The finding. The best-known taper result, a 2-week taper with volume "exponentially decreased by 41-60%" (Bosquet and colleagues, 2007), does not say whether 41 to 60% is the level reached in race week or the average over the taper, and the two readings give different weeks. From a 40-mile peak, a race week 50% below peak averages only a 40% cut over the two weeks; an average cut of 50% needs a race week of 15.3 miles. And for a marathon, the race itself (26.2 miles) is longer than the whole race-week volume at any 50% cut from a peak below 52.4 miles.
What the meta-analysis found
Laurent Bosquet, Jonathan Montpetit, Denis Arvisais and Iñigo Mujika (2007, Medicine and Science in Sports and Exercise) pooled tapering studies in competitive athletes. Of 182 potential studies, 27 met their criteria: competitive athletes, a described reduction in training, and performance measured in competition or a field test. They looked at how cutting intensity, volume and frequency, and the shape and length of the taper, related to performance.
Their conclusion, from the abstract: the most efficient strategy was a taper of about two weeks in which training volume is exponentially reduced by 41 to 60%, without changing training intensity or frequency. In running terms: the same number of runs, some faster running kept in, each run shorter.
The studies mixed sports and levels, and the abstract reports effect sizes, not race times. Nothing below should be read as "this taper will make you X minutes faster".
What "exponential" means for weekly mileage
An exponential decrease takes the same percentage off each week, not the same number of miles. The taper calculator uses that shape: week k = peak x (1 - cut)^(k / weeks). With a two-week taper and a 50% cut, each week is 0.5^(1/2) = 0.707 of the one before, a drop of 29.3% per week. With three weeks, the factor is 0.5^(1/3) = 0.794, a drop of 20.6% per week.
How the calculator reads the cut is our interpretation, not a formula from the paper: it treats the chosen percentage as where race week lands, and spreads it evenly on the exponential curve. Here are the numbers it produces for three common peaks (all in miles; the same works in kilometres):
| Peak week | Cut | Week 1 | Race week | Two-week total | Average cut over the taper |
|---|---|---|---|---|---|
| 30 mi | 41% | 23.0 | 17.7 | 40.7 | 32% |
| 30 mi | 50% | 21.2 | 15.0 | 36.2 | 40% |
| 30 mi | 60% | 19.0 | 12.0 | 31.0 | 48% |
| 40 mi | 41% | 30.7 | 23.6 | 54.3 | 32% |
| 40 mi | 50% | 28.3 | 20.0 | 48.3 | 40% |
| 40 mi | 60% | 25.3 | 16.0 | 41.3 | 48% |
| 60 mi | 41% | 46.1 | 35.4 | 81.5 | 32% |
| 60 mi | 50% | 42.4 | 30.0 | 72.4 | 40% |
| 60 mi | 60% | 37.9 | 24.0 | 61.9 | 48% |
Worked example for the middle row: peak 40, cut 50%. Week 1 = 40 x 0.5^(1/2) = 40 x 0.7071 = 28.3 miles. Race week = 40 x 0.5 = 20.0 miles. Total 48.3 miles over two weeks, against 80 if you had kept training at peak, so the average cut over the taper is 1 - 48.3 / 80 = 40%.
The same 50% cut over three weeks:
| Peak week | Cut | Week 1 | Week 2 | Race week | Each week vs the one before |
|---|---|---|---|---|---|
| 30 mi | 50% | 23.8 | 18.9 | 15.0 | -20.6% |
| 40 mi | 50% | 31.7 | 25.2 | 20.0 | -20.6% |
| 60 mi | 50% | 47.6 | 37.8 | 30.0 | -20.6% |
Race-week level or taper average?
The abstract's phrase "decreased by 41-60%" fits two readings. If it is the race-week level, the calculator's table above applies. If it is the average across the two taper weeks, race week has to go deeper. I solved for the race-week level r that makes the two-week average (r^(1/2) + r) / 2 equal 59%, 50% and 40% of peak:
| Reading of "41 to 60%" | Week 1 (40 mi peak) | Race week (40 mi peak) |
|---|---|---|
| Race week ends 41% below peak (calculator) | 30.7 | 23.6 |
| Race week ends 60% below peak (calculator) | 25.3 | 16.0 |
| Two-week average 41% below peak | 27.8 | 19.4 |
| Two-week average 50% below peak | 24.7 | 15.3 |
| Two-week average 60% below peak | 21.0 | 11.0 |
The gap is not trivial: for a 40-mile runner choosing the middle of the range, race week is 20.0 miles in one reading and 15.3 in the other. I have not read the full paper's definition, so I do not claim either reading is what the authors meant. The calculator's option of 41, 50 or 60% covers a reasonable range either way: its deepest setting (60% in race week) averages 48% over the taper, which sits inside the 41 to 60% band under the second reading too.
The marathon problem: the race is bigger than race week
The calculator's last row is the 7 days that end on race day. If the race is counted inside that week, it works for a half marathon: from a 40-mile peak with a 50% cut, race week is 20.0 miles, so about 6.9 miles of running before the 13.1-mile race.
For a marathon it breaks down at ordinary mileage. The race alone is 26.2 miles, more than the whole race-week figure unless the peak is at least 26.2 / (1 - cut):
- 41% cut: a peak of 44.4 miles or more
- 50% cut: 52.4 miles or more
- 60% cut: 65.5 miles or more
For a 60-mile peak at 50%, race week is 30.0 miles, which leaves 3.8 miles before the race. For the 30- and 40-mile rows, the race-week number cannot include the race. My reading, for marathoners below those peaks, is to apply the race-week figure to the training days before the race and treat the marathon itself as outside the taper volume. That is a practical choice, not something the meta-analysis tested.
Keep intensity and frequency
The second half of the finding matters as much as the volume. The meta-analysis found the best results when intensity and frequency stayed the same. In practice: if you ran five days a week at peak, keep five days and make each run shorter; keep a short block at goal pace in the week before the race.
Here is the arithmetic for that, as an illustration rather than a rule from the study. A 40-mile peak over five runs is 8.0 miles per run on average. With a 50% cut over two weeks and five runs each week, week 1 averages 28.3 / 5 = 5.7 miles per run and race week 20.0 / 5 = 4.0. Cutting runs instead, say from five to three while keeping the same weekly totals, would leave the runs long (9.4 and 6.7 miles) and change frequency, which is the variable the meta-analysis found best left alone.
The calculator does not split the week into days; how you divide the total is your choice. Goal pace for each distance comes from the training pace calculator or the race charts.
Fuel goes the other way
As mileage falls, carbohydrate intake in the last day or two before a long race usually rises. For events over 90 minutes, the current recommendation quoted by a 2021 review is 10 to 12 g of carbohydrate per kg of body weight per day for 36 to 48 hours. For a 70 kg runner that is 700 to 840 g a day. The carb loading calculator does the arithmetic for your weight, and the race fueling calculator plans carbohydrate during the race. For medical conditions such as diabetes, follow your doctor or dietitian rather than a formula.
Limits
A taper cannot add fitness you did not build, and the 41 to 60% range is a pooled average across studies, sports and athletes. The weekly numbers here are arithmetic on that finding, applied in a way we chose and stated. For how much race-day pace a formula can promise in the first place, see why race predictors disagree.
Sources: Bosquet L, Montpetit J, Arvisais D, Mujika I (2007), Effects of tapering on performance: a meta-analysis, Med Sci Sports Exerc 39(8):1358-1365; Holtzman B, Ackerman KE (2021), Recommendations and nutritional considerations for female athletes, Sports Med (carbohydrate loading, citing Thomas, Erdman and Burke 2016). Rules checked . Found a problem? Tell us.
Questions
How much should I cut mileage during a marathon taper?
A 2007 meta-analysis by Bosquet and colleagues found the best results with a two-week taper that cut volume exponentially by 41 to 60%, keeping intensity and the number of sessions. From a 40-mile peak with a 50% cut that is about 28.3 miles, then 20 miles in race week.
How long should a taper be?
About two weeks, according to the same meta-analysis of 27 studies in competitive athletes. Some marathon plans use three weeks; the taper calculator shows both.
Should I run less often during a taper?
The meta-analysis found the best results when training frequency and intensity were kept the same and only volume was cut. Keep the same number of runs and make them shorter.
What does an exponential taper mean?
Each week drops by the same percentage rather than the same number of miles. With a 50% cut over two weeks, each week is about 29% lower than the one before.