Negative Splits: What the Pacing Evidence Actually Says

The context

Almost every argument about pacing is an argument about the opening. Rowers have names for the shapes: a fast start that decays into a finish surge, an even split, a negative split where the second half is faster than the first, and the slow start older hands call conservative. The vocabulary for the version that goes wrong is older than the science: "fly and die" is starting out so fast that you cannot hold the pace, and paying for it in oxygen debt and lactate ([1] Concept2 2018).

So there are four shapes and a great deal of confident opinion about which to pick. The short version of the controlled evidence: the fast start is a convention rather than an advantage, the only shape with consistent evidence against it is a deliberately slow start, and everything else is a question about your training, your distance and your execution rather than about physiology.

The core idea

The convention is not an advantage. Split times were taken for every boat in every heavyweight race at the Olympic Games in 2000 and the World Championships in 2001 and 2002, plus the top 170 competitors at two British Indoor Rowing Championships. Of 1,612 on-water race profiles, 948 fitted the inclusion criteria of a genuine maximal effort ([2] Garland 2005). There were no differences in pacing profile between winners and losers, and none between men and women. Everyone did the same thing, and it did not separate a medal.

The average profile is the shape most people call negative. The first 500 m were rowed at 103.3 per cent of average race speed on the water, then at 99.0, 98.3 and 99.7 per cent. On the ergometer the same race was rowed at 101.5, 99.8, 99.0 and 99.7 per cent ([2] Garland 2005). The fast start is a house style rather than a weapon, and even the house style is not one thing: the ergometer fast start was the smaller one. The convention you inherit depends on where you race.

The controlled evidence is a null, and the null is not "even pacing wins". The best synthesis available is a systematic review and network meta-analysis of 24 studies, 302 participants and 765 observations across cycling, running and swimming ([3] Ramos-Campo 2026). There were no significant differences in performance between self-selected and imposed pacing strategies, with standardised mean differences below 0.20 and no significant differences in peak oxygen uptake, mean oxygen uptake or end-test lactate either ([3] Ramos-Campo 2026).

What that null does not say is that the shapes are interchangeable. The same review found that the effect of slow-start strategies was modulated by the duration of the initial start, and that prolonged slow starts were associated with impaired performance ([3] Ramos-Campo 2026). If you commit to a slow start, the race punishes you for how long you hold it.

The one rowing trial that separated the shapes went the other way. Fifteen male rowers aged 15.4 on average completed three 2,000 m ergometer races with the first 1,000 m manipulated: one self-selected, one negative and defined as starting four per cent slower than the self-selected mean velocity, and one even at that same mean velocity ([4] Dimakopoulou 2018). Performance time, stroke rate and force were all better in the self-selected and even conditions than in the negative one over the first 1,000 m, and the authors conclude that the self-selected profile covered the 2,000 m at a higher stroke rate and in a shorter time than the negative and even strategies, with the same physiological responses ([4] Dimakopoulou 2018).

Two caveats: the rowers were adolescents, and the negative condition was a slow start rather than the even-then-faster split most people mean. It is still the only rowing trial that separated the shapes, and it did not find the deliberate slow start the way to win.

Even pacing is not one instruction. The metabolic balance inside a race moves as the race goes on, and it moves at a different place in a 1,500 m than in a 2,000 m. A systematic review of 102 studies and 311 data points found that the anaerobic systems predominate in short-duration maximal exercise up to about 75 to 80 seconds, with equal contributions at 78.6 seconds ([5] Gastin 2026). The opening minute and a bit of a maximal effort is anaerobically dominant, and that is the window a start is really about.

Distance then changes the arithmetic. In 18 trained junior male rowers, mean power over 1,500 m was 5.2 per cent higher than over 2,000 m, and the anaerobic contribution rose from 17.3 to 22.2 per cent ([6] Astridge 2024). Rowers kept a consistent pacing strategy across the two distances, but the ones with the biggest power jump into the shorter race had the least 1,500 m pacing variance ([6] Astridge 2024). When the race is more anaerobic, an even opening is worth more.

And when the shapes were tested at 1,500 m, the answer was still a null. Thirty-three under-23 international rowers, 11 of them female, raced unprescribed, even and fast-start trials. Despite significant differences in pacing variance, mean power did not differ between them, every strategy still permitted a sprint finish, and prescriptions for the first 1,000 m did not influence performance ([7] Astridge 2026). Rowers kept the same reverse-J profile in their 1,500 m as in their 2,000 m, and the authors conclude a different shape may be needed to use the physiological capacity of the shorter distance.

The cost of a fast start is about when the bill arrives. Thirteen French rowers, four women and nine men at national and ex-international level, completed the first 1,500 m of a simulated 2,000 m race under three power-output strategies over distance: degressive, progressive and stable. There was no significant difference in aerobic energy contribution between the three, and final blood lactate and perceived effort were similar for all of them. What differed was timing: the rise in blood lactate and perceived effort came sooner under the degressive strategy, and the time spent at the higher values was longer. The rowers ranked degressive last ([8] Boillet 2022). That is the real cost of going out hard: the same work, the same aerobic contribution, and a bill that arrives early and stays.

The one genuine mechanism is real, and it comes from another sport at another duration. Eleven active men ran at the upper boundary of the severe-intensity domain. A fast start reduced the mean oxygen uptake response time from 22.2 to 19.3 seconds, let them cover 683 m against 670 m in a performance of similar duration, and raised exercise tolerance at a similar average speed from 114 to 125 seconds ([9] Turnes 2014). The authors describe running performances of two to three minutes. A 2,000 m piece is about seven minutes, and no rowing trial has measured it there.

What actually goes wrong is execution, and execution is the fixable part. The strongest causal evidence comes from cross-country skiing. Thirty-four skiers aged around 17 raced 7.5 km with self-selected pacing, and the 17 with the fastest start pace were told to reduce it to their own average pace from the day before ([10] Losnegard 2022). They slowed their start by 7.7 per cent, their heart rate fell from 83 to 81 per cent of maximum, their perceived effort dropped from 5 to 4, and they finished 1.9 per cent faster. Their improvement was larger than that of the conservative skiers, who changed nothing ([10] Losnegard 2022).

Read that carefully. The people who gained most were the ones starting too fast, and what worked was a deliberate change of shape rather than of fitness. A principal-component analysis of ten years of World Championship racing data for men's sculling put most of the variance in pacing strategy into age and the earliest splits ([11] Oliveira Borges 2025). The opening is where the race is decided, even if its shape is not what decides it.

The how-to

1. Decide the shape from the distance, not from the slogan. A 1,500 m race carries a bigger anaerobic share than a 2,000 m race, and the rowers who adapted to it were the ones who paced it most evenly ([6] Astridge 2024). Inside any piece the balance shifts again: for the first seventy-eight seconds of a maximal effort the anaerobic systems still dominate ([5] Gastin 2026). The decision is about the opening, and the opening is where the race is decided ([11] Oliveira Borges 2025).

2. Pick a shape you can actually hold. The manufacturer's own advice to racers is to practise different strategies and find what works, and the 2,000 m plan it publishes is a deliberate controlled opening, a racing start and ten hard strokes before settling onto goal pace, rather than a full-throttle first 500 m ([1] Concept2 2018).

3. Run the three shapes at the same distance, and record more than the time. The design that produced the only rowing metabolic comparison of shapes is simple enough to run without a laboratory: three pieces of the same distance, one degressive, one stable and one progressive, with the time recorded for each ([8] Boillet 2022).

4. Watch the shape of the effort, not only the finish. The degressive condition's signature is that strain arrives early and stays, and that is legible in a split log and in how the last third feels ([8] Boillet 2022). If the first third felt like a race already won and the third felt like a funeral, the physiology did not fail you. The execution did. And do not trust a single pair of pieces: physiological responses were indistinguishable between shapes even where the times were not.

5. If your opening is the problem, that is the good news. The people with the most to gain from changing their shape were the fast starters, and the change that worked was an instruction to hold a pace they had already proved they could hold ([10] Losnegard 2022).

The trade-offs

The honest limits matter, because the volume of confident advice here is far larger than the volume of controlled measurement. The rowing-specific evidence is small: 13 rowers for the comparison of shapes ([8] Boillet 2022), 15 adolescent males for the trial that separated the shapes ([4] Dimakopoulou 2018), 18 junior males for the energetic comparison of the distances ([6] Astridge 2024), and 33 under-23 international rowers for the pacing trial ([7] Astridge 2026).

The headline null also comes from other sports. The network meta-analysis pooled cycling, running and swimming ([3] Ramos-Campo 2026). A null result there means the strategies were indistinguishable in those studies, and it cannot tell you whether the field is measuring a small effect as noise or whether there is genuinely nothing there.

The clearest mechanism comes from running over two to three minutes ([9] Turnes 2014) and the strongest causal experiment from cross-country skiing ([10] Losnegard 2022). Both transfer plausibly and neither was measured in a rowing race. The striking oddity is that the trial with the best design found no difference, while the trial that found a difference had the weakest design and the least rowing-specific population. And a null is not a verdict on the convention: the 1,500 m trial's authors concluded a different shape may be needed ([7] Astridge 2026).

What to do tomorrow

Take your last 2,000 m split log and work out the shape you actually rowed: your first 500 m as a percentage of your average speed, and the same for the last 500 m. Most rowers will find they are already negative, and some that they are not as negative as they thought.

Then pick one distance and one alternative shape, and run both. If your default is a fast start, run one piece deliberately even. If your default is already even, run one piece that starts slower and finishes faster. Use the same machine, damper and rate target, and separate the pieces enough that you are not comparing a fresh piece with a tired one.

Write down more than the time. Note where the effort went, which 500 m felt hardest, and where the breathing broke. The shape that costs you is the one where the cost arrives early, and you can see that without equipment.

Then pick a shape and change only that. The fastest start is not the winning one, a slow start you cannot sustain is the one shape with evidence against it, and the rest is a decision about your own race rather than a rule you inherit ([2] Garland 2005).