Systematic reviews in sport and exercise science go stale the moment they're published, and new studies outpace the reviews that should synthesise them. Training.Science is a living and evolving web of network meta-analyses — automated searching, screening, extraction, and analysis — that keeps the comparisons between training interventions continuously current.
Where a traditional meta-analysis pools evidence for one treatment against another, a network meta-analysis connects every treatment in a web of direct and indirect comparisons, draws on studies that never tested two protocols head-to-head, and ranks them all against each other.
An open, queryable source of truth for athletes, coaches, researchers and anyone who wants to add to or audit the evidence — rather than a snapshot frozen in time. Free to read, free to use, designed to be built on; the full pipeline source is being prepared for open-source release on methodology paper publication.
Six edges in the index — 3 in production, 3 pre-specified. Each node represents a measurable outcome; each edge will be a network meta-analysis ranking every treatment against every other. Nothing here is published yet — VO₂max leads, currently moving through extraction.
Every edge moves through the same five-stage pipeline. The judgement-heavy stages — abstract screening, full-text screening, data extraction — are each carried out independently by two AI reviewers from different model families, working blind to each other's decisions. Disagreements are then resolved at a structured consensus roundtable, with a chair adjudicating any that the reviewers can't converge on. Only agreed or adjudicated decisions enter the analysis. Automated steps (search, deduplication, validation, arm classification, NMA computation) run without human intervention. Every state change is checkpointed and version-controlled.
Six edges in the index. Each one a network meta-analysis, each one updated monthly. 3 in production — VO₂max furthest along, in extraction — and the remaining 3 pre-specified, queued.
Network meta-analysis of structured exercise interventions on maximal oxygen uptake.
Energy cost of locomotion under exercise interventions.
Sub-maximal blood lactate breakpoints under training interventions.
A network meta-analysis can compare interventions that have never been studied head-to-head in a single trial — by drawing on the studies that connect them through shared comparators.
Training.Science is a living publication. Cite the edge you used, with the access date — content evolves between sync cycles.
Training.Science. A living review of training interventions in sport and exercise science. Issue 01, May 2026. doi:ts·2026·01. Available at training.science. Accessed [date].
DXA-validated fat-mass and lean-mass change across training paradigms.