Scoring runs in two stages. The first prices a single event. The second settles a whole day for one category.

​
Stage 1: the event

For each event: measured activity → category scoring logic → category weight → percentage bonuses
  • The base comes from the category’s scoring logic, computed on what was measured: the balance at the day’s reading, a confirmed operation, or a fixed task value.
  • The weight is the category’s current tuning value. Weights are bounded, and every change to one is announced and versioned, never silent. A change applies forward and never reprices p8ints already credited.
  • Percentage bonuses add on the base. They never multiply each other. Where two duration bonuses could apply, the highest qualifying one applies and they do not stack.
Flat bonuses are separate. A destination-chain bonus, a first-trade bonus, a completion bonus: each is its own line item, credited at x1.0, and sitting outside its category’s daily cap.

​
Stage 2: the day

Once a UTC day is closed, each category settles:
  • The multiplier applies to the multiplier-eligible portion of the day, using the multiplier value as of that day’s end.
  • Multiplier-ineligible recurring p8ints are added at face value.
  • The category’s daily cap applies after the multiplier, to the final figure.
  • Rounding happens once, half up, at the end. Internal maths carries more precision than the number you see, and it is only resolved once.
Flat and one-time bonuses are added separately, at x1.0, outside that cap.

​
When a cap binds

Later events in the day truncate first, in chronological order. A capped credit records that the cap applied and credits whatever capacity was left. It never reverts. An honest action can never cost you p8ints that were already earned.

​
Everything is stored

Each event stores the values it was built from: the source value and its unit, the chain, the transaction reference, the base p8ints, the weight, the bonuses, the multiplier, the cap rule if one applied, and the final p8ints. Two consequences:
  • Any row can be replayed and explained, so a support question has a factual answer rather than an opinion.
  • Nothing is recalculated later. A reversal uses the exact figure that was originally credited.

​
Where the chain and the database each stand

Chain-confirmed balances and events are authoritative for The Network Score and Total Earned. The database holds pending calculations, ownership and custody evidence, and the supporting daily readings. When the two disagree, because of a dropped signature, a reverted transaction or a reorg, the chain wins and the database is reconciled to it.