Which way is it moving?
Asked separately on every timeframe rather than mashed into one number. A market can be climbing on the hour and falling on the day. That disagreement is information, and averaging it away throws the information out.
Engine 1 of 2This one reads prices. The other reads people.
Five indicators, one verdict
It runs the indicators you already know. RSI, MACD, moving average crosses, volatility, volume. Nothing exotic, nothing secret, and nothing you could not check yourself.
One of them firing means nothing on its own. It only acts when enough of them say the same thing at once, and when they disagree it does nothing at all. Doing nothing is a real answer here, not a failure to reach one.
- Indicators
- 5
- RSI, MACD, EMA, ATR, volume
- Output
- 3 fields
- market, direction, confidence
- Runs on
- Shared GPUs
- not a server we own
- Included from
- Starter
- on every plan
The four questions
Every one of them asks one of four questions
Take the names off and each indicator is asking one of these four things about the market. Every question gets asked on every timeframe, the same way, without anyone getting bored on the four hundredth pass.
How rough is it right now?
The same move means one thing in a quiet market and something else in a violent one. This is what decides how big a position should be, which matters more than most people give it credit for.
Is anyone actually behind this?
Price with no volume behind it is a rumour. Volume tells you whether real money moved, and where it piled up tells you which levels are likely to matter next time.
Which way would it tip?
How much is waiting to buy, how much is waiting to sell, and how close to the current price. Thin on one side means a modest order moves things a long way, which is worth knowing before you become that order.
Why there are two
Because they break differently
There are two engines. This one reads prices. The other reads what people are writing about those prices. We could have merged them into a single model and there is a reason we did not.
Price signals go wrong slowly and quietly, when a market changes character and the numbers take a while to catch up. Text signals go wrong fast and loudly, usually because somebody paid for a few thousand posts.
Merge them and a bad call is just a bad call with no explanation attached. Keep them apart and every decision carries a note of what each side said, and that note goes on the chain with it. When something goes wrong, you can see which half got it wrong.
What actually comes out
Three things. Which market, which direction, and a number between zero and one for how sure it is. No price target, no write-up, no recommendation.
Keeping it that small is the point. A number can be weighed against the other engine's number, checked later against what actually happened, and written to a ledger anyone can read back. An opinion can do none of those things.
market BTC/USDT
direction long
confidence 0.71
confluence 3 of 4 directional indicators
recorded on-chain, the moment it was madeFour, not five, because volatility does not get a vote. It tells you how rough the water is, not which way the current runs, so it changes how big a position should be rather than which direction to take. An indicator that cannot disagree with the others should not be allowed to agree with them either.