QuikShift
Sign in Start coaching
Start coaching

Every number here has a paper behind it.

We don’t invent metrics. Every one is built on published sport science and listed below with a link to the original work. As the engine grows and new research lands, this page keeps up.

55 published sources Reviewed May 2026

Power & critical power

Read the explainer Critical power, the work W′ pays for above it, and why FTP is a useful number that is not the same thing.

Training load & form

Read the explainer What TSS, CTL, ATL and TSB measure, and how to read the curve without over-reading it.

Intensity distribution

Read the explainer Polarized, pyramidal and threshold distributions, and where the 80/20 pattern came from. Two people can measure the same week and disagree about it; this covers why.

Thresholds & physiology

Read the explainer Two thresholds, a dozen names between them, and how far VO2max can realistically be moved. Holding your numbers late in a race is a separate quality again.

Recovery & readiness

Read the explainer One morning HRV reading says less than the trend it sits in. What the number can and cannot tell you, and where sleep and self-report come in.

Running & gradient

Read the explainer A slope changes what a metre of running costs. What grade-adjusted pace corrects for, and why economy moves your pace independently of VO2max.

Fueling & hydration

Read the explainer Consensus ranges for carbohydrate, protein, fluid and caffeine in endurance training. Also where those ranges stop being general and start being about you.

Heat & altitude

Read the explainer A heat block changes more than it feels like it should. Why hot-weather targets need derating, and how much of the altitude story the evidence actually carries.

Female physiology

Read the explainer The cycle-phase research is thinner than most summaries admit. What it does show, and the documented sex differences in performance and fuel use.

Health & overtraining

Read the explainer Overreaching through to overtraining, as the consensus defines it. It is diagnosed by exclusion, which is why no single marker settles the question.

Strength standards

Read the explainer The five lifts and what counts as a qualifying rep. How your bodyweight is handled between published classes, and where the app rounds, clamps or declines to answer.

Strength for endurance

Read the explainer Heavy, low-rep work makes an endurance athlete more economical without making them heavier. What the injury evidence actually tested, and where the strength tab's tier badge comes from.

Wind & aerodynamics

Read the explainer On a bike, the air is the biggest thing you push against. What CdA and yaw angle measure, and how far to trust a drag number fitted from a ride rather than a tunnel.

The AI is held to the same standard

Those papers aren’t decoration. They gate what the AI is allowed to say, in three steps.

  1. The engine computes

    Deterministic code, not a language model, calculates every metric from the athlete’s data using the cited methods on this page. It’s the only source of numbers in the system.

  2. The AI narrates those results

    The model gets the computed findings and the relevant citations, and explains them in plain language. It’s never asked to estimate, extrapolate, or fill a gap.

  3. Every claim is checked afterwards

    Each number in the write-up is compared against the computed findings. If one figure can’t be matched, the whole write-up is discarded and the engine’s own plain-language read is shown instead. A hallucinating model degrades to accurate text rather than a fabricated number.

That’s the whole mechanism. It can make the science readable, and it can’t add anything that isn’t already in the numbers.

Spot a method we should add?

The library gets curated as the field moves. If there’s research you’d trust your athletes’ training to, we want it in here.

Tell us about it