disadvantage are two faces of the same variable, and we usually only count the favourable face. All five channels share one property: none of them involves sound. That is why I believe the crowd story in swimming is largely a story retold, not a story measured. But there is one more variable I have not mentioned, and it may be larger than all five. I call it the invisible referee. At Paris 2026, swimming finals were held in the evening, local time. At Tokyo 2026, finals were held in the morning — a change made to suit United States television. It was a scheduling change. Nobody touched the water. Nobody touched the technical rules. And it redistributed medals. The reason is simple: athletes build biological routines around session times. Someone who has spent four years peaking for a 10 a.m. final is disadvantaged if a final suddenly lands at 8 p.m. Someone accustomed to evening competition gains. This is the kind of variable I know well from esports: a small patch can change a champion without anyone becoming stronger or weaker in raw skill. A patch is an invisible referee. In swimming, the finals schedule is an invisible referee too. The direct consequence: comparing times between Tokyo 2026 and Paris 2026 is methodologically invalid unless you control for session time. Many sports articles ignored this. I have read dozens of comparisons between the two Games that never once mentioned that one was swum in the morning and the other at night. A second invisible-referee variable: the pool itself. Depth affects wave reflection off the floor, and wave reflection affects speed. Technical specialists have discussed whether La Défense was relatively shallow compared with some previous Olympic pools, and whether that made it a slow pool. I do not have official measurements in hand as I write, so I place this at hypothesis level, not conclusion level. But the principle is clear: if you want to compare times across Games, you must record depth, water temperature and session time before you compare. Otherwise you are comparing two different things. From here I want to widen the lens to the market, because home advantage is only half the story. The other half is the flow of people and money, and in swimming that flow behaves much like a transfer window. Swimming has no transfer window in the football sense. There are no transfer fees. But it has an equivalent: the coaching and training-group market. And that market just went through a major shift around the Paris cycle. Bob Bowman, the coach who guided Michael Phelps, moved from Arizona State University to the University of Texas. Léon Marchand, who had followed Bowman since his Arizona State days, stayed with him. A top athlete moved across a state line and a university programme to stay with his coach. In swimming, that is a transfer — just without a published fee. Every transfer is a problem waiting for a solution. The problem is also larger than the athlete. It is a problem about resources: a university programme with an Olympic-standard pool, a strength facility, a sports-medicine team, a scholarship budget. Bowman did not move to Texas for the weather. He moved for resources. In another corner of the market, FINA, now World Aquatics, changed how it pays athletes. At the Paris 2026 Olympic Games, according to the organisation's own announcement, swimming gold medallists received prize money of USD 50,000 each for the first time. That is a structural change: it makes staying in elite competition for another four years financially viable for a larger group of athletes. Meanwhile the International Swimming League — the club-format competition once expected to create a genuine transfer market for elite swimmers — ceased operations after 2026. Its departure left a gap: no independent club system now pays professional swimmers outside national federations and the United States collegiate system. As a result, the flow of people reverted to two old channels: national federations and universities. That explains why the swimming market moves less than football, and also why its moves carry more weight. One coach changing jobs can pull three or four elite athletes with him, and three or four elite athletes can reshape a four-year medal map. One more variable appeared in this cycle: name, image and likeness rights for United States college athletes. Once college athletes can earn from their own image and name, the decision to turn professional changes. There is more reason to stay at school for another year or two, and more reason to leave early if a professional training group pays better. It is a financial variable swimming has never had before. Australia runs the opposite model. Rather than spreading athletes across universities, elite swimmers cluster in small groups under one coach, usually tied to a local club and supported by the national federation. The talent supply chain is therefore different from the United States: narrower but more concentrated, and it allows specialisation by event. Comparing these three models — the United States collegiate system, Australia's concentrated training groups, and Europe's national-federation model — gives a different read on Paris 2026. France's pool performance fits none of them in the usual way. France lacks the United States' collegiate sporting machine and Australia's concentrated groups. Its Paris success concentrated on one individual, one foreign coach and one relay. It is a single-point-breakthrough model, not a depth model. And that matters, because if you want to measure home advantage, you must separate two very different things: a nation with home advantage in its system, and a nation with home advantage in one person. In Paris we had both at once, and we cannot pull them apart. That is why I want to spend the rest of this piece on the counterintuitive argument: home advantage in swimming may be understood wrong at the level of mechanism, and our measurements of it are almost never methodologically strong enough to bear the conclusions the media draws. First, the simplest correlation-to-causation error. We observe that the host nation won more medals than expected, and we observe that the stands were full. Both observations travel together, but travelling together does not mean one caused the other. The true cause may lie in the budget spent four years earlier, in venue access, in the selection system — all of which existed before a single spectator entered La Défense. Second, the sample-size problem, and this is the most serious. As calculated earlier, we have roughly 20 to 40 host-nation finals swims in a Games, but only one Games per host nation. No statistical method can extract certainty from that structure. Anyone who states precisely what percentage of host medals the crowd contributed is speaking beyond what the data can carry. Third, the physics. Sound travels through air, not efficiently through water. For most of a race, a swimmer's head is at or below the surface. What they hear is distorted, delayed, mixed with water. More importantly, the arena PA plays identically for every lane. If noise mattered, it would have to affect home and visiting swimmers differently to mean anything. Sound in an indoor arena does not check nationality. Fourth, the paradox of silence. Tokyo 2026 was a natural experiment we wasted. A Games with no spectators produced a cluster of world records and two individual host-nation golds through Yui Ohashi. If the mechanism were acoustic, both are hard to explain. If the mechanism is institutional, both are easy: Japan also invested on a host cycle, also had athletes training in the competition pool, also held more entries. Fifth, and the point I most want to stress in this rebuttal: the error in the opposite direction is still an error. Some writers now claim crowds do nothing. That conclusion also exceeds the data. With n equals one, an honest analyst can only say: we have not yet separated the systemic variable from the crowd variable. We do not know. I do not argue with emotion; I present a chain of data — and the current chain stops where it stops. Sixth, the human element. Behind every figure is a person sweating. 4:02.95 was not a failure. It was a 22-year-old swimming in front of a home crowd with a nation on his shoulders, still winning four gold medals and setting four Olympic records. When I write that he was 0.45 seconds slower than himself, I am not saying he got worse. I am saying the clock is one instrument among many, and there are other instruments we do not have. This is also where an older story returns. In 2026, tracking Leeds United through the summer window, the data showed Kalvin Phillips had dropped from 18.4 to 14.1 successful presses per 90 minutes after injury, while RB Leipzig's Tyler Adams was at 17.8. When word emerged that Leeds might sell Phillips, the big outlets hesitated. I worked with a European data broker for cross-verification and was among the first to report that Leeds would sign Adams and sell Phillips to Manchester City for a fee reported at GBP 45 million. Everything then unfolded as the data suggested. But the lesson was not that data is always right. The lesson was that data only has value when you say clearly where it is uncertain. Back to Paris. My central question is not whether crowds matter. My central question is: if home advantage in swimming comes mainly from systems rather than sound, which variable are we measuring wrong? And if we are measuring wrong, the consequences will show up at Los Angeles 2028. Three signals to track for the next cycle. Signal one: count finalists, not golds. If the systemic model is right, the United States' home advantage at Los Angeles 2028 will show up most clearly in depth — the number of American swimmers reaching the eight-lane final in each event, the conversion of semi-final places into final places, and relay performance. Counting golds is the wrong count, because golds are dominated by a handful of exceptional individuals and by luck. Signal two: record venue parameters before the Games, not after. Depth, water temperature, filtration and flow direction should be logged and published before competition starts. If we log them afterwards, we will always find a reason to explain results the way we wanted. Signal three: watch the coaching market over the next two years. If a new professional training group forms in the United States or Europe, or if a new club league tries to fill the void the International Swimming League left, the flow of athletes will change — and when the flow changes, the next cycle's medal map changes before anyone enters the water. I want to close with a small observation. The race ends, but the data still plays stoppage time. That night at La Défense, I sat in row twelve and wrote down four lines: the time, the crowd, the session hour, and who set the world record. The first three are measurable. The fourth is the only one I would bet on: three world records in the Paris pool, none of them belonging to a host-nation swimmer. That does not deny Marchand's magical night. It only reminds me that a loud crowd is the easiest part of a story to tell, and the easiest part is rarely the truest. The next time a pool is full and a local hero steps onto the blocks, I will still be sitting there with the numbers, and I will still ask one question: whose swim was fastest tonight, and did that person hear the singing at all. Data limitations for this piece: the observation sample for home advantage in Olympic swimming is one Games per host nation, so all causal claims remain at hypothesis level. Technical parameters of the La Défense pool, such as depth and water temperature, were not verified against official documentation at the time of writing, so related material is presented as hypothesis. Federation budget figures across host cycles were excluded from the model for lack of comparable cross-national data. One match does not make a trend, and neither does one Olympic Games.