Two scores arrive. They have to agree with each other, clear the confidence floor you set, and fit inside your risk limits. Miss any one of those and the answer is no trade.
That is not the engine failing to decide. It is the engine deciding. A bot that always has an opinion is a bot that will happily act on noise, and you are the one who pays for that.
The three gates
Not two out of three. All three, every time, in this order. The first one it fails is the reason it stopped, and that reason gets recorded with everything else.
One says prices are turning, the other says the mood is turning. If they point opposite ways, that is not a weak signal to be averaged into something lukewarm. It is two independent reads contradicting each other, and the right response is to wait until they stop.
Agreement on its own is not sufficient, because two engines can both be mildly positive about nothing much. The combined score has to clear a floor you set. Set it low and it will act often. Set it high and it will act rarely and mean it.
Even a strong, agreed signal gets refused if taking it would break the exposure caps you configured, or push you past the drawdown line where you told it to stand down. The risk rules outrank the signal. Always, and without asking.
Who decides what
The two engines produce two numbers. How much each one counts is a question about what you are trying to do, not a question with a correct answer, so we do not answer it for you.
You set the balance between them, the confidence floor, which markets are in scope, and the conditions under which it does nothing at all. The defaults are cautious. The interface tells you plainly when a configuration is unusually aggressive rather than letting you find that out later.
One decision, with the workings attached. Which engine said what, what the weighted result was, which gates it passed, and if it stopped, which gate stopped it.
All of that goes on the chain together. Not just the trades it took but the ones it refused, which is the half nobody else publishes and the half that tells you whether the risk rules are doing anything.
market BTC/USDT
quant +0.62
sentiment +0.31
weights 60 / 40
fused +0.50
agree pass both positive
confidence pass 0.50 over your floor of 0.34
risk refuse exposure cap reached
decision no trade
recorded on-chain, with the reasonThat example is a good signal that did not become a trade, and it is written down exactly like one that did. If the refusals were left off the record, the record would flatter us.