This post covers the daily session that closed on 2026-09-04 across 20 USD-quoted crypto pairs sourced from Coinbase Exchange candles. The session has ended; the figures below are final for that calendar day.

The mechanism described here is the session-summary itself: how breadth, dispersion, and range position interact to produce a regime label, and what that label does and does not carry forward.

What the 2026-09-04 Session Did

Eighteen of the 20 pairs in the panel closed lower on 2026-09-04, placing the session in the broad-decline regime. The median session change was -1.91%, meaning roughly half the panel moved more than that and half moved less — a loss spread across the group rather than concentrated in one or two names. The table carries the full per-pair breakdown; what the summary figures add is the shape of the session: near-universal, moderate, and pulled at the top by a single outlier.

That outlier was NEAR, which closed +11.26% on the session and sat 95% of the way up its own 30-day range — meaning its close was near the highest price it had traded in the prior 30 days. Every other pair that declined sat between 34% and 87% up its 30-day range, suggesting the broader group remained in the middle portions of recent trading territory despite the session loss. ADA, the weakest at -4.46%, closed at 46% of its 30-day range — not at a 30-day low, but in the lower half.

Dispersion — the gap between the strongest and weakest pair — came in at 15.72 percentage points (NEAR at +11.26% minus ADA at -4.46%). That is a wide spread for a session where 18 of 20 names declined; it exists almost entirely because of NEAR's move rather than because the losers were scattered across a wide band. The losing cluster itself ran from -0.83% (ATOM) to -4.46% (ADA), a span of about 3.6 percentage points, which is comparatively tight.

Among the longer-horizon columns, several pairs carried substantial 30-day gains into this session — LINK at +42.67% over 30 days, SOL at +37.83%, XRP at +31.82%, and ETH at +28.80% — yet all closed lower on the day. STX showed the largest 30-day figure in the panel at +98.35% while closing the session at -3.67%. Those 30-day figures describe what happened over the prior month; they do not describe the 2026-09-04 session itself.

The Regime This Session Describes

The broad-decline label applies when the majority of the panel closes lower in the same session. On 2026-09-04, 18 of 20 pairs met that condition. The label is a description of what closed, not a characterisation of the market at any other time.

Broad-decline differs from the dispersed regime in that dispersed sessions show wide variation in direction and magnitude — roughly comparable numbers of advances and declines, or very large moves in both directions. Here, direction was nearly uniform; the high dispersion figure (15.72 percentage points) is almost entirely attributable to NEAR's outlier move rather than to two-sided breadth. A dispersed session and this session can share a similar dispersion number while differing sharply in breadth.

Broad-decline also differs from compressed, where the panel moves together but within a narrow band. The median of -1.91% and a losing range of roughly 3.6 percentage points among the 18 declining pairs is moderate rather than compressed — moves of that size are large enough to be meaningful relative to typical daily ranges, though not extreme by historical standards for this asset class.

What the regime label requires to be true is that the session actually closed: prices are final, candles are settled, and breadth is a count rather than an estimate. What it cannot establish is whether the conditions that produced it were present before the session opened, or whether they will be present in any future session.

Cost Conditions on a Broad-Decline Session

Coinbase Exchange publishes a tiered maker-taker fee schedule; at the lowest volume tier, taker fees run to 0.60% per side, producing a round-trip cost of approximately 1.20% before spread and slippage. At higher volume tiers the taker fee falls, and maker rebates can reduce the round-trip further, but the floor for a retail taker executing two market orders remains material.

On 2026-09-04, the median session change was -1.91%. A position that captured the median move exactly — open to close — would have faced a round-trip taker cost representing roughly 63% of the gross move at the lowest-tier rate (1.20% against 1.91%). That ratio describes how little room existed between the gross price change and the cost of accessing it via market orders for a position held exactly one session.

The pairs with larger moves present a different picture. NEAR's +11.26% session change would absorb the same 1.20% round-trip as a much smaller fraction — about 11% of the gross move. ADA's -4.46% decline sits in between. The practical observation is that on sessions where most moves cluster near the median, the cost-to-move ratio is high for short-duration positions; the outlier moves are the ones where the ratio compresses. Spread and slippage are additional and vary by pair liquidity; the fee schedule provides only the floor.

Strategies that depend on holding period rather than frequency or size — carrying a position across multiple sessions — face the same round-trip cost amortised over a longer gross move. The 30-day figures in the table, where several pairs show gains exceeding +20%, illustrate the scale over which that amortisation occurs, though those figures describe a completed 30-day window rather than any prospective holding period.

What One Session Does Not Establish

A single daily session is a sample of one. The broad-decline label on 2026-09-04 describes what 20 pairs did on that specific day; it does not establish that the conditions producing it were persistent, building, or about to reverse. Both of those framings require more than one observation, and this post provides exactly one.

The regime classification is a label applied after the session closes. It is computed from the session's own data and cannot be falsified by what happens in subsequent sessions — a broad-decline session followed by a broad-advance session does not mean the label was wrong; it means two different sessions had different outcomes. The label carries no predictive content by construction.

NEAR's +11.26% outlier move is the most visually prominent feature of the session. Single-session outlier moves in this asset class are structurally unreliable as signals: they may reflect an idiosyncratic catalyst, a liquidity event, a short squeeze, or noise. The 7-day figure for NEAR (+19.23%) and the 30-day figure (+28.13%) show that the session gain arrived in the context of a broader recent move, but those figures also describe completed windows rather than causes. Nothing in the session data identifies why NEAR moved, and the data presented here would not support that inference even if it did.

Range position figures show where each pair's close sat inside its own 30-day high-low range. A close at 95% of the 30-day range (NEAR) is near the 30-day high; a close at 34% (APT) is in the lower portion. These are positional descriptions of a completed session. They do not indicate whether a pair is more or less likely to move toward either end of the range on any future session.

The 2026-09-04 session closed with near-uniform losses across the panel, a single sharp outlier in NEAR, and most pairs sitting in the middle portions of their recent 30-day ranges — a configuration the regime label captures in one word, and that the data behind it qualifies in several.

Note: This explains how a mechanism works. It is educational and is not financial, investment or trading advice, not a recommendation, and not a signal service. Any figure describes a session that has already closed and says nothing about what happens next. Trading cryptoassets risks losing money, including all of it.