Good catch. International rows feed the gauge too, which is part of why it moves while North America sleeps.

We promise the index does not add rooms when you are not looking. Mostly.

The Fear gauge is a volatility measure, not a mood measure. It rises when genre tickers move a lot relative to each other within a short span. Overnight we saw $SUPR and $PSYC swing in opposite directions within a few windows, mostly on You…

Roughly testable today with the mention breakdown, and making it cleaner is on our list for V1.0. No promises on the exact form yet.

Right now regional heat mostly reads titles and the conversation around them, so a strong shorts scene shows up as creator chatter more than as a title line. We know that undercounts places where shorts are the main format, and it is on the…

Confirming that from the data side. The analog tag now co-occurs with news sources in a way it did not before. Still Reddit and YouTube heavy, but the mix is broadening.

It does on wider screens. On phones that time gets squeezed out, which is not great. Logged.

Data note, since a few of you asked: the $FOLK heat score weights sentiment over raw volume, so a lot of 'I am tired of flower crowns' posts will drag heat even while mentions climb. The fatigue is real in the text, not an artifact of the m…

Good eyes, and nothing is broken. Box office figures arrive in batches, not continuously, so between batches our reader checks every 15 minutes, finds nothing new, and after a while the popup honestly marks that source stale. Stale there me…

That matches what we see. There is also a smaller mechanical piece: a big chunk of rows from a loud weekend aged out of the rolling window midweek. When a busy weekend leaves the window, HXI tends to sag even if nothing new happened. The 90…

Enjoy. And a reminder the footer is telling the truth: signals do not constitute financial advice, or peace of mind.

One nuance from the data: long form channels have less spiky heat but a very slow decay after each upload. On CRYPT that looks like a smooth line. Compare $MRBL style storytelling, which tends to spike harder per upload and settle faster.

That is a real limitation worth saying out loud. Chatter based signals favor things people argue about. Comedies get enjoyed and then forgotten in conversation terms, so a flat $COMH is not the same thing as a weak box office.

Correct, and a limit we should state more clearly on the Method page. Genre tickers pool film, TV, games and creator content unless the ticker is a title or creator. I will add a line about it.

Welcome. Here is how I would read the Index page, top to bottom: 1. HXI at the top is the overall horror index. Think of it as "how much is the horror audience talking, and how warmly". Up is more and warmer, down is quieter or cooler. The …

Reasonable. Logged it. Until then the Method page is the source of truth.

Good instinct. Heat measures how much people are talking, sentiment measures how they feel about it. A film can be loud and disliked. Legs usually need the second one.

Here is the plain reading. HOLD means the rule saw activity but no clear direction: mentions flat, sentiment flat, nothing crossing a threshold. The citations are the rows that contributed the most to that reading, not a curated argument. T…

Good questions. delta is the change versus the previous read, so it is a 15 minute change, not a daily one. If you want a daily number, take two exports a day apart and subtract. signal is the Oracle rule strength for that row, and it is bl…

Logged that. It is a good idea and not hard to do. No promise on timing before launch.

Not obvious at all, and good questions. Quick answers: 1. Heat blends how much people talk about you with how they feel about it, weighted toward recent activity. For a small channel, read it against your own history, not against the big ti…

Data side: this is exactly why CRYPT tracks sources separately. You can see when a heat bump is mostly trade and news versus mostly Reddit and YouTube. The Oracle weighs them differently for a reason.

This is a genuine limit. Horror praise often sounds like distress. We tuned the sentiment model on horror text for exactly this reason, but it still misreads some of it.

What we can say from backtests: divergence on its own is a weak predictor. It gets more interesting when it persists across more than one source. If Reddit is sour and YouTube is sour and news tone is flat, that is different from Reddit alo…

Correct and a known limit. Translation coverage is growing but non English rows are still under counted on the main index.

Worth saying clearly: the creator metrics do not track sales, and we are not going to guess at them. What creators tell us, anecdotally, is that merch-friendly audiences tend to have high sentiment and a strong in-joke vocabulary in the com…

Staff note, since "undervalued" invites it: CRYPT does not have a notion of fair value. The indices measure attention and tone, not worth. A low index can mean an underrated corner or simply a quiet one. This thread is still fun.

Fair. The ungraded calls are counted separately on the Oracle scorecard so they cannot quietly pad the hit rate. You can see how many there are.

Feature, though an uncomfortable one. We call it the zero exit. If the rows supporting a call age out of the window and nothing new replaces them, conviction decays toward zero. When it hits zero the call closes as ungraded. We decided that…

For what it is worth, we never meant conviction to replace judgment. It is a summary of the evidence the rules could see. Your gut sees things we cannot, like a lobby. That is exactly the kind of signal we do not have a feed for.