Losing about 2% of body mass is where performance starts to measurably decline. The useful question is not whether you will get there but how quickly — and at any realistic sweat rate the answer is sooner than most people expect. The curve below is your measured rate applied to a session of any length.
Over the length of a training session, essentially all the mass you lose is water. Fat oxidation accounts for a few grams; respiratory and metabolic losses are small. So the change on the scales, corrected for what you drank, is your sweat loss — and one litre of sweat weighs one kilogram.
sweat = (mass before − mass after) + fluid drunk − urine passed, all
in kilograms and litres, then divide by the hours. Weigh nude or in the same dry kit both
times — a sweat-soaked shirt can hold half a kilogram and will quietly halve your answer.
Every term in that formula is a control on this page, including the last one. A void is body water, so it comes off the scales exactly as sweat does — but counting it as sweat inflates the rate you plan the next session from. On the worked example below, a 400 mL void is the difference between 1.10 and 0.90 L/h: the uncorrected figure is 22% too high.
It does not change how dehydrated you got. The 1.4 kg on the scales is body water you no longer have whichever way it left, so the percentage stays where it is. Urine moves the sweat rate; it does not move the deficit. Those are two different questions and the page answers them separately.
2% of body mass is the point where endurance performance measurably falls. For a
70 kg athlete that is 1.4 kg. To stay inside it, drink at least
(sweat rate × hours) − (2% of body mass) — which for a 1.1 L/h rate
over two hours is 0.8 L, or 400 mL an hour.
70.0 kg before, 68.6 kg after, 0.8 L drunk, two hours. Sweat loss is 1.4 + 0.8 = 2.2 L, a rate of 1.1 L/h. The net 1.4 kg lost is exactly 2.0% of starting mass — right on the threshold. At that rate, drinking nothing at all would put you past 2% after only 1 h 16 m.
There is no evidence behind it. The figure appears to trace back to a 1945 recommendation that also said most of it comes from food — a qualifier that got dropped somewhere. Fluid needs vary with body size, diet, climate and activity, and most people regulate them adequately by drinking to thirst with meals.
That advice has been withdrawn for endurance events. Drinking well beyond sweat losses dilutes blood sodium and causes exercise-associated hyponatraemia, which has killed marathon runners — and the people most at risk are slower finishers with plenty of time to over-drink. Current guidance is to drink to thirst, and use a measured sweat rate to sanity-check that thirst is keeping up.
Because the spread genuinely is that wide — roughly 0.5 to 2.5 L/h between individuals doing the same work. Body size, fitness, heat acclimatisation and genetics all feed into it. That range is precisely why a single generic recommendation cannot fit everyone, and why measuring your own is worth one session.
Sweat sodium varies even more than sweat volume, from roughly 200 to 2,000 mg per litre. A heavy, salty sweater losing 2 L an hour can shed several grams of sodium in a long event, which is why sodium replacement matters more as duration increases. This page measures volume only and cannot tell you your sodium concentration.
That mass lost during exercise is essentially all sweat, which holds over a training session but drifts over very long events as substrate oxidation and respiratory water loss accumulate. It assumes accurate scales and the same clothing both times. Urine passed during the session is an input, because it leaves on the scales like sweat but is not sweat — ignoring a 400 mL void overstates a typical sweat rate by about 22%. Not modelled: sweat sodium concentration, glycogen-bound water, acclimatisation over weeks, individual gastric emptying limits, or any medical condition affecting fluid balance.
Releases in which this page changed, newest last. Derived from the archived copy of every release, not from notes written afterwards — so it reflects what actually shipped. Site-wide passes are left out; they are in the full changelog.