Trust Score methodology
All current scores are Preliminary. They rest only on public facts we can cite, linked as "Evidence" on each card. We never score past performance or vendor return claims.
| Criterion | Points | What earns it |
|---|---|---|
| Transparent logic | 30 | Open-source code, published rules, or user-defined strategies you can inspect |
| Test before live | 20 | Paper trading or backtesting available before you risk money |
| Public pricing | 15 | Price shown publicly before signup |
| Your own regulated broker | 15 | Executes in a brokerage account you control (no deposits held by the vendor) |
| Reproduced by others | 20 | An independent third party has publicly reproduced its results. Earns the ✓ badge. No tool qualifies yet. |
Red flags (planned)
Performance claims without rules, live and paper results mixed without disclosure, costs left out, guaranteed returns, or deposits sent to the vendor. Flagged tools will be hidden by default.
Out-of-sample and costs (planned)
Future versions will reward published out-of-sample or holdout tests, and results that include fees and slippage.
Broker automation levels
On Brokers, each broker gets the level its own official docs support. We never upgrade a broker without a source.
- Full autonomy (no human intervention): the broker's official AI-agent/MCP path is documented to allow agent order placement with no per-order human approval. Having a regular trading API is not enough for this level.
- Programmable API (build your own bot): an official API can run unattended code, but unattended AI-agent trading isn't officially confirmed.
- API automated, AI connector needs approval: the API runs unattended, but the official AI connector needs per-order approval. In our testing, IBKR's AI connector required human approval on every order.
- Per-order approval required: every path needs human approval for each order.
- Manual only / no API: no official trading API.
Machine checks such as order previews or dry-run tokens don't count as human approval. Periodic re-logins (for example, E*TRADE's daily token) are noted but don't change the level. Caveats such as Robinhood's occasional in-app approvals or pre-trade alerts are shown in the notes.
Sponsorship
Sponsored placements are labeled and never change a score.