Speedcubing performance analysis: what Kruxmind measures and what it does not

KRU

MIND

Speedcubing Tracker

Speedcubing performance analysis: what Kruxmind measures and what it does not

Almost every timer gives you an ao5 and a personal best. Kruxmind adds an analysis layer with one hard rule behind it: no figure shows up without enough data to hold it, and each one says where it comes from.

Two separate layers, never mixed

Worth separating from the start, because they read differently:

What is derived from your solves. It happens on its own, with nothing more than timing: averages, spread, session curve, phase splits.
What you measure separately. Six short exercises in the Performance Suite, each with its own result. Here it is you who sits down to measure.

What you declare is never presented as something the app deduced, and what the app computes is never presented as something you declared. Two columns kept apart on purpose.

What comes out of your solves

The analysis panel works on the history you already have:

Ao5, Ao12, Ao50 and Ao100 drawn as a series, so you see the shape of the curve and not just today's number.
Spread (σ) of your last 12: how far apart your recent times are. A statistical measure, nothing more.
Session curve: whether your times degrade as the session goes on. Past a certain level the panel suggests stopping.
Personal best and distribution of times, to see where they actually cluster.
Mode × phase: which phase leaks the most time when a solve comes out irregular.

The threshold, and why it exists

The panel does not light up with three solves. It needs at least 12 logged solves and, among them, 3 in base mode and 3 under demand. Until then it shows as a preview and says so in those words.

The progress verdict only becomes VALIDATED IMPROVEMENT if the difference between the first and the second half of your history beats the threshold; if it does not, it is flagged INCONCLUSIVE instead of inventing a trend.
The phase cross-check only uses solves with real phase splits (the timer's multiphase mode, which ships off). Without that data the panel stays locked and explains why, rather than splitting the time proportionally and always blaming the same phase.

A threshold is not a product limitation: it is what separates a measurement from a nicely drawn hunch. We would rather say INCONCLUSIVE than hand you a green arrow the data does not support.

The Performance Suite: six measured exercises

The Suite is a separate screen with short exercises, each timed on its own. The six:

Memory Match — working memory, with three modes.
Tower of London (adapted) — planning ahead, 3 to 5 balls. Played in three phases, with its result at the end.
JLAP-15 — line orientation, 15 lines. Three clicks mark three times: analysis, inspection and execution.
FMS — fine motor control over a pattern you repeat.
Simulated rotation — you get an algorithm and four cubes, and pick the one it produces. The same three times: study, review the alternatives, decide.
BLD memo · Speffz — memorisation using the lettering scheme used blindfolded.

The IRK index and its weights, in plain sight

Four of those exercises are summarised into one index, the IRK, with the weights published on the screen itself:

Planning (Tower of London): 35%
Visuospatial (JLAP-15): 25%
Memory (Memory Match): 20%
Fine motor control (FMS): 20%

FMS only feeds the index when measured with a smart cube, which is what allows the regularity of the rhythm to be recorded; measured by hand it still gives you your speed, but the index is computed from the other three and says so as partial. The IRK arranges into one number what you already measured yourself: it is not a diagnosis, nor a promise of a faster time.

Plan, weekly review and history

Around the exercises sit three pieces that give the training continuity: a plan built from your profile and your real availability in minutes and days, a weekly review you fill in yourself, and a history by week with its chart and its table.

Correlations: exploratory, and said so on screen

With eight weeks or more logged, the Suite crosses your own series and shows what rises and falls together. It asks for eight weeks because with less data a correlation means nothing.

A correlation is not a cause. Two of your series moving together does not prove the training caused the improvement, nor that it transfers to the cube. It is material to look at with judgement, not proof of anything.

What is free and what is Pro

The split is not arbitrary: all six exercises are free, along with the profile and the scientific basis. What you pay for is what the system does with your results over time — the plan, the review, the history, the IRK index, the correlations and the advanced metrics.

Advanced metrics are shown apart and with their own notice: they are performance proxies and do not feed the IRK. Extra information, not part of the index.

What it does not do

Kruxmind does not measure brain activity, does not issue diagnoses and does not assess internal states: everything you see comes from your times, your phases and what you log yourself in the exercises. It also does not claim that cubing makes you better at things far from the cube — not grades, not general intelligence, not health. What is traceable and documentable is your progress inside the domain itself.

← Home · Dictionary · FAQ · Solving methods · About