Bolt Series · Human Performance Calibration · Article 31
Your best performance and your usual performance answer different questions
A student scores 96%.
It is the best result they have ever produced.
What does it mean?
It means something important.
Under those conditions, on that task, the learner produced a 96% performance.
That expands what we know is possible.
But suppose the next four scores are:
- 78%
- 82%
- 76%
- 81%
Is the student a 96% learner?
Or an 80% learner?
Neither label is especially useful.
The 96 and the repeated 76–82 range are telling us different things.
A peak tells us what the system has reached. A baseline tells us what the system can usually reproduce.
Sport lives with this distinction constantly
An athlete has a personal best.
That number matters enormously.
But coaches do not assume the athlete will reproduce the personal best every competition.
A systematic review of elite competitive performance found measurable within-athlete variability from competition to competition.
Even sprint and endurance events, which showed relatively low variability compared with many other sports, still varied.
For elite sprint and endurance performance, the review reported typical coefficients of variation in roughly the 0.6% to 1.4% range.
That may sound tiny.
At elite level, tiny differences can separate winning from losing.
The important point is this:
Even an elite athlete is not one permanently available number.
The peak is not fake
There is a temptation to treat an unusually high result as luck.
Sometimes luck contributes.
Perhaps the test happened to sample the learner’s strongest areas.
Perhaps conditions were unusually favourable.
But if the work itself demonstrates strong reasoning, accurate execution and independent performance, the peak is still evidence.
It proves that the learner’s current system can generate something at that level.
The correct question is not:
“Was the 96 real?”
It is:
“What conditions made 96 possible, and how often can the learner reproduce them?”
The baseline is not a ceiling
The opposite mistake is just as common.
A learner usually scores around 70%.
Adults begin saying:
“She is a 70% student.”
Now the baseline has quietly become an identity and a ceiling.
But a baseline is only a current pattern of reproducibility.
It can move.
Teaching can improve it.
Practice can stabilise it.
A change in task demands can expose that the old baseline was too narrow.
Development can move the entire distribution.
A baseline is a description of what currently repeats.
It is not a prediction that nothing better can become normal.
A useful performance model has at least three levels
Instead of one label, imagine three questions.
1. What is the current baseline?
What performance appears repeatedly under representative conditions?
2. What is the current peak?
What has the learner demonstrated is possible under at least some conditions?
3. What is the current variability?
How far does performance move around the baseline from one meaningful attempt to another?
Those three quantities tell a richer story.
A learner with a baseline of 80 and a narrow range of 78–83 is different from a learner who averages 80 but fluctuates between 62 and 96.
The average may be the same.
The performance system is not.
Consistency is a capability too
Imagine two students preparing for a major examination.
Student A has scored:
82, 81, 84, 80, 83.
Student B has scored:
64, 95, 73, 91, 79.
Their averages may not be dramatically different.
But their risk profiles are.
Student B may possess a very high peak and an unstable ability to reproduce it.
The educational goal may not initially be to create a higher peak.
It may be to make the existing strong performance more available more often.
Improvement can mean raising the peak, raising the baseline, or reducing harmful variability. Those are different training problems.
This changes how we interpret personal bests
A personal best should make us curious.
What was different?
- Was the learner better rested?
- Was the task format unusually favourable?
- Did a new strategy work?
- Was retrieval stronger after spacing?
- Was time management better?
- Did confidence become more accurately matched to difficulty?
- Did the learner finally transfer a concept that had previously remained procedural?
If we can identify the mechanism, the peak becomes more than a trophy.
It becomes a clue about how to move the baseline.
And this changes how we interpret one bad performance
Suppose a stable 82% learner suddenly scores 63%.
That low result matters.
But it does not automatically move the baseline to 63.
It may instead tell us that the current performance range is wider than we thought, or that a particular condition exposes a vulnerability.
If the next several results also fall, the baseline model should update.
If performance returns to the previous range and the low result has a clear state or task explanation, the model updates differently.
Again, one observation enters the model.
It does not automatically own it.
Baseline depends on the conditions we care about
There is another subtlety.
A learner can have different baselines in different conditions.
- Untimed baseline.
- Timed baseline.
- Supported baseline.
- Independent baseline.
- Routine-question baseline.
- Transfer-question baseline.
Which one matters depends on the decision.
If we are asking whether the child understands a concept, untimed explanation may be highly informative.
If we are asking whether the learner is examination-ready, a realistic timed baseline matters too.
“Baseline” is therefore not one magical number hidden inside the learner.
It is a pattern of performance under specified conditions.
Why this matters for parents
Parents naturally notice peaks.
“You got 95 last time. Why did you only get 78 now?”
But this question assumes the peak was the baseline.
A better conversation is:
“95 shows us what you can reach. Your recent pattern tells us what you can currently reproduce. What would help make more of that 95-level performance available more often?”
That protects ambition without distorting measurement.
Why this matters for education
A single number encourages us to ask:
“How good is this student?”
A performance distribution asks better questions:
- What level appears reliably?
- What level has been demonstrated at the peak?
- How variable is the learner?
- Which conditions shift the distribution?
- What intervention would raise the baseline rather than merely create another isolated high?
That is a much stronger basis for teaching.
Your peak tells us what has been possible. Your baseline tells us what is currently dependable. Education should help more of your best performance become ordinary.
Then, when the baseline rises, the next peak can move again.
