Wearable integration guide — which signals ARES uses

Wearable Signal Fusion Cockpit

# Wearable integration guide — which signals ARES uses

Most modern wearables promise a single glance at your wrist can tell you how your body is doing: “Sleep: 83”, “Readiness: 72”, “Load: 91”. These scores feel authoritative, but in the ARES/Bio.OS view they are proxies, not judges.

Wearables are first and foremost signal streams. Only when you add a baseline window, context and signal fusion do they become decision‑ready inputs for humans or algorithms.

This guide lays out a sober integration model without device hype:

• which devices and data streams actually matter (sleep, HRV, resting heart rate, load),
• how ARES thinks about baselines and direction of change,
• how to map evidence boundaries for sleep and HRV,
• how to prioritize and fuse signals,
• how to handle conflicting inputs,
• and how to implement a minimum viable integration that is robust rather than flashy.

Scope: Educational context only. It supports general learning, simulation and reflection. For personal health questions, consult qualified professionals.

---

1. Wearable scores are proxies, not authorities

Consumer platforms compress large amounts of raw data into a few daily scores: sleep score, readiness score, stress score, training status. Those can be helpful shortcuts, but for ARES the real value lies in the underlying raw streams.

1.1 Why raw signals matter more than scores

A single sleep score can emerge from very different nights:

• shorter total sleep but more deep stages,
• longer time in bed but heavily fragmented,
• normal duration but delayed onset.

For decision support and simulation, the mechanism behind the number matters more than the aggregate.

Key examples:

• Sleep architecture and continuity

Professional sleep labs reference the AASM Manual for defining sleep stages, arousals and respiratory events.¹ (https://aasm.org/clinical-resources/scoring-manual/) That framework emphasizes continuity and transition patterns, not just minutes asleep.

• HRV as a context‑sensitive signal

The classic Task Force paper on heart rate variability (HRV) shows how strongly HRV depends on posture, breathing, recording length and measurement method.² (https://pubmed.ncbi.nlm.nih.gov/8598068/) A single daily value without context can be misleading.

ARES therefore treats platform scores as summary hints, while the engine works mainly with:

• total sleep time,
• sleep timing and fragmentation,
• nightly heart rate and resting heart rate,
• nightly HRV (where available),
• training load and movement,
• subjective check‑ins.

1.2 Consumer wearables in the evidence landscape

In the last decade, consumer wearables have improved substantially. Still, they remain simplified measurement systems compared with full lab setups.

A review by De Zambotti et al. on consumer sleep technologies highlights that wearables are useful for trends and behavioral reflection, while stage‑level details are less reliable and vary by device and population.³ (https://pubmed.ncbi.nlm.nih.gov/30736895/)

For ARES this translates into a conservative stance:

• Sleep duration, timing and gross fragmentation are usually good enough for trend use.
• Detailed stage percentages (deep vs. REM vs. light) are treated as soft signals.
• Optical HRV can be informative if compared across many nights, not as an absolute one‑off number.⁴ (https://pubmed.ncbi.nlm.nih.gov/32897239/)

---

2. Integration model: baseline, direction, interpretation, action

ARES uses a simple but robust four‑step integration model for wearable signals:

1. Baseline window: What is the personal reference range? 2. Direction of change: Is the current value clearly up or down relative to that window? 3. Interpretation with context: Are there external factors (caffeine, late meals, alcohol, travel, intense training) that plausibly shift the signal?⁵ (https://pubmed.ncbi.nlm.nih.gov/24235903/)⁶ (https://pubmed.ncbi.nlm.nih.gov/23550728/) 4. Action layer: How are routines, planning or simulations adapted based on the fused signals?

2.1 Baseline windows: why 14–30 days are a sweet spot

For sleep and HRV, ARES typically works with 14–30‑day windows. This length:

• smooths out outliers from technical glitches or unusual days,
• still reacts quickly enough to meaningful changes (e.g., new work schedule, new training block).

Sleep example:

• Baseline metric: average total sleep of the last 21 nights.
• Baseline band: ±10–15 % around this average.
• Signal: if last night is clearly below the band, this is tagged as a low sleep input; clearly above → high sleep input.

Nightly HRV example:

• Baseline metric: 21‑day median of nightly RMSSD (or similar).
• Baseline band: one standard deviation or an interquartile band (25th–75th percentile).
• Signal: values outside that band trigger stronger checks against context and other streams.

2.2 Direction: trends, not daily mood swings

Single‑night outliers are normal. ARES focuses on moving windows and trajectory:

• 3‑day windows capture short‑term fluctuations (e.g., acute workload, travel).
• 7‑day windows show week‑level balance (e.g., accumulated sleep loss).
• 21‑day windows reveal patterns connected with lifestyle or training cycles.

Direction of change is then defined relative to these windows rather than absolute thresholds.

2.3 Interpretation: the context filter

Several context factors show up consistently in the literature as major modifiers:

• Late caffeine: Drake et al. demonstrate measurable shifts in sleep when caffeine is taken even 6 hours before bed.⁵ (https://pubmed.ncbi.nlm.nih.gov/24235903/)
• Evening alcohol: Ebrahim et al. summarize how alcohol alters continuity and architecture of sleep.⁶ (https://pubmed.ncbi.nlm.nih.gov/23550728/)
• Travel, jet lag, night shifts: phase shifts alter timing and heart rate patterns.
• High training load days: strong muscular and metabolic strain shows up in heart rate and HRV during the night.

In ARES, every numerical signal is therefore combined with a context mask:

• Nights with marked caffeine or alcohol input are down‑weighted for baseline construction.
• Workdays with normal routines are fully weighted.
• Travel days or shift work are tagged as special context, useful for learning but not ideal as reference.

2.4 Action: from numbers to practical decisions

In the ARES philosophy, wearables feed a decision cockpit, not an autopilot. Actions arise from pattern recognition across signals:

• Several nights with short sleep + elevated resting heart rate + falling HRV + high load

→ suggests a heavily taxed system; users or coaches may reflect on load distribution, sleep opportunity, and daily structure.

• High load + stable HRV + stable sleep duration and timing

→ suggests that the present periodization is currently well tolerated.

ARES does not give personal health recommendations. It provides structure so that individuals and teams can:

• run thought experiments and simulations,
• refine routines and planning,
• formulate better questions for qualified professionals.

---

3. Evidence map: sleep and HRV boundaries

An evidence map ties together:

• professional reference standards,
• how consumer wearables approximate them,
• and how ARES uses the difference productively.

3.1 Sleep: lab standards vs. wearables

The AASM Manual defines reference rules for stages (N1, N2, N3, REM) and related events.¹ (https://aasm.org/clinical-resources/scoring-manual/) Polysomnography in the lab combines EEG, EOG, EMG, airflow and more. Wearables, by contrast, rely mainly on accelerometry and optical heart rate, sometimes plus skin temperature.

De Zambotti et al. and related work highlight key points:³ (https://pubmed.ncbi.nlm.nih.gov/30736895/)

• Many devices reasonably estimate sleep duration and bedtime/waketime.
• Stage classification (light vs. deep vs. REM) is notably less accurate and highly device‑dependent.

For ARES this means:

• Stage percentages rarely drive any critical branch in the logic.
• Duration, timing and fragmentation are considered primary.
• Stage information can be displayed for curiosity and reflection, but with lower weight.

3.2 HRV: standards and measurement context

The Task Force standard on HRV remains a cornerstone reference.² (https://pubmed.ncbi.nlm.nih.gov/8598068/) It emphasizes:

• HRV is strongly modulated by recording length, respiratory pattern, posture and time of day.
• Time‑domain measures like RMSSD and frequency‑domain measures capture different aspects.
• Consistent measurement protocols are critical for reliable interpretation.

Nelson et al. review the validity of wearable HR and HRV measurements.⁴ (https://pubmed.ncbi.nlm.nih.gov/32897239/) They find that:

• In rest and sleep, many wearables capture HR with good accuracy.
• HRV accuracy is more variable, but adequate for trends in several devices.
• Movement and sensor artifacts are major challenges.

ARES responds by:

• Focusing on night‑time HRV in supine rest, where artifacts are reduced.
• Using extended windows and distribution‑based summaries rather than relying on a single value.
• Linking HRV trends with load, sleep and subjective feel before drawing any practical conclusions.

---

4. Signal priority: what matters most?

Not all signals are equally robust or equally relevant. The table below shows a typical priority matrix that ARES users adopt in their pipelines.

| Signal type | Source / sensor | Typical reliability (trend) | Typical reliability (single night) | Priority in ARES logic | Main practical use | |---------------------------------|-------------------------------|-----------------------------|-------------------------------------|------------------------|------------------------------------------------------------| | Total sleep time | Wearable (actigraphy + HR) | High | Medium | High | Sleep behavior, catch‑up patterns, rhythm analysis | | Bedtime & wake‑time | Wearable + journal | High | Medium | High | Social jet lag patterns, phase shifts | | Sleep interruptions | Wearable (movement + HR) | Medium | Medium | Medium | Fragmentation, environmental disturbance | | Average nightly HR | Wearable (PPG) | High | Medium | High | Overall strain, relation to daily workload | | Resting heart rate (RHR) | Wearable (PPG) | High | Medium | High | Baseline shifts, response to training and life stressors | | Nightly HRV (e.g., RMSSD) | Wearable (PPG) | Medium–high | Medium–low | Medium–high | Trend analysis of autonomic balance | | Training load index (e.g., TL) | Wearable + training platform | High | High | High | Periodization, block planning, workload progression | | Subjective daily state | Manual check‑ins | High (if consistent) | – | Medium | Interpreting numbers, discovering person