Wearable integration guide — which signals ARES uses

Wearable Signal Fusion Cockpit

# Wearable integration guide — which signals ARES uses

Wearables promise clarity: a single score for sleep, a readiness bar for the day, a colored ring for how “prepared” you are. In reality these numbers are proxies, not authorities. They compress many raw signals into a convenient index – but they do not know your context.

This guide explains how ARES/Bio.OS treats wearable data in an educational and simulation setting: which signals matter most, how they are prioritized, and how conflicting streams can be reconciled. The focus is on sleep, HRV (heart rate variability), RHR (resting heart rate) and load.

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

---

1. Hook: wearable scores are proxies, not authorities

Most platforms ship a bundle of daily scores: sleep quality, readiness, strain, recovery index and more. In the ARES model these are treated as derived products, not as the core data.

Key implications:

1. Scores are device-specific Each vendor blends raw data differently: weighting sleep duration vs. deep sleep, heart rate vs. HRV, motion vs. temperature. A score of 85/100 on device A does not equal 85/100 on device B.

2. Scores hide measurement noise Aggregation smooths outliers but also hides why a change occurred. A “bad readiness score” might reflect shorter sleep, high caffeine intake, recent travel, or plain sensor artifacts.

3. Scores react to context the device does not see Research shows, for example, that caffeine can shift sleep duration and structure Drake et al., 2013 (https://pubmed.ncbi.nlm.nih.gov/24235903/), and alcohol can reshape sleep continuity and architecture Ebrahim et al., 2013 (https://pubmed.ncbi.nlm.nih.gov/23550728/). Only you know these influences; the device does not.

4. ARES puts raw signals first ARES prioritizes directly measured quantities (sleep duration, wake episodes, night-time HRV, night RHR) and treats vendor scores mainly as meta-hints.

In short: wearables provide valuable streams, but signal fusion and context are what make them decision-ready in a learning and simulation environment.

---

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

ARES uses a four-layer integration model:

1. Baseline window — understand usual patterns. 2. Direction of change — focus on deviations from your baseline. 3. Interpretation — add context layers around the numbers. 4. Action — design learning experiments and adjustments.

2.1 Baseline window

A baseline is a personal reference corridor. Instead of asking “Is 6 hours of sleep bad?” the model asks “How does 6 hours compare to the last 30 days for this person?”

For ARES the baseline window typically includes:

• Sleep: mean sleep duration, variance, typical bedtime and wake time, frequency of wake episodes. For reference, the American Academy of Sleep Medicine provides standardized scoring rules for lab-based sleep staging in its Scoring Manual (https://aasm.org/clinical-resources/scoring-manual/), which serve as a conceptual boundary, even though consumer devices are less precise.
• HRV: typical nightly RMSSD/SDNN range under calm conditions, aligned with the measurement conventions in the Task Force HRV standards paper 1996 (https://pubmed.ncbi.nlm.nih.gov/8598068/).
• RHR: average night-time heart rate in restful periods, with spread over weeks.
• Load: typical distribution of activity and training throughout the week.

2.2 Direction of change

ARES pays more attention to movement relative to baseline than to absolute numbers:

• Is sleep shorter or longer than usual?
• Is nightly HRV clearly lower or higher than the personal corridor?
• Is RHR unusually elevated compared with previous nights?
• Is load much higher or lower than the usual pattern?

Once these differences are flagged, the question becomes: What changed around those days?

2.3 Interpretation layer

Raw data points are then wrapped in context layers:

• Behavioral context: training timing and intensity, late meals, caffeine or alcohol intake.
• Environmental context: jet lag, travel, time zones, room temperature, light exposure.
• Subjective context: perceived fatigue, mood, muscle soreness, focus.

Evidence on consumer sleep technologies emphasizes that trends and changes are typically more robust than single-night values De Zambotti et al., 2019 (https://pubmed.ncbi.nlm.nih.gov/30736895/). ARES leans heavily into that insight.

2.4 Action in a learning & simulation setting

In Bio.OS, “action” mainly means experiments and adjustments, such as:

• Designing a more consistent sleep window and observing signal shifts.
• Periodizing training load and deliberately planning recovery days.
• Reflecting on timing of caffeine and alcohol based on research like Drake et al. (https://pubmed.ncbi.nlm.nih.gov/24235903/) and Ebrahim et al. (https://pubmed.ncbi.nlm.nih.gov/23550728/).
• Improving sensor placement, strap tightness and device wear habits.

ARES provides maps, dashboards and decision trees for this, aiming at general learning and reflection, not individualized instructions.

---

3. Evidence map: boundaries for sleep and HRV

3.1 Sleep as a multi-layered signal

Sleep trackers attempt to estimate stages: light, deep, REM, wake. The gold-standard reference for sleep staging from lab polysomnography is documented in the AASM Manual for the Scoring of Sleep and Associated Events (https://aasm.org/clinical-resources/scoring-manual/). Consumer wearables instead rely on optical sensors (PPG), accelerometers and proprietary algorithms.

Systematic reviews, such as the consumer sleep technology review by De Zambotti et al., 2019 (https://pubmed.ncbi.nlm.nih.gov/30736895/), highlight key points:

• Sleep duration is often captured reasonably well, especially for longer nights and stable schedules.
• Sleep stages (deep vs. REM vs. light) are notably less accurate than lab recordings.
• Wearables are better suited for trend tracking than for precise night-by-night classification.

ARES derives three operating principles:

1. Duration beats fine-grained architecture for most users Outside of research labs, a robust estimate of how much you slept is more actionable than precise minute-level REM percentages.

2. Stability beats single nights Instead of overreacting to a single “bad” night, the model highlights patterns that repeat over multiple days.

3. Large jumps beat small wiggles A drastic drop in sleep duration or recurrently fragmented nights matters more than a 20–30 minute fluctuation.

Sleep architecture as a signal layer

3.2 HRV as a context signal

Heart rate variability (HRV) is used as a marker of autonomic nervous system dynamics. The foundational Task Force HRV standards paper 1996 (https://pubmed.ncbi.nlm.nih.gov/8598068/) laid out measurement protocols and time windows (e.g. 5‑minute resting segments).

Modern wearables typically implement continuous PPG-based HRV estimation. A broad review of wearable HRV validity by Nelson et al., 2020 (https://pubmed.ncbi.nlm.nih.gov/32897239/) notes that:

• Some wearables provide reasonably valid HRV trends at rest.
• Motion, poor contact and light artifacts can distort readings, especially during exercise.

ARES therefore focuses on:

• Night-time HRV windows during still, movement-free segments.
• Relative deviation from the person’s own baseline, not comparison with population averages.
• Combined interpretation with RHR and sleep metrics.

3.3 How sleep, HRV, RHR and load interact

In ARES each signal is a layer. Layers stacked together form a pattern. For example:

• Shortened sleep over several nights
• Markedly lower nightly HRV vs. baseline
• Elevated RHR compared with previous weeks
• Recent high training load

→ In a learning scenario this may be interpreted as “overall stress is high; experiment with load and recovery windows.” No single data point is definitive, but convergence across layers is informative.

---

4. Signal priority: which streams win when?

ARES uses a priority matrix to decide how strongly each signal contributes to the current assessment.

4.1 Priority table

| Rank | Signal layer | Typical source | Context where it dominates | Typical ARES weighting | |------|-----------------------------|---------------------------------|------------------------------------------------------------|--------------------------| | 1 | Total sleep duration | Ring, watch, sleep tracker | Learning goals around restoration, focus, performance | Very high | | 2 | Sleep continuity (wake time)| Wearable + subjective logging | Fragmentation, jet lag simulations, shift work patterns | High | | 3 | Night-time HRV (RMSSD etc.)| Wearable with reliable PPG | Stress/recovery trends, training periodization | Medium–high | | 4 | Night-time RHR | Most wearables | Load estimation, cross-check with HRV | Medium | | 5 | Daily activity / steps | Watch, tracker, smartphone | Baseline movement, sitting time, NEAT-focused learning | Medium | | 6 | Training load (TRIMP, TSS) | Sports watches, training apps | Periodization blocks, endurance/performance experiments | Medium–high (if relevant)| | 7 | Vendor “readiness” scores | Proprietary vendor apps | Rough guidance where raw data are ambiguous | Low–medium | | 8 | Subjective ratings | Journals, in-app surveys | Context anchor, comparison to wearable trends | Medium |

Weighting refers to common educational use cases and can be adjusted.

The main ideas:

• Raw signals outrank composite indices.
• Subjective context is part of the fusion, not a competing truth.

---

5. Stream-by-stream interpretation

This section walks through the key streams and how ARES typically reads them.

5.1 Sleep streams

Sources: Oura, Whoop, Garmin, Apple Watch, Fitbit, and similar devices. Key raw signals: sleep onset and offset, estimated sleep duration, wake episodes, sleep stages.

ARES references formal sources like the AASM Scoring Manual (https://aasm.org/clinical-resources/scoring-manual/) conceptually, and practical consumer evaluations such as De Zambotti et al., 2019 (https://pubmed.ncbi.nlm.nih.gov/30736895/), while staying device-agnostic.

In a typical Bio.OS setup, questions include:

• Is total sleep time in the target corridor for this person (which may differ across training phases)?
• Is sleep efficiency (time in bed vs. time asleep) reasonably stable across weeks?
• Are there repeatable patterns, such as shorter sleep after late training or heavy evening meals?

Example learning prompts:

• “Run a 2‑week experiment with 30–60 extra minutes in bed and observe signal changes.”
• “Keep bedtime within a 30‑minute window for one month and compare variability before/after.”

5.2 HRV streams

Sources: the same wearables, optionally chest straps for high-precision snapshots. Signals: RMSSD, SDNN, frequency-domain metrics, often summarized into a nightly HRV score.

Evidence from Nelson et al., 2020 (https://pubmed.ncbi.nlm.nih.gov/32897239/) suggests that wrist-based HRV can capture resti