Australian Swimming: The 0.87 Seconds That Never Appeared on the Scoreboard

There is one number I have kept in my notebook across four seasons: 0.87 seconds. It is the average gap between heats and finals in the women's 200m freestyle at Australian national championships over the past three years, measured across a sample of 186 swims. That number never appears on a scoreboard. It wins nobody a medal. But it is the number I use to predict who will break down in lane 4, and who will swim 1.2 seconds slower than they did twelve hours earlier.

Swimming is the sport where people assume data has already said everything. There are lanes, electronic timing accurate to the hundredth of a second, 50m splits, stroke rates, kick frequencies. No sport is more transparent. And yet at every Olympic Games, every World Aquatics Championships, there is always at least one case where the probability model is completely wrong. An unknown girl swims nearly two seconds faster than her personal best in a final. A reigning world champion swims half a second slower than in the heats and misses the final.

That is why I never finish an analysis without a limitations section. Data has a ceiling. And swimming's ceiling sits exactly where the scoreboard ends.

Context: Why Australian swimming became the harshest data laboratory

In six years of sports betting analysis, I have never seen a swimming system where the pressure on every single swim is as dense as in Australia. A 17-year-old athlete may have to swim four events in three days at a National Championship, each heat with its own time standard, and each final with three Olympic slots or four national team slots hanging in the balance.

This structure produces a type of data that European analysts often do not know how to read: the heat-to-final delta. I began collecting this metric in 2026, when a UK data company hired me to reassess their prediction model for women's events. The initial database had 2,400 swims, drawn from four Australian national championships, three Olympic Trials, and two World Aquatics Championships.

The results forced me to rewrite my entire set of assumptions. On average, a female athlete in my sample swims 0.42 seconds slower in the final than in the morning. But the distribution is uneven. Among athletes appearing in a national final for the first time, the average drop is 0.91 seconds. Among athletes with at least three prior national finals, the average drop is only 0.18 seconds. Among Olympic medallists, they swim on average 0.34 seconds faster than their heats.

Numbers have no gender, but the people who read them do. And in swimming, the reader is often a coach anxiously watching a protégé sit in the ready room.

Core analysis: A chain of evidence from four national seasons

I will take one specific case to clarify how to read this data. At the Australian National Championships, women's 400m freestyle, an 18-year-old athlete swam 4:08.12 in the heats, finishing 6th overall. She dropped 3.2 seconds from her season-best. The scoreboard placed her in lane 7 for the evening final.

The data I had at that moment: the swimming equivalent of PPDA — the average number of touches before an opponent begins to accelerate — does not apply to this sport. But I had three other things. First, the 100m opening and closing splits. Second, stroke rate and distance per stroke. Third, maximum heart rate recovery within three minutes of leaving the pool.

That 18-year-old swam the opening 100m with a stroke rate of 42.1 cycles per minute and a distance per stroke of 2.08m. This is the configuration of an energy-efficient swimmer. But over the closing 100m, her stroke rate dropped to 39.4 cycles per minute, while distance per stroke held at 2.05m. This means she did not accelerate by increasing cadence; she merely held distance through good technique. This is the signature of an athlete whose conditioning has not yet reached the threshold for two rounds of finals swimming.

In my 186-swim sample, athletes with a similar split pattern — where the closing 100m holds stroke rate but does not increase distance — are 2.7 times more likely to slow down in the final compared to the rest. Specifically in this case, I predicted she would swim 1.0 to 1.6 seconds slower in the final and finish outside the top 5.

Final result: 4:09.81. Slower by 1.69 seconds. Finished 6th.

This is not magic. It is the result of reading a chain of data that most people ignore because it is not printed on the official scoreboard. Organisers print times. They do not print stroke rate, distance per stroke, or the correlation between the two at each segment.

But here is what forced me to rewrite my model

If I had stopped there, I could have confidently claimed that swimming data is sufficient to predict finals. But at one recent Olympic Trials, I was wrong. And how I was wrong matters more than being right.

Again in the women's 400m freestyle, a 21-year-old swam 4:05.44 in the heats, finishing 4th. She had a stroke rate of 43.7 cycles per minute in the opening 100m and 41.2 in the closing 100m. Distance per stroke fell from 2.14m to 2.07m. According to my model, this was the signature of someone at peak conditioning who would lose about 0.6 to 1.0 seconds in the final.

Final: 4:03.18. She swam 2.26 seconds faster than the heats, broke her personal best, and won silver.

I sat in the stands and asked myself what I had missed. The answer came from a source data never records: she had swum the morning heat with a different purpose. She was not swimming to qualify. She was swimming to adjust her distance-per-stroke technique after changing her hand-rotation mechanics four weeks earlier.

Numbers have no gender. But numbers also have no intention. And ignoring the intention of the person producing the numbers is the most serious mistake in this profession.

I rewrote my entire model after that meet. I added a new variable: heat-swim context. Not every athlete swims a heat with the goal of qualifying. Some swim to conserve energy, some to test technique, some because the coach demanded a specific pace. These three situations produce three completely different data patterns, and if you read them the same way, you will be wrong as I was wrong.

Counterintuitive angle: When swimming data contradicts itself

Here is what I want to stress to anyone using data to evaluate swimming: the more accurate the metric, the easier it is to reach a wrong conclusion, if the reader does not understand the limits of the measurement.

Stroke rate is the clearest example. In my sample, the average stroke rate in the women's 400m freestyle final at Australian national championships is 42.3 cycles per minute. But this average conceals two completely different groups of athletes. The first group swims at a high cadence above 44 cycles per minute and short distance below 2.00m per stroke. The second group swims at a low cadence below 40 cycles per minute and long distance above 2.15m per stroke. Both groups have roughly similar average final times.

If you only read stroke rate and conclude that the athlete with the higher cadence is swimming better, you will be wrong about half the athletes in the sample.

This is why I always refuse to make a prediction based on a single metric, no matter whether it comes from Opta, Stats Perform, or the official measurement system of World Aquatics. I do not trust emotion. I trust a data chain longer than your emotion. But I also believe a data chain without context is meaningless.

One more thing swimming data often misses: pool effects. Not every pool produces the same conditions. Water temperature, depth, filtration systems, and even airflow above the surface all affect results, especially in sprint events. In an analysis I did in 2026, I found that the same athlete, on the same day, swimming in two different pools could differ by as much as 0.31 seconds in the 100m. This number appears in no official data sheet.

And this is where the analyst's personal memory becomes data. I have swum. I know the feeling when the water surface reflects light back into your eyes in a pool whose overhead lighting is not up to standard. I know the feeling when the water is heavier than normal because the temperature is 1.5 degrees Celsius below standard. This is not subjective feeling. It is physical data the scoreboard does not record.

Valuing swimmers: A lesson from a deal that failed

A few years ago, when I was asked to evaluate a young swimmer for a sponsorship programme, I applied exactly the process I use for football: collecting injury data, average weekly swim volume, finals-appearance frequency, and age-based growth patterns.

That athlete had a near-perfect data profile: 11 national finals in two years, an average improvement of 1.8 seconds per season in the 200m, and no shoulder-injury history. I recommended sponsorship.

Two seasons later, that athlete quit swimming.

The reason was in no data sheet. She told her coach she no longer wanted to swim. No injury. No technical problem. She had simply been swimming since age 6 and at 19 realised she had never done anything else.

After that case, I added a new metric to every evaluation: consecutive years of engagement with the sport. Not age, not results, but the number of consecutive years of high-intensity training. Among the 47 young athletes I tracked afterwards, those with more than 12 consecutive years of training had a 3.1 times higher probability of quitting before age 22 than the rest.

This is the kind of data nobody wants to publish. It is not pretty. It wins no medals. But it is the truth.

A view from two shores of the ocean

I grew up in Vietnam and work in Australia, but I will not use cultural difference to explain everything. When I compare youth swimming systems in two places, I use the same set of questions: training hours per week, competitions per year, age at specialisation, and the level of financial support from family.

In both places, I see the same pattern: the athletes with the fastest times are not the ones who train the most hours. They are the ones whose families do not have to worry about travel and accommodation costs during the season. The difference is not discipline. It is resources.

This is what sports data sheets often fail to record. And this is why I always add a limitations section to every piece. Not for self-defence. But to remind the reader that behind every number is a human being with a gender, with emotions, and who can die even when the probability is 99%.

Kazan was the day I learned that a 99% probability can still die on the betting table. That day, Germany controlled 74% of possession and lost 0-2. Every model was wrong. Every expert was silent. And I realised that the scoreboard is not the truth. It is only a tool.

Swimming is the same. A model predicting the women's 400m freestyle final has 78% accuracy in my sample. That means one out of five predictions is wrong. That 22% is not a failure of data. It is the limit of data. And a good analyst is not the one with the most accurate model. A good analyst is the one who knows when their model will be wrong.

What I am watching in the next round

Over the next three weeks, there are two things I will track closely. First, structural changes to the competition schedule at Australian national meets, where organisers are testing shortening the gap between heats and finals from eight hours to six. If the trial succeeds, heat-to-final delta data will change, and I will have to recalculate my entire model.

Second, the number of athletes under 18 competing in at least three events at a national meet. Across the past two seasons, that number is rising. In 2026 it was 14 athletes. In 2026 it was 19. If the trend continues, we will see a generation of swimmers trained to handle heavy loads from a very young age. That may produce better results, or it may produce a wave of quitting before age 20 that nobody predicts.

I do not know the answer. Nobody does. But I do know that if you only look at the results board, you will never see this question. You will only see times, rankings, and medals.

As for me, I look at the gap between two swims. That is where the real data lives.