Each completed trading day, the Session desk publishes a single dated post covering twenty cryptocurrency pairs quoted in USD, sourced from Coinbase Exchange daily candles. The post contains a six-column table, five summary statistics beneath it, and one coarse regime label. Together those elements describe the session that just closed — nothing more and nothing less.

The Regimes desk is concerned with how those labels and statistics are derived, what they measure reliably, and where a reader can be misled by treating a descriptive classification as something it is not. This piece works through each part of the table in the order it appears, then covers the summary line and the regime label.

What Each Column in the Table Actually Records

Pair. The first column names the trading pair — a crypto asset ticker followed by USD. Twenty pairs appear in every post. The set is fixed per post; it does not expand or contract based on volume or news.

Close (USD). The second column records the last traded price of the daily candle for that pair on Coinbase Exchange. It is a single point in time — the candle's close — not an average and not a bid or ask. It carries no forward implication.

Session (% change, open→close). The third column expresses the percentage change from the candle's open to its close for that same day. A session figure of −1.94% on a given date means the pair's last price was 1.94% below its opening price on that date. The figure describes intraday drift — how far the market moved between the open print and the close print within the single session. It says nothing about where price sat at any point in between, because the table does not render the high or the low.

7d (% change). The fourth column records the percentage change in closing price over the seven calendar days ending on the session date. It spans seven daily closes, not seven trading sessions, and does not adjust for any gap or holiday convention. A positive figure means the most recent close is higher than the close seven days prior; a negative figure means the reverse. The 7d column is the shortest lookback in the table and therefore the most sensitive to recent momentum shifts.

30d (% change). The fifth column applies the same logic over thirty calendar days. It provides the medium-term frame against which the 7d figure can be read in context. A pair showing a strongly negative 7d reading against a strongly positive 30d reading is in a different structural position than one where both figures point the same direction — but the table makes no judgment about which state is preferable. It records the arithmetic; the interpretation is left to whoever is reading.

Range position (0%–100%). The sixth and final column answers a different question from the others. Rather than measuring change from a fixed prior point, it places the session close inside the pair's own price range over the preceding thirty days. A reading of 0% means the close matched the thirty-day low exactly; 100% means it matched the thirty-day high. A reading of 50% means the close sat at the midpoint of that range. This column is the only one that normalises across pairs — because it expresses each pair relative to its own history, a volatile pair and a quiet pair become comparable on the same 0–100 scale. The thirty-day range used here is the same thirty-day window as the 30d change column, so the two columns share a lookback but measure different things: the 30d column measures net drift, while the range position column measures where the close sits within the full extent of price movement over that period.

Summary statistics beneath the table. Below the six-column grid, each post states five additional figures. Breadth counts how many of the twenty pairs closed higher on the session than they opened — it is a whole number between 0 and 20. The median session change is the middle value of the twenty session-percentage figures, providing a centre-of-distribution read that is less affected by outliers than a simple average. The strongest and weakest pair are identified by name alongside their session-percentage figures. Dispersion is the arithmetic difference between the strongest and weakest session figures, expressed in percentage points; a dispersion of 8 percentage points on a given date means the best-performing pair outpaced the worst by 8 percentage points on that session.

The regime label. Each post carries exactly one of five coarse labels: broad-advance, broad-decline, compressed, dispersed, or mixed. The label is assigned to the session that closed and to no other session. Broad-advance and broad-decline are driven primarily by breadth — how many pairs moved in the same direction. Compressed describes a session where most pairs moved very little in either direction, producing low dispersion and a median session change close to zero. Dispersed describes a session where the spread between the strongest and weakest pair was wide, regardless of the direction of the median. Mixed describes a session that does not satisfy the threshold for any of the other four labels cleanly — breadth is split, dispersion is moderate, and no single character dominates.

What the Table Needs to Be True That It Cannot Make True

The table's usefulness depends on several conditions that the table itself cannot establish or verify.

First, the candle data must be complete and consistent across all twenty pairs. If a pair experienced a trading halt, a data gap, or an anomalous print on the source exchange during the session, the close figure, the session percentage, and the range position for that pair are all affected. The table does not flag anomalous prints; it records whatever the candle reports.

Second, the thirty-day range used for range position is only as meaningful as the trading activity within it. In a period where a pair traded in an extremely narrow band for most of the thirty days and then moved sharply in the final few sessions, the range position figure will cluster near 0% or 100% for nearly every close in that final burst — not because the pair is at a structural extreme, but because the range denominator is dominated by a single move. The column describes position within the observed range; it does not describe whether that range is representative of normal conditions.

Third, the regime label depends on thresholds. Those thresholds are fixed rules applied to the five summary statistics. A session where breadth is 11 out of 20 — one pair above the midpoint — may receive a different label than a session where breadth is 10, even though the underlying market character is nearly identical. The label is a coarse classification, not a continuous score, and the boundaries between categories are definitional rather than natural.

Fourth, the table covers one venue. Coinbase Exchange prices for these twenty pairs may differ from prices on other venues at the moment of the candle close, due to differences in order flow, liquidity, and the precise timing of the close print. The table does not represent a consolidated tape or a volume-weighted cross-venue price.

The Cost of Using the Table as an Input to a Trading Decision

The table itself has no direct financial cost to read. The cost structure discussed here concerns what a trader faces if the table's outputs are used to inform a round-trip trade on a crypto spot venue.

Coinbase Exchange publishes a tiered maker-taker fee schedule. At the lowest volume tier (trailing 30-day volume under $10,000), the taker fee is 0.60% per side as of the schedule in effect at the time of writing, implying a round-trip taker cost of approximately 1.20% of notional on that tier. Maker fees at the same tier are 0.40% per side, implying a round-trip maker cost of approximately 0.80% of notional. These figures are per-trade costs before spread and slippage.

Spread — the difference between the best bid and best ask at the moment of execution — adds to the round-trip cost and is not fixed. For liquid pairs in normal conditions, the spread on Coinbase Exchange has historically been a small fraction of a percent, but it widens during periods of low liquidity or rapid price movement. The table does not report spread; it reports close prices, which are last-trade prices and may not reflect the mid-market at the moment of the close.

Slippage — the difference between the expected execution price and the actual fill — is a function of order size relative to available liquidity at each price level. The table provides no information about depth of book and therefore no information about the slippage a given order size would incur.

A strategy that uses the session table as its primary signal must clear the full round-trip cost — taker fee plus spread plus slippage — before it produces a positive result. The three structural paths to clearing that floor are high trade frequency (requiring many small edges to accumulate), large position size (requiring the edge per trade to be large enough in absolute terms), or a long holding period (allowing a smaller percentage edge to compound over many sessions). The session table, by design, publishes one reading per day; it is a daily-frequency instrument, which constrains the frequency path and points toward holding period as the primary mechanism for clearing costs in any strategy that relies on it.

Where the Table's Classification System Breaks Down

The most common structural failure of a regime-classification system is label persistence — the tendency of a label assigned today to persist into tomorrow not because market conditions are genuinely stable, but because the inputs to the classification change slowly. The thirty-day lookbacks used in the range position and 30d change columns are heavily weighted toward past sessions; a single new session represents 1/30th of the lookback. This means the range position column and the 30d column are slow to register regime transitions. A market that has shifted character two or three sessions ago may still produce table readings that resemble the prior regime, because the thirty-day window has not yet absorbed enough new data to reflect the change.

The regime label itself is assigned on the basis of a single session's summary statistics. A label of "broad-advance" means that on the session that just closed, most of the twenty pairs moved higher. It does not mean that a trend is in place, that momentum is sustained, or that the following session will share the same character. Treating the label as a trend signal is a category error — it is a description of one closed session, not a forecast of the next.

Dispersion, which drives the "dispersed" label and informs the "mixed" label, is itself an unstable measure at small sample sizes. Twenty pairs is a small cross-section. A single pair with an idiosyncratic move — a protocol event, a liquidity incident, or a data anomaly on the source exchange — can shift the dispersion figure and the regime label for the entire post without any change in the behaviour of the other nineteen pairs. The label, in that case, reflects an outlier rather than a genuine market-wide condition.

Finally, the compressed label presents a specific failure mode for strategies designed to trade breakouts from compression. Compression — low dispersion, median session change near zero — is a necessary but not sufficient condition for a subsequent expansion in range. The table identifies compression accurately when it is present. It cannot identify whether the compression will resolve in the direction of the prior trend, against it, or simply persist for additional sessions. A strategy built on the premise that compression precedes directional expansion will encounter sessions where compression persists or resolves in a direction that costs the strategy its round-trip fee. That failure is not a flaw in the table's construction; it is the terminal condition of the pattern being traded.

The daily session table is a structured record of what twenty crypto pairs did on one exchange during one completed day — its columns are arithmetic, its labels are definitional, and its summary statistics are descriptive; none of those properties change based on how the table is used or what a reader hopes to find in it.

Note: This explains how a process works. It is not legal advice, it is not specific to any debt, and it is not a substitute for a licensed attorney in your state. Rules and time limits vary by state and change over time — check the cited sources.