Paper — every trading figure here (positions, orders, P&L, equity, returns) is simulated on an Alpaca paper account. No real capital is at risk, and no order is ever placed with your money. Running costs shown under ops, such as LLM spend, are real.

Going deeper

Public

Thesis Management

SentinelHub forms its own view of the market from the data — and when an outside opinion arrives, it holds that opinion up against its own evidence rather than simply going along with it.

SentinelHub has a view of its own

Before it weighs anyone else's opinion, the system builds its own read of the market from the numbers — the trend, how calm or stormy conditions are, where the business cycle sits, and how much room a move has left. That read becomes a written thesis: a clear stance (bullish, bearish, or neutral), a level of conviction, the evidence behind it, and — crucially — a plain condition that would prove it wrong.

That last part matters most. A view you can never be wrong about isn't a view; it's a horoscope. Every thesis SentinelHub holds comes with a line in the sand — if the market crosses it, the thesis is marked wrong and retired. That's what stops a wrong view from quietly living on.

The Adversarial Gate

Plenty of outside opinions land on any given day — a market newsletter, an analyst note, a confident take on where things are headed. SentinelHub treats each of these as information to test, not an instruction to follow. It lines the outside view up against its own evidence and asks a simple question: does this agree with what our own data actually shows?

That deliberate skepticism is the Adversarial Gate. An outside opinion only shifts the system's conviction if it brings evidence the system can check and confirm. A confident claim with nothing verifiable behind it is noted and set aside, not acted on.

Why disagreement is the valuable part

When the outside view and SentinelHub's own evidence disagree, the system does not split the difference. It flags the disagreement clearly and keeps its evidence-based view until the evidence itself changes. Knowing exactly where you and the crowd part ways — and who turned out right — is how you learn which voices are worth listening to.

The three verdicts, in plain English

  • ·Corroborated — the outside view agrees with what our own evidence shows. Two independent reads pointing the same way is a genuine strengthening of the case.
  • ·Divergent — the outside view disagrees with our evidence. We hold our evidence-based view and flag the split clearly, rather than being talked out of what the data shows.
  • ·Qualified agreement — same direction, different conviction. We agree on which way things are leaning, but our evidence says the enthusiasm is overdone. The plain- English read: constructive, but don't chase.

A worked example

A market newsletter arrived urging caution on semiconductors while staying upbeat on the big technology names, advising readers to keep riding them. SentinelHub held both claims up to its own evidence:

  • ·On semiconductors— the system's own read already leaned cautious. Outside view and our evidence agreed: corroborated.
  • ·On the big technology names — the newsletter said keep riding them. Our evidence showed them stretched and running out of room. Same broad direction, very different conviction, so the verdict was qualified agreement: extended, don't chase.

The value wasn't in agreeing where it was easy. It was in the "don't chase" on the crowded trade — the place where quietly following the outside opinion would have cost the most.

The doctrine

This is decision support, not a trader. Thesis Management informs your thinking and leaves a clear record of what the system believes and where it parts ways with the crowd — but it never places, sizes, or closes a trade on its own. The judgement stays with you.

To see how the system arrives at its own view in the first place, read Market Weather and How Opportunities Are Vetted.