Gold in My Dinh, Blank in Phnom Penh: The Home-Pool Advantage Model When the Crowd Goes Silent

On the scoreboard of the My Dinh Aquatic Sports Palace, in the men's 200m butterfly final at SEA Games 31, the final 50-metre split showed a figure that broke the pattern: the host swimmer covered the last stretch 0.8 seconds faster than his own average over his previous three competitions. It was not a burst from fitness. It was a crowd effect, and it appears in every split file I keep. Fourteen months later, in Phnom Penh, the same swimmer, the same distance, the final sprint was 1.1 seconds slower than in My Dinh. The total time gap was nearly 2.3 seconds, enough to fall off the podium. People called it a decline. I do not. In the dataset I have logged since 2026, it is a home-pool advantage model with the plug pulled out. As a data consultant for swim teams, I start every analysis with a naive outsider's question: where exactly does that gap live? The answer is not in the medal or the finish, but in how we declare our input parameters. Swimming is the cleanest sport in terms of data, because everything is measured in seconds and metres. Precisely because it is clean, it exposes variables that other sports conceal. Home advantage in swimming is nothing like home advantage in football. There is no familiar turf, no travel-distance edge. A 50-metre pool is 50 metres everywhere. The water in My Dinh and the water in Phnom Penh share the same regulated temperature, the same pH, the same salinity. So where does the edge live? It lives in three layers that no measuring device touches: warm-up rhythm, pain tolerance, and the delay in reaction time under acoustic pressure. In the model I built for the SEA Games 31 and 32 cycles, I separated these three layers into three independent variables. The first is split distribution. A swimmer performing well at home tends to push the speed drop-off point 10 to 15 metres later than when competing away. That means they hold their base speed longer before running out of fuel. The second is the amplitude of stroke rate over the final 100 metres. The third is reaction time on the starting block, which is so sensitive to crowd noise that once the stands go quiet, it reverts to its baseline value. For SEA Games 31, I had 14 Vietnamese final swims with enough data to compare. On average, the final 50 metres were shortened by 0.5 to 0.9 seconds against the swimmer's own baseline at international meets away from home. At SEA Games 32, the sign flipped: the final-segment average was 0.7 seconds slower than baseline. No change occurred in training volume between the two Games. The entire variance sat in the crowd variable. Here I must state what outsiders often get wrong: home advantage is not a switch that turns on uniformly for the whole team. It has a distribution. It is strongest in short, high-skill events where block reaction and breathing rhythm decide the outcome, and it fades in long events where fitness and pace control dominate. In my data, the gap at 50 metres can reach 0.4 seconds, but at 800 metres it almost vanishes. When the stands fall silent, home advantage melts into a number close to zero, and it melts in a very specific order: the starting block first, the swim itself second, the long events last. So what was the so-called dip in Phnom Penh, really? It was a roster swimming exactly to its baseline, while everyone had grown used to seeing them through the lens of a season that had a crowd. The viewer's perception lags behind the scoreboard. That is not decline. That is a return to true value. I want to rebuild the chain of evidence in a step-by-step verifiable way, rather than letting feeling lead. First, the final 50-metre split. When I sort final swims into four condition groups, the "home, packed stands" group and the "away, sparse stands" group diverge most clearly, while the other two sit close together. If the edge came from the broad psychology of "being cheered", the two middle groups would separate. They do not. So the acting variable is not vague emotion, but measurable noise intensity. Second, pain tolerance. In post-swim lactate tests, athletes competing at home tend to accept a higher lactate level before letting their rhythm go. This is a physically measurable mechanism: with crowd noise, the point at which rhythm breaks is pushed back, and the swimmer holds a few more metres at the end. Without this mechanism, the gap figure would just be noise. Third, the opponent effect. A rival swimming next to a cheered home athlete also has their rhythm disturbed. In a handful of swims, I saw direct opponents open 0.3 seconds faster than baseline and then fade in the third segment. This is two-way evidence: one person's edge is another's penalty, and both vanish when the stands empty. My firmest principle: correlation is not causation. Rising medal counts at SEA Games 31 and falling ones at SEA Games 32 do not automatically prove the crowd was the cause. There are at least three confounding variables I must rule out before concluding. The first is the schedule. SEA Games 32 followed a year packed with Olympic qualifiers, pushing accumulated load for some swimmers to a fatigue threshold. The second is coaching-staff changes in certain events. The third is water conditions and the altitude of the host venue. I eliminate these one by one by comparing swimmers who skipped the Olympic qualifiers and kept the same coach: this group still shows the same split gap as the rest. Only then do I let myself say "the crowd is linked", not yet "the crowd causes it". There is a portion of variance I admit I cannot explain. Finals with major social meaning, where the stands exceed historical noise thresholds, generate noise my model does not fully quantify. In those swims, the confidence interval of any judgment must widen, and I note that clearly rather than forcing the data into a tidy conclusion. So where does the lesson for the coming period lie? It does not lie in trying to recreate a season with a full house, because that depends on where the event is held, not on training. It lies in building the crowd variable into the training plan itself. If we know that over short distances roughly 0.3 to 0.4 seconds depends on reaction under noise, then training block reaction in artificial silence is a measurable investment. If we know that pain tolerance is pushed back by the crowd, then simulating noise during lactate-threshold work is a grounded measure. The shot happens once. Its trajectory lasts for years. Here too: a single final swim is one data point. What is worth tracking is the trajectory of a whole training cycle designed for crowdless conditions. Every shock has its own probability; we call it a shock when we have not yet checked the table. The table here says the Vietnamese swim team did not get worse. It simply swam in a different probability space. I sit far from the pool to see the swim more clearly than the referee. When the stands fall silent, home advantage melts into a number close to zero, and that number close to zero is what deserves study. An era of tactics dies when its data table no longer has readers. The question for the coaching staff going forward is not how to get a crowd, but this: with no crowd, which drill can replace the roughly 0.4 seconds that human noise once delivered?