Reflection · Descriptive Analytics
Patterns & Discoveries
Four patterns surfaced from a decade of training data — things you might not have noticed manually. Reflective, narrative-led.
Patterns
4 / 4
Sample Size
1,925
Top Discoveries
Auto-surfacedChangepoint Detection
PELT — Pruned Exact Linear Time
4 structural shifts detected in your training between 2015 and 2025.
The algorithm scans your weekly volume and flags moments where the underlying pattern fundamentally changed — not just normal variation.
2016-W44
Volume rose 197% (26.8→79.6 km/wk)
Confidence: 72%
2023-W03
Volume rose 157% (75.1→192.8 km/wk)
Confidence: 99%
2025-W00
Volume rose 40% (171.6→239.4 km/wk)
Confidence: 74%
2025-W39
Volume dropped 44% (259.6→146 km/wk)
Confidence: 72%
Training Era Clustering
Hierarchical · Ward Linkage · Year DNA
Your training history clusters into 3 distinct eras.
Each era is a group of years with similar volume, sport mix, HR profile and time-of-day patterns.
Foundation Era
2015 – 2015
Cycling-dominant, avg 65 km/yr
Build & Compete
2016 – 2022
Mixed sport, avg 3,498 km/yr
Recent Evolution
2023 – 2025
Cycling-dominant, avg 9,383 km/yr
STL Seasonal Decomposition
Seasonal-Trend via Loess · 12-mo period
Underlying trend rose 825% across the period — improvement is genuine, not seasonal noise.
Seasonal swings span ±125 km/month — that's the predictable summer-vs-winter rhythm.
Anomaly Detection
Multi-feature Z-Score (Isolation Forest proxy)
10 standout sessions sit far outside your normal range.
These are statistical outliers across distance, duration and pace — your standout days, including the proof that the competitive year was genuinely exceptional, not just remembered as such.
★ Standout Performances (10)
Relay for Life 2024
2024-10-12 · Walk · 45.5km · 590min
Seven Solo training camp Day 2
2025-05-01 · Ride · 148.5km · 457min
Morning Ride
2018-10-27 · Ride · 180.9km · 375min
Gravel World Champs - Maastricht 2025
2025-10-12 · Ride · 179.9km · 394min
Afternoon Ride
2018-12-02 · Ride · 179.2km · 351min
Jipped! Flat 6km from home.
2018-09-29 · Ride · 173.9km · 363min
Night Ride
2025-09-26 · Ride · 171.1km · 365min
Seven
2023-05-12 · Ride · 124.5km · 386min
Morning Gravel Ride
2025-09-19 · Ride · 144.5km · 367min
Morning Gravel Ride
2025-09-12 · Ride · 138.9km · 357min
HR Age Benchmarking
Linear Regression vs Tanaka norms
Observed peak 211 bpm exceeds the Tanaka age-predicted ceiling of 181 bpm (Δ +30).
Peak HR is drifting +2.5 bpm/year vs the population average of −0.7 bpm/year.
Four reflective patterns plus an HR age-benchmarking reference · all algorithms run client-side on your filtered activity set (1,925 sessions).