Revision 7T · Full Tutorial Edition · Updated 1 September 2026
PWR-055 · SELF GUIDED full tutorial
Calibrate a recovery prediction against what actually happens
Recovery forecasts may help planning, but a score is only as good as its device, model, context and calibration. A recovery forecast is useful only when made before the outcome and checked later. The learner predicts one benign task from declared inputs, measures the actual result, calculates error across repeated days and changes only one forecast rule. Wearable scores remain inputs, not authority.
What you will produceThe learner completes ten prospective predictions for the supplied 60-second code-check cards and reports signed error, mean absolute error, prediction-band calibration and whether one revised rule helped on fresh cards.
Full step-by-step individual tutorial · TLU-PWR-055
The learner completes ten prospective predictions for the supplied 60-second code-check cards and reports signed error, mean absolute error, prediction-band calibration and whether one revised rule helped on fresh cards.
Write a 0/5/10 prospective score before opening each benign 60-second code-check card, then run the card while ordinarily rested.
Calculate hits, misses, false alarms, actual score, signed error and absolute error from the supplied key/formula and test one revised forecast rule only on fresh cards.
Record sleep/wearable/context inputs as clues while keeping the forecast away from driving, medicine, work clearance or another consequential decision.
Qualified help is required for
Interpreting persistent fatigue, sleepiness, illness or unexpectedly poor function and deciding whether medical assessment/treatment is needed.
Any fitness-for-duty, driving, sport-risk, medication, patient-care or occupational use of a readiness forecast.
Changing sleep treatment, medication, workload or rehabilitation from wearable/recovery scores.
Never do this from the page alone
Shorten sleep, intensify exercise or create illness/fatigue to generate harder forecast days.
View the key before predicting, rewrite a prediction after the card or omit a bad day from calibration.
Let a predicted 10 or a wearable badge override symptoms, an objective error, a safety rule or professional restriction.
2 · Get ready
Gather what you need and check the starting conditions
What you need
A ten-row forecast sheet and ten exact 60-second code-check cards. Each card has eight left/right code pairs and exactly two mismatches: A 4821/4821, 7306/7306, 9145/9154, 2680/2680, 5573/5573, 6419/6419, 8032/8033, 1194/1194; B 3148/3148, 7250/7520, 6681/6681, 9024/9024, 1573/1573, 4862/4867, 2395/2395, 8106/8106; C 5740/5748, 1326/1326, 9951/9951, 4073/4073, 6812/6812, 2509/2509, 7438/7438, 3264/3624; D 8610/8610, 2457/2457, 7039/7039, 1186/1816, 9342/9432, 5270/5270, 6604/6604, 3891/3891; E 4408/4408, 1725/1729, 6931/6931, 8264/8264, 3510/3510, 9072/9072, 5846/5486, 2183/2183; F 7602/7602, 3319/3319, 8457/8475, 1096/1096, 6243/6243, 5780/5788, 2925/2925, 4137/4137; G 9350/9358, 2046/2046, 7812/7812, 5684/5684, 3279/3729, 6501/6501, 1198/1198, 4423/4423; H 2861/2861, 7705/7705, 5314/5314, 9402/9042, 1638/1638, 6257/6257, 8091/8091, 3746/3748; I 6173/6173, 2085/2805, 9541/9541, 4860/4860, 7329/7392, 1507/1507, 8634/8634, 3298/3298; J 7031/7031, 2648/2648, 8195/8159, 4470/4470, 5362/5362, 1904/1904, 6827/6872, 3251/3251.
A folded key giving mismatch positions A 3,7; B 2,6; C 1,8; D 4,5; E 2,7; F 3,6; G 1,5; H 4,8; I 2,5; J 3,7. For each card count hits, misses and false alarms, then calculate actual task score = maximum of 0 and 10 − 2×misses − false alarms. Keep raw counts beside this non-clinical 0–10 score.
A fixed readiness-prediction scale: 0 means expect no reliable code checking today, 5 means expect several uncertain pairs, and 10 means expect all eight decisions correct without a false alarm.
Optional wearable value recorded as one input, with device and version named.
Before you start
Choose a task whose poor result cannot harm anyone; do not use driving, medication, water, heights or work clearance.
Use the supplied 60-second card rule and actual-score formula unchanged on all ten rows.
Keep task difficulty and time similar across days.
If severe or unexplained symptoms exist, seek care rather than running a forecast.
3 · The method
Follow these steps in order
Define the future task
Write the next unused card letter, 60-second limit, mismatch-marking rule, actual-score formula and planned clock time before looking at the folded key.
Why: A forecast needs a fixed target, not a general feeling.
Check: Another person could run the same task from the description.
Choose declared inputs
Record only the inputs chosen in advance: sleep opportunity, symptoms, prior load, caffeine timing and optional named device score.
Check: The row has the same input columns each day.
Make the prediction first
Before the task, predict readiness from 0 to 10 and the concrete task result, then lock the row.
Why: Prospective prediction can be tested; retrospective explanation cannot.
Check: The prediction has a timestamp earlier than the task result.
Perform the benign task
Open the named code card, compare its eight pairs for exactly 60 seconds and circle every suspected mismatch without opening the key or changing aids.
Why: A stable outcome makes forecast error interpretable.
Check: Time, item version and support are recorded.
Score the actual result
Open the folded key, count hits, misses and false alarms, then compute actual score = maximum of 0 and 10 − 2×misses − false alarms.
Why: Confidence and effort are not the outcome.
Check: The raw counts remain beside the converted score.
Calculate forecast error
Subtract predicted from actual score and also record the absolute difference.
Why: Signed error shows optimistic or pessimistic bias; absolute error shows size.
Check: Both numbers are calculated for every row.
Check calibration after ten rows
Group low, middle and high predictions and compare each group with its mean actual result.
Why: One lucky forecast does not establish a useful model.
Check: The report includes all ten rows and group averages.
Change one rule
If consistently optimistic or pessimistic, alter one input weight or decision threshold and predeclare the next ten-row test.
Why: A single change can be evaluated; wholesale rewriting cannot.
Check: The revised rule is written before new outcomes occur.
4 · Worked example
See the whole method used once
Scenario
Priya forecasts her score before using code-check Card A for 60 seconds.
Walkthrough
Before opening Card A, she records sleep opportunity and symptoms, predicts 8/10 and timestamps the row.
In 60 seconds she circles pairs 3, 6 and 7; the folded key says only positions 3 and 7 differ.
She records two hits, zero misses and one false alarm, so actual score = 10 − 2×0 − 1 = 9.
Signed error is actual minus predicted = 9 − 8 = +1; absolute error is 1. She keeps the false alarm rather than editing the circle.
After Cards A–J, her high-prediction rows average 8.5 predicted and 7.0 actual, so she predeclares one revision: a prediction above 8 requires both her stated sleep opportunity and low symptom input, then tests fresh cards rather than refitting A–J.
Result
Priya has a calibrated personal forecast for this proofreading task. It is not permission to drive, work unsafely or treat the device score as recovery.
5 · Right and wrong
Compare correct or safer execution with the common wrong version
Right and wrong comparison
Moment
Right / safer
Wrong / riskier
Why it matters
Timing
Lock the prediction before the task.
Explain afterward why the result was predictable.
Hindsight cannot test forecast accuracy.
Outcome
Use raw errors and a fixed scoring rule.
Use confidence as the actual result.
Confidence can be miscalibrated.
Wearable
Name the device score as one optional input.
Let the device label override symptoms or hazards.
Consumer scores have device-specific error and are not clearance.
Sample
Evaluate at least ten prospective rows.
Promote one correct prediction as a reliable model.
Calibration needs repeated outcomes.
Update
Change one rule and retest prospectively.
Rewrite the rule after every miss.
Continuous hindsight fitting hides failure.
6 · Common mistakes
Spot the error and apply the correction
Common mistakes and corrections
Mistake
Fix
Explain afterward why the result was predictable.
Lock the prediction before the task.
Use confidence as the actual result.
Use raw errors and a fixed scoring rule.
Let the device label override symptoms or hazards.
Name the device score as one optional input.
Promote one correct prediction as a reliable model.
Evaluate at least ten prospective rows.
Rewrite the rule after every miss.
Change one rule and retest prospectively.
7 · Practice
Turn the steps into a usable skill
First session
Copy the three readiness anchors, actual-score formula and fixed input columns onto the forecast sheet.
Write Card A, the planned clock time and a 0–10 prediction before opening the card.
Compare all eight Card A pairs for 60 seconds and freeze the circles when the timer ends.
Use the folded A key to count hits, misses and false alarms, calculate actual score, signed error and absolute error, then check the arithmetic.
Schedule Cards B–J on comparable occasions before interpreting the low, middle and high prediction bands.
Repeat plan
Collect one forecast per day or per comparable occasion until ten are complete. Review bias and absolute error, change one rule, then collect ten new forecasts. Retire the rule if it does not improve or creates burden.
Progress when
Predictions are timestamped before outcomes.
All raw task errors and contextual deviations are retained.
A revised rule reduces error on a new set rather than the old rows.
Do not progress when
The predicted task becomes safety-critical or consequential.
Illness, medication or device change makes old rows incomparable.
Forecasting increases anxiety, compulsive checking or pressure to ignore symptoms.
8 · Check the result
Measure what changed
Prospective calibration and decision benefit for a declared future task
How: For each row calculate signed error = actual score − predicted score and absolute error = |signed error|. Mean absolute error is the sum of all absolute errors divided by the number of completed rows. Also report mean prediction and mean actual score within the predeclared low, middle and high prediction bands.
Good result: All predictions precede their unused cards, every score reproduces from the folded key and formula, and any revised rule is judged only on fresh cards. A lower fresh-card mean absolute error is task-specific evidence to continue checking, not recovery or safety clearance.
This does not prove: Forecast discrimination, calibration and decision utility must be tested against a future declared outcome; a readiness score is not the outcome itself.
Self-check
Can you demonstrate “Choose declared inputs”? The row has the same input columns each day.
Can you demonstrate “Score the actual result”? The raw counts remain beside the converted score.
Can you demonstrate “Change one rule”? The revised rule is written before new outcomes occur.
9 · Stop, adapt or get help
Keep the safety boundary practical
Stop and get help
Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
Seek clinical advice for persistent fatigue, sudden sleep episodes, chest symptoms, breathlessness or marked functional decline.
Stop forecasting if it drives compulsive checking, sleep loss or risky attempts to prove the score wrong.
Accessibility and adaptations
Use a spoken or pictorial task with an objective key if proofreading is inaccessible.
Let a helper timestamp and score while keeping the learner blind to the answer key.
Retain normal aids and score supported performance as the declared configuration.
10 · Evidence and limits
Why these instructions are here
primary research
Direct comparison with polysomnography shows that consumer sleep devices differ in performance, so device output should be treated as an input with error.
Forecast discrimination, calibration and decision utility must be tested against a future declared outcome; a readiness score is not the outcome itself.
Unsupported claim: your wearable knows when you are recovered.
Evan D Chinoy; Joseph A Cuellar; Kirbie E Huwa; Jason T Jameson; Catherine H Watson; Sara C Bessman; Dale A Hirsch; Adam D Cooper; Sean P A Drummond; Rachel R Markwald · 2021 · Primary research
Progress is saved only in this browser on this device.
Step-by-step learner mode
Each activity includes its success check, a nearby accessible alternative and an “I’m stuck” correction path. Alternatives preserve the target where possible; when they change the task, Titan labels them as related rather than equivalent.
01
Define the future task
Write the next unused card letter, 60-second limit, mismatch-marking rule, actual-score formula and planned clock time before looking at the folded key.
Why this step exists
A forecast needs a fixed target, not a general feeling.
Success check
Another person could run the same task from the description.
I’m stuck on this step
Reset: Re-read this authored instruction — “Write the next unused card letter, 60-second limit, mismatch-marking rule, actual-score formula and planned clock time before looking at the folded key.” — and its success check, then attempt only this step.
Possible snag: Use confidence as the actual result.
Correction: Use raw errors and a fixed scoring rule.
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
02
Choose declared inputs
Record only the inputs chosen in advance: sleep opportunity, symptoms, prior load, caffeine timing and optional named device score.
Reset: Re-read this authored instruction — “Record only the inputs chosen in advance: sleep opportunity, symptoms, prior load, caffeine timing and optional named device score.” — and its success check, then attempt only this step.
Possible snag: Let the device label override symptoms or hazards.
Correction: Name the device score as one optional input.
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
03
Make the prediction first
Before the task, predict readiness from 0 to 10 and the concrete task result, then lock the row.
Why this step exists
Prospective prediction can be tested; retrospective explanation cannot.
Success check
The prediction has a timestamp earlier than the task result.
I’m stuck on this step
Reset: Re-read this authored instruction — “Before the task, predict readiness from 0 to 10 and the concrete task result, then lock the row.” — and its success check, then attempt only this step.
Possible snag: Explain afterward why the result was predictable.
Correction: Lock the prediction before the task.
Possible snag: Promote one correct prediction as a reliable model.
Correction: Evaluate at least ten prospective rows.
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
04
Perform the benign task
Open the named code card, compare its eight pairs for exactly 60 seconds and circle every suspected mismatch without opening the key or changing aids.
Why this step exists
A stable outcome makes forecast error interpretable.
Success check
Time, item version and support are recorded.
I’m stuck on this step
Reset: Re-read this authored instruction — “Open the named code card, compare its eight pairs for exactly 60 seconds and circle every suspected mismatch without opening the key or changing aids.” — and its success check, then attempt only this step.
Possible snag: The result from “Open the named code card, compare its eight pairs for exactly 60 seconds and circle every suspected mismatch without opening the key or changing aids.” does not yet meet this declared check: Time, item version and support are recorded.
Correction: Return to the start of “Perform the benign task”, reduce complexity or pace, and repeat only the part needed to satisfy: “Time, item version and support are recorded.”
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
05
Score the actual result
Open the folded key, count hits, misses and false alarms, then compute actual score = maximum of 0 and 10 − 2×misses − false alarms.
Why this step exists
Confidence and effort are not the outcome.
Success check
The raw counts remain beside the converted score.
I’m stuck on this step
Reset: Re-read this authored instruction — “Open the folded key, count hits, misses and false alarms, then compute actual score = maximum of 0 and 10 − 2×misses − false alarms.” — and its success check, then attempt only this step.
Possible snag: The result from “Open the folded key, count hits, misses and false alarms, then compute actual score = maximum of 0 and 10 − 2×misses − false alarms.” does not yet meet this declared check: The raw counts remain beside the converted score.
Correction: Return to the start of “Score the actual result”, reduce complexity or pace, and repeat only the part needed to satisfy: “The raw counts remain beside the converted score.”
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
06
Calculate forecast error
Subtract predicted from actual score and also record the absolute difference.
Why this step exists
Signed error shows optimistic or pessimistic bias; absolute error shows size.
Success check
Both numbers are calculated for every row.
I’m stuck on this step
Reset: Re-read this authored instruction — “Subtract predicted from actual score and also record the absolute difference.” — and its success check, then attempt only this step.
Possible snag: The result from “Subtract predicted from actual score and also record the absolute difference.” does not yet meet this declared check: Both numbers are calculated for every row.
Correction: Return to the start of “Calculate forecast error”, reduce complexity or pace, and repeat only the part needed to satisfy: “Both numbers are calculated for every row.”
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
07
Check calibration after ten rows
Group low, middle and high predictions and compare each group with its mean actual result.
Why this step exists
One lucky forecast does not establish a useful model.
Success check
The report includes all ten rows and group averages.
I’m stuck on this step
Reset: Re-read this authored instruction — “Group low, middle and high predictions and compare each group with its mean actual result.” — and its success check, then attempt only this step.
Possible snag: The result from “Group low, middle and high predictions and compare each group with its mean actual result.” does not yet meet this declared check: The report includes all ten rows and group averages.
Correction: Return to the start of “Check calibration after ten rows”, reduce complexity or pace, and repeat only the part needed to satisfy: “The report includes all ten rows and group averages.”
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
08
Change one rule
If consistently optimistic or pessimistic, alter one input weight or decision threshold and predeclare the next ten-row test.
Why this step exists
A single change can be evaluated; wholesale rewriting cannot.
Success check
The revised rule is written before new outcomes occur.
I’m stuck on this step
Reset: Re-read this authored instruction — “If consistently optimistic or pessimistic, alter one input weight or decision threshold and predeclare the next ten-row test.” — and its success check, then attempt only this step.
Possible snag: Rewrite the rule after every miss.
Correction: Change one rule and retest prospectively.
Stop / get help: Do not use the forecast to override drowsiness, acute illness or a workplace stop rule.
Correct versus incorrect execution
These accessible process diagrams are built from the tutorial’s own right/wrong teaching. They are not anatomical illustrations and do not add technique beyond the canonical tutorial.
Timing — Hindsight cannot test forecast accuracy.
Correct / safer
Lock the prediction before the task.
Wrong / riskier
Explain afterward why the result was predictable.
Outcome — Confidence can be miscalibrated.
Correct / safer
Use raw errors and a fixed scoring rule.
Wrong / riskier
Use confidence as the actual result.
Wearable — Consumer scores have device-specific error and are not clearance.
Correct / safer
Name the device score as one optional input.
Wrong / riskier
Let the device label override symptoms or hazards.
Sample — Calibration needs repeated outcomes.
Correct / safer
Evaluate at least ten prospective rows.
Wrong / riskier
Promote one correct prediction as a reliable model.
✓ Materials or supplied records are declared — present
✓ Measurement or assessment rule is present — present
✓ Tutorial-specific troubleshooting is present — present
✓ Stopping or escalation boundary is present — present
✓ Every activity has an adjacent alternative — present
✓ Correct-versus-incorrect comparison is present — present
✓ Evidence context is bound to the Power record — present
✓ Planning metadata is present — present
The method-readiness band and presence checklist assess tutorial presentation and are separate from evidence quality for the underlying Power. They are automated editorial aids, not human approval.
Manual editorial sign-off: Pending. This tutorial must not display a human-approved state until an identified editor signs the exact content hash.