User manual · 10/12
Analytics and AI reports
Analytics turn an athlete's history into guidance: are they ready or fatigued? What happens if I increase the load? Who should I look at today? These are estimates for information only, not medical advice.
Load numbers in a nutshell
| Number | What it means |
|---|---|
| Fitness (CTL) | underlying condition: average load over roughly the last 6 weeks. Rises and falls slowly. |
| Fatigue (ATL) | recent tiredness: average over the last week. Reacts quickly. |
| Form (TSB) | fitness minus fatigue. Positive: fresh. Strongly negative: accumulated fatigue. |
| ACWR | last week's load divided by the average of the last 4. Between 0.8 and 1.3 the load is building gradually; above 1.5 it is rising too fast. |
Each session's load is computed from heart rate; without it, from pace; without that either, from duration. For reliable numbers set max HR and resting HR in Athlete settings.
Readiness
Athlete → Readiness (the page is titled "Load, form & records"). At the top is the verdict: Ready (green), Caution (yellow) or Fatigued (red). Pick the Method with three buttons; those without data are greyed out.
| Method | How it decides |
|---|---|
| HRV | compares today's heart-rate variability with the personal average of the previous 7 days. It needs 7 measurements in 7 days: until then you see "Baseline: n/7 days". |
| Load (cardio + strength) / Load (cardio only) | uses Form: −10 and above Ready, between −10 and −20 Caution, below −20 Fatigued. |
Below you will find the fitness, fatigue and form chart, VO2max estimated from runs, and records: best time over 1 km, 1 mile, 5 km and 10 km, and the longest workout by time and by distance.
Digital Twin: simulating a plan
The Digital Twin projects the athlete's condition forward and answers two questions: "where will they be in a few weeks if they carry on like this?" and "is the change I have in mind risky?". It uses the same models as the Readiness page; it is not artificial intelligence and it does not predict injuries.
- Open the athlete → Digital Twin.
- At the top check Data reliability (High, Medium or Low) and the reason. High reliability needs at least 8 weeks of history, 2 sessions a week and a recent last session. Check calibration shows how far off the twin usually is for this athlete.
- Choose the Horizon, from 2 to 12 weeks.
- In the Levers tab move volume, intensity, sessions per week, deload week or initial break. Or, in the Plan tab, pick one of the athlete's plans and how many times a week to do it.
- In the charts compare the coloured line (Scenario) with the grey one (If nothing changes). What matters is the gap between them.
- Go through the Risk indicators week by week: green no signal, yellow keep an eye on it, red attention.
The Proposal tab builds a week-by-week load plan by itself. It suggests a Phase (Build, Maintain, Deload, Return) with an explanation and sticks to cautious limits: at most +10% a week, a deload every 4 weeks (every 3 from age 50), ACWR no higher than 1.3. You can correct the loads in the table by hand.
Time machine: restarts the twin from 30, 60 or 90 days ago and compares its forecast with what actually happened. It helps you judge how far to trust it.
Think Center: every athlete at a glance
Menu Think Center. It works without AI too: it shows how many athletes need attention, are inactive or on track, and one row per athlete with a coloured dot.
| Dot | When |
|---|---|
| Red | inactive for more than 7 days, or readiness "Fatigued", or ACWR above 1.5 |
| Amber | readiness "Caution", or ACWR outside 0.8–1.3, or last session more than 3 days ago |
| Green | everything else |
Filter by colour and sort by last activity, readiness, 7-day load or name. Click an athlete to open the report panel.
Setting up AI for reports (Ollama)
Written reports are produced by a language model you host, with Ollama, on the same PC or on a computer on your network. cronowiz Studio ships no model and uses no external service. It is optional: everything else works without it.
- Install Ollama from ollama.com and download a model from a terminal, for example
ollama pull llama3.1. Bigger models write better but need more memory. - In cronowiz Studio open Settings → Virtual PT (LLM · Ollama) and turn on Enable.
- In Base URL keep
http://localhost:11434if Ollama runs on the same PC; otherwise enter the address of the computer hosting it, for examplehttp://192.168.1.50:11434. - Press Test connection: you should see "Connected". The API key is only needed if Ollama sits behind a proxy that requires one.
- Choose the Model from the list and press Save: without Save the settings are not kept.
Generating a report
- In Think Center click an athlete.
- Choose the Report type: Period summary, Readiness & load, Performance progression, Next session, Load progression (explains the Digital Twin proposal) or Full report, which combines them all.
- Choose the Window in days (28 by default) and press Generate. The text appears as it is written, in the program's language.
- At least 2 sessions are needed in the chosen window; otherwise widen it.
- Every report is saved. If new data arrives in the meantime a notice appears: press Regenerate; the previous version goes to the Archive.
- Copy puts the text on the clipboard; Export saves it as Markdown or PDF.