Evidence checklist
instrument decision for Ethiopia: compare the exact symbol by contract size, cash value per price move, trading hours, spread or commission, margin, holding cost and recent volatility before choosing exposure.
| Claim | Required evidence |
|---|---|
| Prediction | Defined target and horizon |
| Training | Dated, relevant data |
| Validation | Unseen chronological sample |
| Performance | Net of all costs |
| Execution | Exact symbol/account test |
| Control | Loss, exposure and shutdown limits |
Leakage test
Confirm that future prices, revised data and post-event information did not enter training features. Random shuffling can invalidate time-series evidence.
Market specificity
A model tested on EURUSD cannot be presumed valid on XAUUSD or BTCUSD. Contracts, sessions, costs and regimes differ.
Execution reality
Apply live spread, commission, holding, slippage, breaks and latency. Compare the model signal with actual fills on demo.
Risk controls
Set independent volume, daily loss, shared-exposure and kill-switch limits. The model must not be able to expand risk after a losing sequence.
Fraud screen
Reject guaranteed returns, unverifiable screenshots, secret broker deposits, cracked software and requests for credentials or remote access.
Frequently asked questions
Does AI prove an edge?
No. Confirm the exact symbol, contract, cost and current account condition before applying this answer in Ethiopia.
Why use chronological validation?
To avoid future information contaminating the test.
Can one model cover every market?
Not without separate evidence. Confirm the exact symbol, contract, cost and current account condition before applying this answer in Ethiopia.
Should controls be independent?
Yes. Confirm the exact symbol, contract, cost and current account condition before applying this answer in Ethiopia.