Sleep targets

Updated 2 Nov 2026

Objective: achieve consistently excellent sleep, then expand the measurements and improve its demonstrated contribution to alertness, cognitive performance and training recovery.

This is version 1 of the target specification, clarified through 2 November. The numerical score milestones below are proposed project benchmarks. They are not clinical cutoffs, adopted treatment targets or evidence that a particular lifespan has been gained. The 21 dated entries include 20 eligible Oura whole-night sleep scores and 21 available COROS sleep scores; a full personal baseline and a functional test protocol are still to be established.

Two score milestones

Oura defines 85–100 as “Optimal.” The 14/28-day windows, coverage rules and 90-point stretch are deliberate choices for this project, not validated biological thresholds. A single 95-point night does not satisfy either sustained milestone. Oura definitions.

The latest eligible whole-night Oura score is 83 on 2026-11-02. The latest available COROS score is 88 on 2026-11-02. Each remains a separate benchmark. We do not switch providers to select the higher score or average their grades.

Current practical question, 2 November: keep the poor experienced night beside both provider grades. Oura 83 / COROS 88 estimate 8h11m / 8h37m asleep, with 30 / 7 minutes awake. Both end at 09:44 PST, starting 01:03 / 01:00. Oura timing falls to 50, REM to 74 and total sleep to 87, but efficiency improves to 99/100 (94% estimated sleep efficiency). Both HRV summaries are 44 ms; Oura Readiness is 82. The new curves retain the native sequence. The follow-up reports attempting sleep around 01:00, waking around 09:45 and not sleeping well. Both windows fit the approximate times, but their sleep/wake labels still differ. Oura falls 96 → 83 while COROS rises 87 → 88; neither grade replaces the poor experienced outcome. The user now identifies timing as the main suspected contributor and reports many thoughts before bed. This is a personal attribution; the timing grade does not quantify a causal share. Exact onset, recalled awakenings, morning/later-day alertness and other evening conditions remain unspecified. Fixed milestone criteria remain unchanged.

1 November comparison: does the new single-night high coincide with better experienced sleep? Oura 96 / COROS 87 estimate 9h59m30s / 10h26m asleep. Oura timing is 96 and total sleep 100, but restfulness falls to 83 and awake time rises to 52 minutes. The curve preserves elapsed time across the clock rollback. COROS’s clock labels have an unresolved one-hour offset discrepancy, so endpoint agreement is not asserted. The follow-up reports bed around 23:00–23:30, feeling kind of falling asleep from 22:30 and waking around 09:30, close to Oura’s 09:26:53 ending. Continuous sleep from 22:30 and COROS’s endpoint offsets are not established. The user reports good sleep while distinguishing conditions: the short sleeping surface and restricted turning were uncomfortable; the pillow was better, the blanket comparable, and temperature much better, perfect and stable. Background noise felt helpful. Pre-bed light and late food remain perceived drawbacks, but the user suspects they are not the highest-impact optimizations. That relative ranking remains a working hypothesis. Incoming morning light reportedly woke him earlier; its time and relationship to final waking remain unspecified. A main meal 19:30–20:00, small desserts 20:00/21:30, about 60 minutes walking after food and social contact are now known. Main-meal-to-bed and last-dessert-to-bed intervals were roughly 3–4 and 1.5–2 hours. Morning/later-day alertness, fluid timing and cold awakenings remain unreported; no individual cause or optimal cutoff is established. The fixed milestone assessment is unchanged.

31 October comparison: how should the new Oura high be judged alongside COROS and the diary? Oura 93/COROS 83 estimate 8h57m30s/8h47m asleep, with awake totals 46/27 minutes. The user reports no naps, onset around midnight, last food 19:00 and waking 09:15–09:30. COROS’s09:16 ending fits that range more closely than Oura’s09:42. The later follow-up reports good overall sleep and a good morning, with time lying in bed thinking. Later-day functioning remains unknown. A recurring cold-feet pattern around 07:30, plus overheating or removal of heavier starting bedding, is now a separate comfort comparison. An earlier schedule is proposed; sufficient sleep opportunity and the overheating constraint stay visible. The main curve is unchanged; the two-minute evening detection is disputed. The five-hour meal-to-reported-onset gap is an observation, not a demonstrated optimum. Fixed milestones remain unchanged.

30 October comparison: how did the later recorded start and shorter sleep feel? Oura 85/COROS 89 estimate 8h10m/8h20m asleep, with close windows around 00:45–09:25/00:49–09:22. Oura timing falls to 61, restfulness 81 and total sleep 87, while REM improves to 96. HRV rises to 69/68 ms. The curves preserve these different changes. The follow-up reports feeling good, last meal 18:00, a 00:30 sleep attempt and waking around 09:15–09:30. That is about 6h30m from the last meal to the attempt; both device endings fit the waking interval. Exact onset, remembered awakenings, fluid timing and later-day function remain unspecified. The five-minute Oura latency starts from its later main episode, not from 00:30. The good experience and lower grade remain separate outcomes; no meal effect is isolated. Fixed milestone criteria remain unchanged.

29 October comparison: did the earlier recorded start reflect the intended routine, and how did the night feel? The user confirms sleep during 08:27–09:19. Oura’s recheck revises 88 → 90, asleep 8h05m30s → 8h57m30s, and end 08:27 → 09:37. COROS remains 86 / 9h09m, ending 09:19. Current timing is 73, REM 80 and efficiency 97/100 with 92% estimated efficiency; the earlier 100-point efficiency statement is superseded. The curves now use the revised sequence. Actual meal/attempt times, final waking and experienced outcomes remain unreported. No battery cause or meal effect is established; fixed milestones remain unchanged.

28 October comparison: do the lower awake estimates coincide with better experienced continuity and functioning? Oura 91/COROS 94 estimate 9h/9h16m asleep and 30/10 minutes awake. Main sleep starts later, around 00:44–00:47. Oura efficiency contributor reaches 100 (estimated efficiency 95%), while timing falls to 48; HRV is 55/54 ms. The new curves retain those changes. The follow-up reports a roughly 00:30 attempt delayed by meeting people all day, food/lights before bed and pulse less settled than wanted, then an okay night, good rest and morning, and possible few unremarkable awakenings. Exact onset/exposures, current body pains and later-day functioning remain unspecified. The 13-minute API latency concerns a main episode beginning at 00:44, not the whole reported attempt. No isolated cause or new fixed cutoff is established. The first fixed score-window result and the 28-day benchmark remain unchanged.

27 October comparison: does the earlier main episode coincide with better experienced sleep and functioning? Oura 91/COROS 90 report 9h39m30s/9h50m asleep, with starts around 23:50–23:56 and different end times. Oura timing 64 and restfulness 91 rise; HRV is 64/63 ms. The curves retain the actual labels and pulse samples. The follow-up reports a 23:10 sleep attempt with initial difficulty falling asleep, feeling good during the night and morning, and almost all body pains gone. Exact onset and sustained later-day functioning remain unknown. The separate Oura pre-main detection is not explicitly confirmed; its 7m30s main-episode latency does not cover the full interval from the reported attempt. No prior-night exposures are assigned to today. The completed first-window 82.31 and the unchanged 28-day benchmark remain separate.

26 October comparison: keep the reported bad sleep and symptoms alongside the improved measurements. The record now has Oura 85/COROS 94, with 8h33m/8h55m main sleep, HRV 53/51 ms and Oura temperature deviation +0.01 °C. The follow-up reports bad sleep, waking sleepy, ongoing body pains and feeling unhappy overall. It also reports no naps, disputing COROS’s detection; neither extra window is confirmed as actual sleep. Sleepiness duration and later-day functioning remain unknown. The first fixed 14-date window has 13 eligible Oura scores and mean 82.31: coverage is adequate, but the 85-point milestone is not achieved. October 25 alcohol and pre-bed emotions are not assumed repeated. Improved grades do not establish experienced recovery; no threshold or inclusion rule is changed.

Earlier 26 October retrieval: clarify the circumstances and experienced outcome of the October 25 changes before choosing another optimization. The newest available record returns Oura 69, 7h19m asleep, HRV 37 ms and temperature deviation +1.34 °C, after October 24’s 86. The follow-up explicitly reports alcohol, sickness, late bedtime and pre-bed emotions for October 25; amounts, symptoms and individual effects remain unspecified. There are thirteen dated entries and twelve eligible Oura scores. October 26 returned no Oura sleep yet. The later COROS retrieval adds 94 on October 24 and 65 on October 25, with official HRV 55 → 40 ms. COROS’s October 26 response has only a short detected nap and an invalid −1 grade, with no main sleep episode; it is not counted as a valid nightly score. No empty October 26 diary, imputed score or completed sustained milestone is created. Numeric thresholds and inclusion rules remain unchanged.

23 October comparison: compare an earlier opportunity and experienced functioning while retaining externally triggered waking as context. The latest record reports food around 20:00, gym 21:30–23:00, a midnight shower, bed around 01:00 and a cleaner awakening around 09:30–09:40. Oura returned 83 and 7h44m30s asleep; COROS 88 and 8h13m. Their windows end around 09:28. Oura timing 52 remains its lowest grade. Today restores eligible whole-night coverage; October 22’s partial 73 remains excluded. Eleven dates contain ten eligible Oura scores, without completing either sustained milestone. The follow-up confirms staying awake after the cleaner awakening, feeling sleepy initially and still a little sleepy at reply. Elapsed time, resolution and performance effects remain unspecified; no 10–15-minute recovery is inferred. No new exact bedtime or causal routine rule is adopted.

22 October comparison: restore full recording coverage and compare an earlier bedtime with experienced sleep quality. The latest diary describes middling sleep, bed around 00:40 and uncertain waking after 10:00, perhaps 10:15. COROS returned 91 and 9h15m asleep in a 00:44–10:07 episode. Oura stopped at 06:30 after reported low battery; its 73 and timing-100 grade are partial-recording outputs, not full-night results or achievements. Ten dated entries contain nine eligible Oura whole-night scores. The user wants an earlier bedtime without specifying a new exact target; no sustained milestone is awarded.

21 October comparison: can fewer remembered awakenings repeat alongside sufficient sleep and useful daytime function? Today’s record adds fewer—but still several—awakenings, a roughly 00:10 bedtime and 10:15 waking, with initial sleepiness clearing within 10–15 minutes. Oura returned 90 and COROS 91, estimating 9h09m30s and 9h33m asleep. Oura restfulness fell 88 → 85 and timing 66 → 56; the lower grade does not erase the subjective improvement. Last food around 19:00–20:00 was approximately 4h10m–5h10m before bed; other routine changes remain unknown. Nine recorded dates do not complete either sustained milestone. Thresholds and inclusion rules remain unchanged.

20 October comparison: can rising device grades coexist with more comfortable sleep and useful daytime function? Oura returned 92, its highest recorded grade; COROS returned 89. Estimated sleep was 8h55m30s and 9h04m, respectively. Oura timing rose 51 → 66, still its lowest contributor, while restfulness stayed 88. The diary describes okay-ish sleep and many brief awakenings, plus plentiful dream recall. Last food around 18:30 preceded a roughly midnight bedtime; small water sips were unquantified. The record retains that experience and the different device waking estimates. The follow-up reports waking around 09:30 feeling super sleepy, trying unsuccessfully to sleep more for 5–10 minutes, then deciding to get up. No awakening led to a bathroom trip; the brief episodes were estimated under 30 seconds, with dreams suspected as triggers. The second follow-up clarifies normal functioning within 10–15 minutes, with sustained later-day experience unreported. The user finds COROS closer to awake-period timing and Oura probably closer to total awake duration; these impressions remain separate from independent validation. The recent Oura sequence 90, 91, 92 is progress on the score benchmark; eight dates do not meet the assessment window. Continue comparing timing, comfortable sleep and function through the routine history. No sustained milestone is awarded; thresholds and inclusion rules are unchanged.

First window, 13–26 October: eligible score sum 1070 ÷ 13 = 82.3077, displayed 82.31. October 22 remains an explicit battery-related coverage gap. All other eligible dates are included, including difficult nights; no missing score is imputed. Minimum 12-of-14 coverage is met, while the mean 85 criterion is not. The 28-day benchmark remains unassessed.

How a milestone is assessed

Choose the calendar window before assessing it; count each provider wake date once. Show every date, the number of valid scores and missing dates. Include recorded travel, illness and difficult nights in the headline result, with contextual subgroup analyses labeled separately. No imputation of missing scores. Known incomplete recordings, including a battery interruption, are flagged and excluded from whole-night milestone calculations regardless of whether their grades are high or low. Preserve their returned values and their calendar dates as coverage gaps; do not substitute another provider’s score. The October 22 record is the first such flagged case. If an algorithm changes, mark that boundary before interpreting a trend. Retain the component distributions, not just the mean.

Coverage alone does not fix selective missingness. If low-quality or difficult nights are disproportionately absent, the window needs review before a milestone is awarded. Two well-measured windows can test whether a result is repeatable.

Targets beyond a composite score

The duration reference comes from adult consensus and the National Sleep Foundation's age-specific recommendations. Their purpose is population guidance. An individual optimum requires repeated observations and suitable outcomes. AASM/SRS statement; NSF recommendations.

A functional check alongside the score

Start with a validated psychomotor vigilance task rather than treating a generic reaction-time game as equivalent. Keep its duration, thresholds, device and testing conditions consistent. Retain lapses, false starts and reaction-time distribution, plus time since waking and caffeine exposure. Practice and hardware latency can move results independently of sleep.

The future target is a repeatable improvement beyond baseline variability, with no increase in false starts or unacceptable cost in sleep time, training or daily functioning. Its numerical size remains unresolved until a protocol and baseline exist. A rested baseline may already have very few lapses; a suitable response-time distribution or another validated outcome may then offer a more informative challenge. Measurement choices.

A measurable environmental reference: light

Expert consensus for healthy adults aged 18–55 with regular daytime schedules offers quantitative light-exposure references using melanopic equivalent daylight illuminance (melanopic EDI), measured at the eye:

These are exposure recommendations derived from expert review, not a randomized demonstration of my personal optimum. Ordinary photopic lux or an uncalibrated phone reading cannot substitute for melanopic EDI. Spectrum, direction and measurement position matter. The paper discloses lighting-industry relationships among some authors. Brown and colleagues, 2022.

The useful next question is whether the current light pattern conflicts with sleep and waking goals, and whether a feasible change improves a defined outcome. That is more informative than buying equipment solely to fill a table. Light investigation.

How the project expands

1. Establish the measurement: enough repeated nights, units, source and limitations to know what the proposed target means.
2. Pursue a specific improvement: select a practical change and its expected effect before examining the result.
3. Check repetition and function: retain successful and unsuccessful nights, functional observations and confounding context.
4. Expand: add an outcome, a reference measurement or a new hypothesis when it could change a consequential decision.

Expansion can begin before a score milestone is achieved. A clinically relevant issue or a promising measurement question should not wait for 90 points. Additional instrumentation earns its place by answering a question; the number of measured variables is not itself the achievement.

What a result would establish

Three decisions remain separate: was the benchmark achieved, should the intervention be retained, and what broader benefit was demonstrated? A change can achieve the score criterion while worsening functioning or imposing an unacceptable burden. The grade is still recorded, but that change need not be retained as an overall improvement.

A score milestone establishes consistently meeting a named vendor's grading criteria. A reproducible task improvement establishes improvement on that task under the recorded conditions. A controlled personal comparison provides evidence about a change in my setting. None of these alone establishes comprehensive health, better memory in every domain, or added years of life.

The goal remains ambitious. Its claims become stronger as the measurements, comparisons and outcomes improve. Investigations hold the reasoning that can revise these targets.

Target history

13 October 2026: initial specification. An 85-average benchmark follows Oura's grading category; a 90-average stretch is a chosen challenge with feasibility still unknown. Specific wake time and stage quotas remain unassigned pending investigation. No milestone has been achieved or awarded.

14 October 2026: clarified that the first milestone is an average ≥85, not a requirement that every night receive Optimal grading. Thresholds, windows and coverage rules are unchanged. Added the current thermal-comfort/continuity question; its routine check remains a proposal.

15 October 2026: a comfortable but shorter main sleep shifts the immediate question toward sufficient opportunity and final waking. Both providers reported REM despite no recalled dreams. Score milestones are unchanged; midnight and earlier meal/drink timings are proposals, with their benefit unresolved.

16 October 2026: the diary was initially recorded with both providers pending. Oura became available after device/app synchronization and returned a 90-point score and an estimated 8h 37m 30s asleep; COROS remained unavailable at that check, then a later on-demand retrieval returned 98 and 8h59 asleep. Both providers are now recorded with their separate retrieval times. The account reports more opportunity, continuity until a morning water break, and a positive impression. Sleepiness later on 15 October is retained alongside its alert first impression. Earlier meal timing, sleep timing, activity and light changed together, so this is not an isolated meal-timing result. No sustained milestone has been awarded; score thresholds and coverage rules are unchanged.

17 October 2026: recorded the social-night exception with its shorter sleep, late food/alcohol, uncomfortable floor and lower HRV. The date stays in all applicable headline windows. Added the retrospective refreshed-day report for October 16 and the separate October 17 follow-up: awake and sharp on waking, working and feeling good so far, with the rest of the day unassessed. Returning to the earlier routine is planned, not yet an observed recovery; no milestone has been awarded or redefined.

18 October 2026: added a longer night with scores 90/98 and Oura total-sleep/deep-sleep contributors 100. High grading occurred with a roughly four-hour food interval and two-hour gym-to-bed interval, broadening the comparison beyond the earlier six-hour meal gap. Timing 46 remains the weakest contributor. The follow-up reports feeling awake on waking October 18 and feeling good during October 17; October 18 later-day experience remains unreported. This records another high-scoring night, not completion of a 14- or 28-day milestone. The October 17 exception remains included; no threshold or coverage rule changed.

19 October 2026: added matching 91-point grades with close sleep totals, while preserving different episode windows and waking estimates. The diary explicitly reports no urinary awakening after 500 mL water at 22:30–23:00. The follow-up adds approximate 00:10–10:30 opportunity, onset taking a while and initial sleepiness resolving to the usual pace within 10–15 minutes; out-of-bed time and later-day experience remain unreported. Marked Oura firmware 2.1.36 at retrieval after previously observing 2.1.25; update timing and effect are unknown. Seven recorded dates do not complete either sustained milestone. Thresholds and inclusion rules remain unchanged.

20 October 2026: added Oura 92/COROS 89 and the okay-ish impression, many brief awakenings and plentiful dreams. Oura timing rose 51 → 66; its composite improved while estimated sleep shortened and restfulness stayed 88. The recent 90, 91, 92 series does not complete either sustained milestone. The follow-up adds waking around 09:30, strong sleepiness, an unsuccessful 5–10-minute attempt to sleep more, no bathroom trips after awakenings and estimated under-30-second wake episodes. Dreams remain a suspected trigger. The second follow-up clarifies feeling normal within 10–15 minutes, vivid dream recall, no awareness of breathing symptoms, and different subjective strengths for device awake-period timing versus total duration. Sustained later-day experience remains unreported; native estimates are unchanged. No previous-night water amount or other unreported practice was carried forward. Firmware 2.1.36 and algorithm v2 match the preceding retrieval. Thresholds and inclusion rules are unchanged.

21 October 2026: added the fewer-awakenings report, approximate diary timing, short-lived morning sleepiness and 19:00–20:00 last food. Both providers estimated more sleep and less awake time, while Oura’s composite/restfulness fell and COROS’s composite rose. Unusable Oura saturation is unavailable. No new cutoff, sustained milestone or causal conclusion is adopted; previous entries and inclusion rules remain unchanged.

22 October 2026: added a middling diary and COROS 91 alongside the Oura recording ending at 06:30 after reported low battery. Preserved Oura 73 and all component values with an incomplete flag. Explicitly excluded known incomplete recordings from whole-night milestones and high-component achievements without changing numeric thresholds, windows or coverage requirements. Latest eligible Oura score remains October21’s 90; October22 remains visible as a coverage gap. Earlier bedtime is the reported direction, not an established optimum.

23 October 2026: added the cleaner-triggered awakening, late gym/shower context and Oura 83/COROS 88. Today’s Oura episode extends through the main night and its score is eligible; the previous incomplete record remains unchanged and excluded. Ten eligible scores across eleven dates do not meet the existing sustained windows. Timing 52 identifies scoring room, not an established optimal clock time or a causal explanation. The follow-up confirms staying awake and some continuing sleepiness; its duration, resolution and performance effect remain unknown. Native values, thresholds, windows and coverage rules remain unchanged.

Earlier 26 October 2026 retrieval: added native Oura records for October 24–25; October 26 has no returned Oura values yet. COROS was initially unavailable, then successfully retrieved: 94/65 for October 24/25, with native sleep and official HRV retained separately from Oura. The October 26 nap-only detection has no valid main-night score. Twelve eligible Oura scores exist across thirteen dated entries, retaining the October 22 battery-gap exclusion. The follow-up attributes alcohol, sickness, late bedtime and emotions specifically to October 25, with individual effects unresolved. No sustained milestone is awarded; thresholds, calendar windows and coverage rules remain unchanged.

26 October main-night update: added the later complete main-night responses, separate evening detections and first 14-date assessment. Oura 85/COROS 94 replace no earlier recorded provider values; invalid −1 remains only in the earlier availability history. The observed 13-score mean 82.31 falls below 85 despite adequate coverage. The subsequent follow-up reports ongoing body pains, bad sleep, waking sleepy and feeling unhappy overall, plus no naps. COROS’s nap is disputed. Native values, thresholds and inclusion rules remain unchanged; symptom recovery is not established.