A plan starts with your circumstances
A goal alone is insufficient for a training week. NuroStride combines your goal, completed training, experience, available time and feedback. The skill instructs your assistant to ask about missing information and explain changes. Each session has a purpose, an intensity anchor and a shorter alternative.
This is a structured working method for your AI assistant. The assistant generates recommendations and can make mistakes. The profile establishes context; it promises neither a particular race time nor medical safety.
What the calculator actually computes
- Training volume and session RPE: duration × your reported exertion. Missing days or RPE values are marked unknown. A rest day counts only when explicitly documented as complete.
- Optional Banister TRIMP: requires your documented resting/maximum heart rates and model coefficients. Heart-rate segments can be integrated; a mean-heart-rate calculation is labelled an approximation.
- CTL, ATL and TSB: smoothed load curves with explicit initial values, an initialization explanation and configurable time constants. Missing load interrupts the curve. TSB describes the state before the day being considered.
- 7-/28-day load ratio: a descriptive quotient requiring 28 complete days and a positive denominator. The windows overlap. It establishes neither an injury risk nor a safe training zone.
- Target splits and fueling arithmetic: splits from your chosen time and distance; carbohydrate totals and portions from your chosen duration, rate and serving size. The calculator does not choose those targets for you.
The deterministic calculations were checked using complete and incomplete histories, unknown values, invalid inputs and synthetic examples. This checks calculation behaviour and limits, not the effectiveness of an AI training plan.
The method goes beyond numbers
The references cover weekly reviews, periodization, long runs, tapering, recovery, strength work, race scenarios and practiced fueling. A desired goal is distinguished from an option supported by comparable recent performances. Missed hard sessions are not accumulated and made up together. Revised goals and changes are discussed with you.
EF, heart-rate drift and GAP are explained as optional analytical methods. They require comparable segments, data quality and a stated definition. The kit does not automatically calculate these metrics; FIT/GPX import, phase detection and performance forecasts are also not implemented.
Who builds NuroStride
Arne Kellmann develops NuroStride from his background as an ambitious amateur athlete and intensive independent study of training science. The practical question behind it is how research, completed training and a runner’s daily circumstances can inform reasoned decisions.
The result is a method whose sources, assumptions and calculations you can inspect. Personal experience provides a perspective; it does not establish the product’s effectiveness. Technical Premium Setup helps with installing and using the files. More about Arne and his projects ↗
Research, model and coaching heuristic
These levels are kept separate: a study describes its sampled groups, a load model depends on definitions and inputs, and a coaching heuristic must be checked against your feedback. No single metric grants clearance to train. A general percentage rule does not guarantee injury prevention.
- Foster et al. (2001) — Session RPE as a method for describing training load.
- Morton et al. (1990) — A modeled relationship between training and performance, not a personal health measurement.
- Impellizzeri et al. (2021) — Methodological limits of ACWR and injury associations.
- Cohort of 5,205 runners (2025) — Observational evidence about individual distance spikes; no individual safe threshold.
- Bosquet et al. (2007) — Tapering as a starting point for individual planning.
- Minetti et al. (2002) — Energy costs of gradients under treadmill conditions.
- Thomas, Erdman & Burke (2016) — Nutrition in relation to exercise and individual context.
NuroStride provides no diagnosis, biomarker interpretation, injury forecast or medical clearance. A plan does not replace professional help for acute symptoms. MCP is only a possible data source you configure; the kit includes no ready-made Garmin, Strava or watch connection.