Narratives

• Gaining strength — the categories where participation is rising on each of the three moving averages. A category only qualifies if the share of its coins trending is higher than yesterday and higher than a week ago, so a single green day does not put it there. The number is the gain in percentage points against a week ago.
• RS heatmaps (3M / 1M / 1W) — colour is the category's median RS percentile, box size is its share of the day's turnover. Reading the three left to right shows whether a category is arriving or leaving. Click any box to filter every block below it.
• RS Rating Comparison — one row per category with a bar per window. The SHOW pills add or remove windows; 6M greys out on past dates, and 3M on the oldest ones.
• Advances — the share of a category's coins that are up over the month, week and day. High RS with low advances means a couple of leaders are carrying the bucket.
• Top categories by volume — where unusual volume is showing up: pocket pivots over the last week, pocket pivots today, and average relative volume (today's turnover against each coin's own 20-day average).
• Key moving averages — participation, each row showing today, yesterday and a week ago, the direction mattering more than any single reading. The shortlists this feeds sit at the top of the page.

The date picker rebuilds the whole page as of any session the history sheet covers — currently about four months — straight from raw daily candles. The Today / 1W / 1M buttons are quick jumps to three of those sessions; the dropdown and the arrows beside them reach every one. Use it to check whether a rotation you are looking at started today or has been building for two weeks.

On a past date every category listed grows an outcome line: what that category went on to do over the next 7 and 30 sessions, minus the whole universe's median over the same stretch. A rotation signal claims out-performance rather than a rise, so flat in a falling market counts as a win. The date lives in the URL, so a handful of cherry-picked days can be opened side by side in tabs, and Download carries both outcome columns for tabulating them.

Show sub-categories drills the 14 buckets into 34: DeFi splits into DEX, Lending, Perps, Liquid Staking, Launchpads, Prediction Markets and Payments; Gaming/NFT into Gaming, NFT and Metaverse; Infrastructure into Oracle, Bridge & Interop, ZK & Proving, Identity & Naming, Data & Indexing, Wallets and Rollup Infra. Useful once a broad bucket is moving and you want to know which part of it.

Click any category — a heatmap box, a table row, a bar — to filter every block to it and list its coins with their individual ranks and changes. Click it again to clear.

To act on what you find, use the Category filter in the screener's MISC panel: it works alongside every other filter, so you can ask for pocket pivots inside the one narrative that is actually being bought. Download exports the table as it currently stands.

Each coin sits in exactly one bucket, so a category's numbers mean something. That is harder than it sounds: most coins carry half a dozen tags, and nearly every project claims to be an AI project.

The assignment starts from CoinGecko category membership, weighted by how central a coin is inside each category, and then every judgment call is frozen by hand. Buckets that move together are merged — gaming, NFT and metaverse are one bucket; DEXes, lending, perps and launchpads are all DeFi — because splitting them adds noise, not signal.

Tokenized stocks, ETFs and metals (Nvidia, Samsung, gold) trade on the same venues but are not a crypto narrative, so they sit in their own bucket and are hidden by default. The checkbox at the top brings them in.

• The date picker reaches back as far as the history sheet goes, minus a 60-session warm-up — the moving-average block needs that run-up before a 50-day average settles. Dates before it are not offered rather than served inaccurate.
• 6M RS works on the latest date only, and 3M drops off the oldest dates. 6M needs a year of history, which only today's snapshot carries; 3M is recomputed from 90 sessions of candles wherever that much run-up exists, and on the oldest dates it is simply blank.
• The outcome line is measured on survivors. A coin delisted inside the next 30 sessions has no price to compare against, so it leaves the median rather than scoring −100%. Read the outcomes as slightly generous.
• The outcome line needs the future to have happened: on dates inside the last 30 sessions the 30-day figure is blank, and inside the last 7 both are.
• The moving-average block recomputes its EMAs from 90 days so all three readings share one method. The screener's own Rising-50EMA column uses a year, so a category's rising-50 share can differ here by a point or two.
• Heatmap boxes are sized by turnover share, not market cap — the screener carries no market-cap column.
• A category with fewer than 5 coins is marked thin and never ranks — a two-coin median is not a narrative.
• Momentum is a relative measure. In a market-wide selloff a category can gain rank while still falling in price; check the advances columns before reading it as strength.
• Rotation is context, not a signal. It tells you where to hunt — the setup still has to be there on the screener.