172+ scientific papers. One coach who read them all.

What actually reduces running injury risk, and what does not.

This is the largest section of the library, and the one where the evidence is most uncomfortable for the running industry. The consistent finding across systematic reviews here is that strength training reduces injury rates in a dose-dependent way, while stretching and most gait-retraining interventions show far weaker effects than their popularity suggests. Training history and load errors predict injury better than any single biomechanical measurement.

Where the literature still argues: whether tendon pathology follows a true continuum, how much asymptomatic imaging findings matter in runners who feel fine, and which risk factors are genuinely causal rather than merely correlated. Leo uses this section conservatively. When a body signal you report matches a documented pattern, he cites the evidence and grades his response, and he refers you to a clinician rather than guessing when the picture crosses into medical territory.

Fuelling, hydration and supplements, from position stands rather than marketing.

This section is built on position stands rather than single trials: the International Society of Sports Nutrition and IOC consensus documents anchor most of it, supplemented by reviews on carbohydrate periodization, fat adaptation and gut training. That matters, because nutrition is the field where isolated studies are most often quoted out of context to sell something.

The consensus is firmer than most runners assume on carbohydrate intake during long efforts, on caffeine, and on the fact that the gut is trainable. It is far less settled on fat adaptation and on low-carbohydrate approaches for endurance, where the reviews here are genuinely mixed. Leo prescribes race fuelling from the consensus range adjusted to your weight, expected duration and your own reported gut tolerance, and he defers to a registered dietitian for anything resembling a full meal plan.

Critical speed, running economy, and how finish times are actually predicted.

Two threads run through this section. The first is the machinery of endurance: running economy, the critical power and critical speed models, the long-running argument over what the anaerobic threshold even is. The second is prediction, with a set of papers on estimating half-marathon and marathon times from training data rather than from a recent race.

That second thread is the one Leo leans on most, because it is what lets him give you a finish-time range without asking you to race first. The papers here are also honest about the limits: prediction degrades badly when training volume is low, when the terrain changes, and in the specific case of late-race collapse, which one large-scale analysis in this section treats as its own phenomenon rather than as a pacing error. Leo therefore presents estimates as ranges with a stated reliability, never as a single confident number.

Intensity distribution, periodization, tapering and concurrent training.

The backbone of how Leo builds a plan. The meta-analyses here converge on a few things: tapering produces a real and quantifiable performance gain, strength work improves running economy without compromising endurance when sequenced properly, and high-intensity intervals and continuous training both raise VO2max by comparable amounts when total work is matched.

The live argument is intensity distribution. Several reviews in this section compare polarized training against pyramidal and threshold-heavy models, and the honest summary is that polarized wins in some populations and ties in others rather than being universally superior. Periodization itself is questioned here too, in a paper asking whether the classic models are evidence-led or simply inherited. Leo builds phases from these frameworks but treats the intensity split as a starting point to be adjusted against what your own data shows, not as doctrine.

Biomechanics, foot strike, and what the shoe evidence really supports.

Mostly a footwear section, and a usefully deflationary one. A Cochrane review here on running shoes for injury prevention, alongside randomised trials on motion-control shoes and minimalist transitions, gives a picture much more cautious than any shoe wall would suggest: there is stronger evidence that footwear changes running economy than that it prevents injury.

The carbon-plated racing shoe literature is the clearest signal in this section, with several studies converging on measurable economy gains. Foot strike remains contested, and the papers here do not support telling a healthy runner to change theirs. Leo applies this directly in shoe wear analysis: he tracks mileage and inspects wear zones because wear patterns carry real biomechanical information, but he will not prescribe a form change you did not ask for on the strength of evidence this thin.

Cardiac screening, RED-S, hyponatremia, rhabdomyolysis and overtraining syndrome.

The section where Leo's job is to recognise a pattern and then step back. It holds formal guidelines rather than exploratory research: ESC sports cardiology, the AHA and ACC statement on competing with cardiovascular abnormalities, the IOC consensus on Relative Energy Deficiency in Sport with its clinical assessment tool, the international consensus on exercise-associated hyponatremia, and the joint ECSS and ACSM statement on overtraining syndrome.

These documents share a structure: named warning signs, and a clear threshold past which the answer is a clinician, not a training adjustment. Leo uses them exactly that way. If what you describe matches a documented red flag, he names it, cites the guideline, and tells you to see a doctor. He does not diagnose, and he will not coach around a symptom that belongs in a consulting room. This section is also why his overtraining response is graduated rather than binary.

Strength, cycling and CrossFit as complements to running, with their real risks.

Two distinct bodies of evidence sit here. The first covers strength work for distance runners, where the meta-analyses are unusually consistent: it improves running economy, and the order in which you sequence strength and endurance within a week changes the outcome. The second is a cluster of systematic reviews on CrossFit, covering both its benefits and its injury epidemiology, plus comparative physiology of running against cycling in trained athletes.

That second group is why Leo treats cross-training as a decision rather than a filler. A ride and a run are not interchangeable stimuli, and the CrossFit injury literature here is specific enough that volume matters. Every non-running session in your plan gets an explicit role, recovery, aerobic development, muscular reinforcement or injury prevention, and it is counted in your weekly load rather than sitting outside it.

Sex differences, menstrual cycle, postpartum return and youth running.

Evidence that a single average runner does not exist. This section covers physiological sex differences in the response to exercise, VO2max and their consequences for endurance performance, the effect of the menstrual cycle and oral contraception on training response, return to running postpartum for elite and sub-elite athletes, and a consensus statement on youth running.

The menstrual cycle papers here deserve a caveat that the section itself makes: the meta-analytic effect on performance is small and the individual variation is large, so population-level findings translate poorly into individual prescriptions. Leo's use of this section is deliberately restrained. It informs how he reads your data and what he does not assume, rather than producing a different plan template based on who you are.

Mental fatigue, motivation, athlete mental health and the running-wellbeing link.

A small but load-bearing section. Its most directly usable finding is that mental fatigue measurably impairs physical performance, which means a hard day at work is a legitimate input to today's session rather than an excuse. Alongside it sit brain endurance training as an emerging method, self-determination theory applied to exercise adherence, and mental toughness studied specifically in trail runners.

The rest is athlete mental health: the IOC consensus statement and its screening tools, and guidelines for supporting athletes at competitions. Leo does not run those screening tools on you and does not attempt any form of mental health assessment. What this section gives him is the self-determination work, which is why he asks rather than instructs, and why he will not use guilt as a motivator: the evidence says pressure-based coaching produces resistance, not adherence.

ACWR, session-RPE, and the honest limits of load monitoring.

The most self-critical section in the library. It contains the consensus statement on monitoring athlete training loads and the validation work on session-RPE, but it also contains a paper explicitly on the conceptual and methodological pitfalls of load research, and a recent meta-analysis testing whether the acute-to-chronic workload ratio actually predicts injury.

It largely does not, at least not with the confidence with which it is marketed. That finding shapes how Leo works: he tracks your load and watches the trend, but he does not hand you an ACWR number and call it a risk score. Load is one input among several, cross-referenced with rest days, how you say you feel, and recurring body signals, precisely because this section shows that any single metric used alone will mislead you.

Downhill damage, mountain-running injuries and predicting trail performance.

Small, specialised, and the reason Leo does not simply apply road logic to the mountains. The papers here cover injury epidemiology in trail runners and a five-season study of mountain running, a narrative review on the effects of downhill running and how to adapt to them, and work on predicting short trail race performance from lactate thresholds.

Descent is the thread. Downhill running damages muscle in a way that flat running does not, which is why time-based rather than distance-based recovery makes sense after a mountain race, and why Leo excludes trail results from road-based prediction models instead of letting them pollute your VDOT. Trail estimates are given as ranges adjusted for elevation, because this section is honest that technicality and fatigue accumulation are not yet well captured by any published model.

Heat acclimation and hypoxic training, with realistic timelines.

Four documents, each carrying a practical protocol: the mechanisms and applications of human heat acclimation, consensus recommendations for training and competing in the heat, a meta-analysis of sea-level performance after altitude adaptation, and work on combining hypoxic methods to peak.

What makes this section useful is that it is specific about time. Heat acclimation follows a known course over roughly one to two weeks and decays when you stop, and the altitude meta-analysis is measured about how much sea-level benefit to actually expect. Leo uses it when your race conditions differ from your training conditions, which is more often than runners plan for. If your goal race is hot and your build is not, that is a gap in the preparation, and it is one of the few that can be closed cheaply in the final fortnight.

What sleep loss costs, and which recovery methods survive scrutiny.

The smallest section, and deliberately so, because the honest evidence base on recovery methods is thinner than the market around it. It holds a meta-analytical review of how acute sleep loss affects physical performance, a systematic review of sleep interventions aimed at athletes, and an evidence-based framework for choosing between post-exercise recovery techniques.

The signal is that sleep is the recovery intervention with the strongest evidence behind it, and that most of the rest is small, situational, or poorly supported. This is also why Leo reads your readiness from how you say you feel rather than from a wearable score: the papers here measure the consequences of sleep loss on performance, not the accuracy of consumer devices at detecting it. If you tell him you slept badly, that is data. He would rather have that than a number from a ring.