Shownotes

The real problem with FatMax

01:13 -

• This is a solo episode on why FatMax is one of the most overrated things to test and track in triathlon, and that holds even for Ironman athletes, contrary to a common belief in long-course circles.
• The episode title points at the real issue. The usual argument against FatMax is that carbohydrate is a more oxygen-efficient fuel than fat, so chasing fat burning is pointless for performance. That argument is correct and I come back to it at the end, but it is not the main point of the episode.
• The main point applies to every performance and physiological marker you will ever pay to measure, or measure for free at home: test-retest reliability and the coefficient of variation (CV).
  Most listeners have never tested FatMax and never will, so FatMax is just the worked example. The tools here let you decide for yourself whether any test is worth doing, or whether it is just noise dressed up as data. That includes FTP, VO2max, HRV and blood tests.
• I reference a fair number of CV values from various studies through the episode. I don't read the references out, but every study I take a number from is linked in the show notes.

What FatMax actually is

04:47 -

• When you exercise you burn a mix of carbohydrate and fat. Plot substrate oxidation against intensity and carbohydrate oxidation rises continuously as intensity increases. Fat oxidation starts low at rest, rises to a peak, then falls towards zero around and above threshold, where you run almost entirely on carbohydrate.
• Two different numbers come off the fat oxidation curve and they are easy to confuse. The height of the peak is the maximal rate of fat oxidation (MFO, sometimes PFO), measured in grams of fat per minute. That is a rate.
• The intensity at which that peak occurs is FatMax. It can be expressed as power, pace or percentage of VO2max. FatMax is an intensity, not a rate.
• Training FatMax usually means finding the intensity where you oxidise the most fat and doing a lot of work there, with the aim of pushing that intensity higher over time, for example from 100 watts to 120 or 140 watts.
• It can probably work to some degree, but classic endurance training gets you the same effect faster: volume at low intensity plus quality threshold and VO2max work. Threshold and VO2max are the rising tide that lifts all boats, and a higher VO2max comes with a higher FatMax. As rough illustrative numbers, a cyclist with a VO2max of 70 might have FatMax around 200 watts, and one at 85 around 280 watts.
• Testing, tracking and training FatMax was a notable theme in the low-carb Ironman space, treated as a way to hack long-course performance.

You have to trust a number before you train it

09:41 -

• Before you set out to improve a physiological number, you need to be able to trust it. If FatMax tested today differs from FatMax tested next week, you cannot tell whether your physiology changed or the test is just noisy. For FatMax, it is mostly noise.
• The coefficient of variation quantifies how much a test bounces around when repeated on the same person with nothing real having changed. It captures both biological noise (day-to-day differences in your body) and measurement noise (equipment, protocol), expressed as a percentage.
• A CV of 2% means that, with no real change, about 95% of your repeat results fall within roughly 2% of each other from noise alone. A CV of 20% means they can fall within 20%. Lower is better.
• My practical bands, as a coach rather than a sports scientist, are stricter than the textbook ones. Under 2% is good to excellent, with 1% being excellent. 2 to 5% is good to okay. Beyond 5%, things get shaky, because most measures we care about don't change enough for us to see a real change through that much noise. Sports science would call 5 to 10% medium and only beyond 10% bad.
• The logic: if a real 2% improvement in VO2max matters to you, for example 50 to 51 ml/kg/min, then you need a test with a CV below 2% to measure it. Otherwise the change is buried in noise.

The smallest detectable change

15:35 -

• The smallest detectable change (SDC) is CV multiplied by 2.77. It is the smallest difference between two tests you can be 95% confident is real rather than noise. The 2.77 is what applies that 95% confidence level. Any change smaller than the SDC might be real, but you cannot know.
• FTP is a good example. A good FTP test in experienced athletes has a CV of roughly 0.5 to 2%, lower in fitter and more experienced athletes, closer to 2% in more recreational ones. Using 1% for an experienced age-grouper, which I'm comfortable with, the SDC is 2.77 × 1%, about 2.8%.
• At an FTP of 300 watts that is about 8.3 watts. A later test of 308 or 292 watts could be noise, so you cannot say for certain you improved or declined. At 200 watts FTP and the same 1% CV the SDC is about 5.6 watts, so the absolute number you need to beat is smaller at lower FTPs.
• A practical aside: a happy, motivated athlete is a faster athlete, so there is no harm in taking a small gain as a win in your own head. The danger is in the decisions you draw from it. Doubling down on a training block because a sub-SDC change looks like proof it worked is where you have to be careful, whether you are self-coached or coaching others.
• One technical caveat: using 20-minute power as a stand-in for FTP as a true physiological measure introduces additional error, but that is a separate discussion. Either way, an FTP or 20-minute time trial carries a lot of signal and little noise, especially for experienced and fit athletes. Most other tests are noisier.

FTP versus FatMax

22:31 -

• FatMax CV varies a lot between studies, from about 3% in the best to 26% in the worst. The 26% figure comes from the study with by far the largest sample, 99 subjects, while the others ranged from 8 to 21 subjects. The same large study reported about 21% for the maximal rate of fat oxidation. Another, in 16 trained cyclists, found about 18% for FatMax. Several others sit around 5 to 11%.
• Weighting by sample size, the median is probably in the mid-teens. As a deliberately generous working estimate I use 10% for FatMax.
• At a 10% CV the SDC is 2.77 × 10%, about 28%. To be confident your FatMax genuinely changed you would need a change of more than 28%. At a FatMax of 200 watts that is 56 watts, which is not happening without years of training.
• Put that standard on FTP. If an FTP test had a 10% CV, the SDC would be 28%, about 84 watts at 300 watts FTP. You would not trust an improvement from 300 to 384 watts as real. Nobody would keep using a test that blind. That is exactly the standard we accept when we track FatMax.
• The headline problem with FatMax is not carbs versus fat. It is that any true change is buried in measurement error, so you have no way of knowing whether the number on your test report reflects a real change at all.

Reliable tests: performance trials, thresholds and VO2max

28:59 -

• You can apply the 2.77 × CV formula to any test using the CV values below. These are representative numbers from the literature rather than an exhaustive review, with sources in the show notes.
• The most reliable measures are actual performance tests where you go as hard as you can over a fixed distance or time. A well-designed cycling time trial (a 10-mile or 20-minute effort) has a CV of roughly 0.5 to 3%, often 1 to 2.5%, giving an SDC of about 2.8 to 6.9%. A 5 km treadmill running time trial sits around 1%, which is excellent. Peak speed at the end of a treadmill ramp test is around 1 to 2%.
• Time-to-exhaustion tests are far less reliable than time trials, even though they appear as outcome measures in many studies. In one study a 5 km time trial had a CV of 2% while time to exhaustion at 5 km speed had a CV of 15%. When you read research, put more weight on studies that found real time-trial improvements than on those resting on time-to-exhaustion results.
• A ramp test's end speed or power is reliable if you always use the same protocol, but derived and gas-exchange markers from the same test are not as clean. Don't treat them as equivalent.
• VO2max with a good protocol is around 2%, giving an SDC of about 5.6%. Some studies do report 5.6% gains from a training block, but that is exceptional. I don't think most athletes should be doing lab VO2max tests. The CV is more useful for reading the literature: a 2 to 3% group improvement is real as a group effect, but be careful extrapolating that magnitude to the gain you can expect as an individual.
• Power at maximal lactate steady state is around 3%. Lactate thresholds are around 3 to 3.5%, depending heavily on method. Avoid the DMAX method, which came in around 10%. I read lactate curves visually for the two inflection points. That is not ideal for research, where another observer might read it differently, but applied consistently to my own athletes it is reliable, and far better than the algorithmic transforms.
• Running economy is around 2 to 3.5% and gross efficiency on the bike around 4%. Both change very little within an individual, so even a 4% CV is larger than the realistic signal, particularly in well-trained athletes. The lesson is to judge CV against the size of the change you are actually trying to detect, not in isolation. More experienced athletes get smaller CVs but are also chasing smaller true changes, which is part of why experienced athletes tend to test less and beginners more.

Noisy markers: W', lactate, HRV and blood tests

39:02 -

• From a critical power test, anaerobic work capacity (W') has a CV around 8% with familiarisation and around 25% without it. Critical power itself is around 4% without familiarisation and 2.6% with, still larger than a pure time trial. A critical power test is really several time trials, so look at the individual efforts too. I often build a 20-minute test into the protocol to also capture the change in 20-minute power. More measures mean more chances of finding signal.
• Blood lactate concentration at a fixed intensity is very noisy. At maximal lactate steady state the power was 3% but the lactate concentration was about 16 to 17%. Another study found 27% for lactate concentration at fixed running speeds, as bad as FatMax or worse. This is why I don't recommend lactate spot-checking, and why fixed 2 and 4 mmol thresholds are a poor idea. A proper lactate step test is different: multiple measurements, interpreted through the inflection points rather than the absolute values.
• A single morning HRV reading is meaningless. Raw RMSSD from one reading can have a CV of 30%, an SDC over 80%. Best practice is to measure daily and track a seven-day rolling average, because averaging shrinks noise by the square root of the number of measurements. A weekly averaged HRV CV can drop to around 5%, which becomes a usable signal.
• The same averaging principle applies to aero testing. Either run multiple passes per configuration and average them, or make each run long enough to average out noise within it. Good aero testing does this (a shout-out to Bernardo at AeroEdge). Bad aero testing relies on single short runs, like a 1 km out-and-back or six laps of a velodrome, which is too short to be anything but noise.
• Blood markers carry a lot of noise. Ferritin is around 15% in males and 20 to 27% in females. Testosterone is around 9% in saliva and 17% in a blood draw. TSH is around 22%. A single value can swing widely on noise alone, which is why a good doctor, such as Jeff Sankoff, looks at symptoms and multiple tests over time rather than reacting to one number. The point is not to avoid these tests, but to interpret them in light of how noisy they are.
• For reference, ventilatory thresholds came in at 1.6 to 3.5% in one study, generally a bit more reliable than lactate thresholds, though it depends on protocol and equipment.

Practical takeaways

47:12 -

• For any test or marker you are considering, ask the provider or check the literature for its CV, then multiply by 2.77 to get the smallest detectable change. That tells you how much the measure has to move before you can rule out noise.
• If the SDC is small relative to the improvement you can realistically make, the marker is trackable. That covers FTP, 5 km time-trial performance, VO2max in a good lab, ventilatory thresholds and critical power.
• If the SDC is large, you have two options. Either accept that a single test is useless and measure repeatedly to track the average and trend, the way HRV and good aero testing work. Or, if repeated measurement is too expensive or invasive, accept that the marker is not suited to individual tracking. It can still be valid for population research, where studies average across many individuals.
• FatMax fails on both counts. The SDC is enormous, you will not test it often enough to average out the noise, and any real change tends to track alongside improvements in VO2max and threshold anyway. It is not a marker I would advise anyone to track as an individual.

Why FatMax fails even with a perfect test

50:24 -

• Set the reliability problem aside and imagine FatMax had a CV of zero. It still would not be the marker to chase, which closes the carbs-versus-fat loop.
• Your substrate stores are not what limit endurance performance. The limiter is how much oxygen you can deliver and use, your VO2 over the course of the race, which sits closer to maximum in a sprint than in an Ironman. In an Ironman, muscle breakdown becomes more of a limiter than oxygen, but substrate availability still is not the limiter, because you can carb load and fuel during the race. I can't recall an athlete describing anything bonk-like in an Ironman in years. When an Ironman goes wrong it is usually muscle breakdown, not running out of fuel.
• The key fact is that carbohydrate and fat are not equally efficient to oxidise. Carbohydrate yields about 5.0 kcal per litre of oxygen and fat about 4.74, so fat gives roughly 7% less energy per litre of oxygen.
• Fat-adaptation work in elite athletes, including the Louise Burke race-walker study, found that higher fat oxidation at race intensity came with a 5 to 8% higher oxygen cost. You go slower for the same effort and cap your performance ceiling lower, regardless of distance or level. Maximising fat oxidation at race intensity, which is the whole aim of chasing FatMax, works against you.
• FatMax intensity is also well below race intensity for most athletes, even for Ironman, so it is not a special intensity from a performance standpoint either.
• There are exceptions. An athlete who genuinely cannot take on enough fuel in racing, after training the gut and consulting registered dietitians, may be more limited by fuel intake than oxygen uptake, mainly in very long, low-intensity events. That can apply to Ironman and more so to long running or cycling ultras. Gut training is always the first port of call, but in those cases a bit more attention to fat oxidation at race pace makes sense. These edge cases are real but few.

Summary

56:05 -

• FatMax is not worth testing and tracking for triathlon performance, for two reasons. First and most important, the test is unreliable, with a CV around 10% or higher, so you would need very large changes to know a change is real. Second, even with a perfect test, fat is the less oxygen-efficient fuel, so a high FatMax is not an advantage on the race course, including in Ironman.
• The bigger takeaway is the method itself: take 2.77 times the coefficient of variation of any test or marker to get the smallest detectable change at 95% confidence, and use that to decide whether the test is worth doing or whether to throw it out.

Bibliography

• The day-to-day reliability of peak fat oxidation and FATMAX - Chrzanowski-Smith et al. 2020
• Reproducibility of Fat(max) and Fat Oxidation Rates during Exercise in Recreationally Trained Males - Croci et al. 2014
• Methodological issues related to maximal fat oxidation and FATmax reproducibility: a narrative review - Guevara et al. 2025
• The Reliability of a 5km Run Test on a Motorized Treadmill - Driller et al. 2016
• Reliability of Time-to-Exhaustion versus Time-Trial Running Tests in Runners - Laursen et al. 2007
• Test-retest reliability of a 16.1 km time trial in trained cyclists using the CompuTrainer ergometer - Sparks et al. 2016
• The reproducibility of 10 and 20 km time trial cycling performance in recreational cyclists, runners and team sport athletes - Borg et al. 2018
• Reliability of the Functional Threshold Power in Competitive Cyclists - Klitzke Borszcz et al. 2020
• Reliability of Mean Power Recorded During Indoor and Outdoor Self-Paced 40 km Cycling Time-Trials - Smith et al. 2001
• Reliability of peak running speeds obtained from different incremental treadmill protocols - Peserico et al. 2014
• Reliability of the parameters of the power-duration relationship using maximal effort time-trials under laboratory conditions - Triska et al. 2017
• Validity of 3 Protocols for Verifying VO2max - Kirkeberg et al. 2011
• Validity and Reliability of Ventilatory and Blood Lactate Thresholds in Well-Trained Cyclists - Pallarés et al. 2016
• Reliability of Maximal Lactate-Steady-State - Hause et al.r 2012
• The reliability of cycling efficiency - Moseley et al. 2001
• The reliability of running economy expressed as oxygen cost and energy cost in trained distance runners - Shaw et al. 2013
• Reliability and variability of running economy in elite distance runners - Saunders et al. 2004
• One-week test-retest recordings of resting cardiorespiratory data for reliability analysis - Schumann et al. 2025
• Day-to-day variation in iron-status indices in healthy men and women - Borel et al. 1991
• Within- and between-subject biological variations of follitropin, lutropin, testosterone, and sex-hormone-binding globulin in men - Valero-Politi et al. 1993
• Biological variation of thyroid stimulating hormone, free triiodothyronine and free thyroxine in healthy subjects in Turkey - Yildiz et al. 2025
• Carbohydrate and fat utilization during rest and physical activity - Melzer 2011
• Adaptation to a low carbohydrate high fat diet is rapid but impairs endurance exercise metabolism and performance despite enhanced glycogen availability - Burke 2020