Three wins and a loss can price four different ways.

The structure

Four positions, four different weights, and the pattern repeats across gates.

Read the growth rows and the arithmetic is exact. A clean sweep prices the top of the band. Move a single loss through the four positions and the pool steps down by four points, then two, then one, then nothing — so the same three-and-one record produces four distinct pool figures. The weights combine additively, so two losses cost the sum of their positions' weights, and the worst case lands exactly at the bottom of the band. Recovery uses the same ladder of weights against a lower base. Buffer uses a slightly compressed version. The construction is consistent enough across gates to be a deliberate scheme rather than a set of hand-tuned rows.

Why grade at all

A counting rule would make three of the four positions meaningless.

The obvious alternative is to count wins, which would map sixteen sequences onto five outcomes and treat every three-and-one identically. Grading by position is more expensive to build and buys two things. It produces a smoother response, so the pool moves in small steps rather than jumping between five levels, which matters when the pool is being subtracted from and divided by four. And it lets the system express that not all losses in a cycle carry the same information — a claim that is either true or false about trading, and which the design has evidently decided is true. A counting rule cannot express it at all, because a count discards the ordering before any weighting could be applied.

The shape of the weighting

One position dominates, one is nearly free, and the middle two grade between.

The weights are not evenly spread. In a growth row the heaviest position costs four times what the third costs and the fourth is free entirely, which means a single position carries most of the sequence's influence and the remainder taper away. That concentration is what makes the scheme informative rather than merely smooth: if all four weights were equal, the ordering would carry no signal and the design would be a counting rule with extra steps. The taper also bounds the total, so the worst case — every position lost — reaches the bottom of the band and no further. The band's width is therefore the sum of the weights, which is why growth spans six points and the deeper gates span less.

The convention question

Which end of the sequence carries the heavy weight is worth confirming before relying on it.

One thing the matrix cannot settle on its own is the reading direction. The weighting is unambiguous — one terminal position costs four and the other costs nothing — but whether the heavy end is the most recent trade or the oldest depends on how the sequence is written, and the table does not say. The interpretation that fits the rest of the system is recency: the trade nearest the pending decision carries the most weight, and a loss four trades ago has been superseded. That reading is consistent with how the rolling windows treat recent evidence and with the tactical layer's role generally. It is nonetheless an inference from design intent rather than something the data states, and it belongs in the same category as any other convention that should be confirmed once and then documented, because reading it backwards would invert the response to every cycle.

What it does not do

The weighting moves the pool inside the band and never moves the band.

It is worth restating the boundary, because a graded response to recent results can look like performance earning authority. It is not. The entire weighting scheme operates within a band whose edges were set by the gate, so the best possible sequence reaches the top of the current band and the worst reaches the bottom, and neither touches the band above or below. Results decide where inside the capital state the account sits; drawdown decides which capital state that is. The two questions are answered by different mechanisms with different inputs, and the outcome matrix is only ever answering the second one.

• The matrix reads sequence, not tally — the same record prices several ways.
• Weights concentrate in one position and taper to zero, which is what makes ordering informative.
• Confirm the reading direction once; reversing it would invert every response.

The key idea

Graded responses are how a system reacts to small samples without over-reacting.

Four trades is not enough evidence to justify a large change in deployment, and it is not nothing either. A binary rule would have to choose between ignoring the cycle and treating it as decisive. A graded weighting takes the third option: respond, in proportion, within limits set elsewhere, so that a poor cycle costs a few percentage points of pool rather than a tier and a good one earns the same in return. The response is real enough to matter and bounded enough that no single cycle can reprice the account — which is exactly the property a four-trade sample deserves.