Regime Duration Profile

What is it?

Most market analysis measures distance: how far from the high, how far above the average, how deep the drawdown. This one measures time. It counts how long the market has occupied its current structural state, and places that count against every previous episode of the same state. The difference matters because the two dimensions frequently disagree, and the disagreement is informative. A market can be barely down and yet have spent an unusually long stretch in a weak configuration, which is a slow grind that wears holders out without ever producing a dramatic move. The reverse also happens: a violent fall that resolves in weeks. Looking only at price, these look nothing alike. Looking at duration, the first is the more advanced situation. The reference statistics come from closed episodes only. When the chart reports a median for a state, that median is built exclusively from episodes that had already finished before the day being displayed. A day never sees its own eventual duration, and never sees an episode that came after it.

How to read

Read the banner first. It names the state in force, gives the number of days elapsed, and places that number against the median and the third quartile of past episodes of that same state. Without it, a reference line at 152 days does not say the median of what. The main line is a sawtooth: it climbs one unit per day while the state holds, and drops back to the start the day the state changes. Each tooth is one episode, and its height is that episode's length. The reference lines are the median and the third quartile of past episodes of whichever state is in force. They step rather than curve, because they only change when a new episode closes and joins the historical record. When the current tooth is below the median line, this episode is shorter than most of its predecessors. When it climbs above the third quartile, it is longer than three quarters of them. The reference lines are absent early in the history and whenever a state has not yet accumulated enough closed episodes to support a statistic. That absence is deliberate: a median computed from a single prior episode would look like a measurement while carrying almost no information. An empty area means the honest answer is that there is not yet enough history. Overlaying the BTC price shows the disagreement directly. Long teeth with a flat price line are exactly the slow-wearing periods that price-based analysis tends to miss.

Key zones

There are no fixed thresholds here, because a meaningful duration depends entirely on which state is being measured. Some states are structurally long and some are structurally brief, and comparing a raw day count across states would be meaningless. What matters is the position of the current tooth relative to its own state's history. Three readings are worth naming: • Below the median: this episode is young by the standards of its own state. • Between the median and the third quartile: it has outlasted most past episodes, but not unusually so. • Above the third quartile: it is in the long tail of its own distribution, which is worth noting without treating as a countdown. The strongest and the weakest states have very different natural lengths. The constructive configurations tend to persist far longer than the stressed ones, which is consistent with an asset that rises over long horizons and falls over shorter ones.

What to observe

• A tooth that keeps climbing while the price line stays flat. This is the pattern that duration analysis exists to surface, and it is invisible on a price chart. • Whether reference lines are drifting upward or downward as episodes accumulate. A state whose episodes are getting longer over successive cycles is telling you something about how the market has changed. • Very short teeth in sequence, which mean the market is oscillating around a boundary rather than settling. The regime map shows the same thing from the other side. • The gap between the median and the third quartile. A wide gap means episode lengths vary enormously and any single reading deserves caution. • How many closed episodes support the statistic. With few prior episodes, the median is a weak summary and should be read as such.

Historical context

Across the full history, the constructive states account for the majority of elapsed time while the stressed ones are comparatively rare. The distributions are also asymmetric: for most states the mean sits well above the median, because a small number of very long episodes stretch the average. This is why the median and the third quartile are shown rather than the mean, which a single unusual episode can distort beyond usefulness. Episodes shorter than a week are excluded from the reference statistics. A configuration that lasts three days is a crossing artefact rather than a state the market genuinely occupied, and including such episodes would drag every median toward zero and make the comparison useless. The number of closed episodes per state is measured in dozens across the whole record, not hundreds. That is enough for a median to be indicative and nowhere near enough for a confidence interval to be meaningful.

Expert notes

The anti-look-ahead discipline is the whole point of this chart, and it is stricter than it may appear. Statistics for a given day use only episodes whose end date precedes that day. Building them from the complete record instead would produce a chart that appears remarkably prescient in backtest and delivers nothing in live use, because it would have been drawing on knowledge that did not exist at the time. The consequence is visible and intended: the reference lines are missing over the early years and appear progressively as history accumulates. A chart drawn without this discipline would show clean reference lines from the very first day, which would look better and mean less. The day counter itself resets on the confirmed state change, after the deadband has been applied. It therefore counts days in the state as the regime map displays it, not days in the raw unfiltered ordering. The two would differ around every graze of a boundary, and using the raw ordering would produce teeth that reset on noise.

Common mistakes to avoid

• Reading the median as a due date. It is the middle of a wide and skewed distribution. Episodes have run at a small fraction of the median and at several times it. • Assuming a long episode is closer to ending. Duration alone carries no such implication, and states have no memory of how long they have lasted. • Comparing day counts between different states. Their natural lengths differ substantially and the comparison is meaningless. • Reading a missing reference line as a data problem. It means the state has not yet accumulated enough closed episodes, which is a statement about the sample, not a fault. • Forgetting the sample size. A median drawn from a handful of episodes is a rough indication and should not be treated as a stable property of the market. • Expecting the counter to react to price. It reacts to the ordering of cost bases, and a large price move that does not change that ordering will not touch it.

Programmatic access

REST API

curl -sS \
  'https://api.trinityinsights.io/api/v1/cycle-intelligence/metrics/cycle-regime-duration-profile/data?timeframe=90d' \
  -H 'X-API-Key: $TRINITY_API_KEY'

MCP server

{
  "tool": "get_chart_value",
  "metric_id": "cycle-regime-duration-profile",
  "timeframe": "1y"
}

Required tier: pro. See the pricing grid for the tier list and the MCP documentation for multi-client configuration.

Related metrics

• cycle-cost-basis-regime-map
• cycle-cost-basis-regime-plane
• cycle-duration-dashboard
• cycle-drawdown-profile
• cycle-halving-block-progress

Institutional disclaimer

Trinity Insights is an educational and analytical tool. The metric above does not constitute investment advice. Trinity Insights is not a Crypto-Asset Service Provider (CASP) registered under MiCA Regulation (EU) 2023/1114. See the full disclaimer.