The endurance training argument has a Bayesian answer, and it is 'no clear winner'

Training intensity distribution describes how an athlete's total training time is allocated across low-, moderate- and high-intensity zones [s1]. It is among the most-argued variables in endurance training, and the argument has a favourite: polarised training, in which the great majority of the work sits at low intensity and the remainder sits high, with little in between.

A Bayesian network meta-analysis published in the Journal of Strength and Conditioning Research put the competing models side by side. The result is less a verdict than an admission [s1].

What the analysis compared

The authors retrieved randomised controlled trials from PubMed, Web of Science, EBSCO, the Cochrane Library and Scopus, assessed risk of bias with the Cochrane tool, and compared the common distribution models on two outcomes: maximal oxygen uptake, and time-trial performance [s1].

Compared with polarised training, no other model showed a definite advantage for improving either outcome — because all 95% credible intervals crossed zero [s1].

That sentence is the headline finding, and it cuts in every direction at once. Polarised training did not beat the alternatives. The alternatives did not beat polarised training. On the evidence assembled here, the distinction that generates the most argument generates the least measurable difference.

The ranking, and what it is worth

Bayesian analyses also produce posterior rankings, which express how likely each option is to be best rather than whether it beats the others. Here the rankings did separate.

Lactate-threshold training was most likely to be the optimal model for maximal oxygen uptake, with a rank-1 probability of 65.4% and a SUCRA of 84.8% [s1]. High-intensity interval training was most likely to optimise time-trial performance, with a rank-1 probability of 53.9% and a SUCRA of 81.5% [s1].

Note what that implies. The model ranked first for aerobic capacity is threshold work — the moderate-intensity zone that polarised training exists specifically to minimise.

A ranking is not a result. When every credible interval crosses zero, the rank ordering describes which option the data lean toward, not which option is better. The authors' own framing is conditional: threshold and interval training emerge as the most promising options for those two targets respectively, and coaches should tailor distribution to athlete characteristics and sport demands [s1].

They also identified age, training status, the method used to quantify intensity, intervention duration and sex as potential moderators, and their cluster analysis found both shared and sport-specific distribution patterns [s1]. A field where five variables plausibly moderate the effect, and where studies define intensity zones by different methods, is a field where a null network meta-analysis is close to the expected outcome.

What people actually do

Separately, a study in the International Journal of Sports Physiology and Performance tracked what recreational runners do when nobody is prescribing their intensity distribution [s2].

Forty-eight recreational runners — 22 preparing for a marathon, 26 for a half-marathon — were monitored across the 12 weeks before their races using GPS and heart rate monitors, with training sorted into three intensity zones and internal load calculated as individualised training impulse [s2].

Both groups adopted a pyramidal distribution [s2]. Across the full sample, most training was performed in Zone 1 (69.1%, SD 8.5%), followed by Zone 2 (17.3%, SD 5.2%) and Zone 3 (13.6%, SD 5.3%) [s2]. There were no significant differences between the marathon and half-marathon groups in weekly training impulse, which averaged 911 arbitrary units (SD 210), or in session load [s2].

Pyramidal, not polarised — a descending staircase of volume rather than a barbell. Recreational runners, left alone, converge on a distribution that sits between the models the trials keep comparing.

What this does and does not settle

Two cautions belong on this reading.

The first is that zone labels are not portable. Studies define intensity boundaries by different physiological anchors, and the meta-analysis lists the intensity-quantification method as a moderator of the results [s1]. A percentage of training time in "Zone 2" means different things in different protocols.

The second is that both papers concern trained and recreational endurance athletes pursuing performance outcomes. Neither speaks to health outcomes, and neither followed anyone long enough to say anything about durability of adaptation beyond the intervention periods studied.

What the evidence supports is narrower than the argument it is used to settle: across the trials run so far, how you distribute intensity matters less than the certainty of the debate suggests, and the distribution most likely to raise aerobic capacity is the one polarised training was designed to avoid.

This article is informational and is not medical advice.