Strategy settings
Select relevant methods. Preferences are saved on this device; scoring is deterministic, not a win probability.
Analytical methods
Documented confluence model
Each independent group earns its fixed weight only when directional, directly observed evidence is present and uncertainty is empty. Similar signals within one group never stack; shared correlation IDs also prevent the same event earning points across groups. The original image is re-examined before scores can support orders.
Conflicts deduct 10 points per opposing strategy group; uncertain selected groups deduct 5 each; chart-level uncertainty deducts 5. The total penalty is capped at 30. Opposing structure or entry confirmation blocks actionable trades even if the score remains high.
Score interpretation & essential gates
0–49
NO TRADE
50–59
WEAK · Watchlist
60–69
MODERATE · Missing confirmation
70–84
STRONGER · Subject to checks
85–100
HIGH CONFLUENCE · Subject to checks
All proposed trades require readable identity, timeframe, price scale and exact source labels; directional structure and a technical zone; valid stop/target geometry and minimum reward/risk. NOW additionally needs observed confirmation and no missing conditions. Pending orders require an explicit activation condition. Scores never override a failed gate. No historical win rate is claimed because no documented backtest dataset is connected.
Staged reliability workflow
- Read the exact screenshot and preserve extracted labels, candle evidence and uncertainty.
- Re-examine the same original image in a separate AI request. Reject unsupported evidence, levels and contradictory proposals.
- Apply deterministic geometry, risk/reward, independent confluence and conflict penalties.
- Review original/annotated images, then optionally save public reports and record hypothetical outcomes. These do not train the AI.
Real vision accuracy tests require user-provided screenshots. Test bullish, bearish, ranging, blurry, incomplete and misleading labels through Upload and inspect each image-review verdict. Unit fixtures and injected request failures are not screenshot backtests.
Safety validation tests
Synthetic unit fixtures test the deterministic engine only—not live vision or screenshot accuracy.
Bullish: valid geometry can pass
Risk points reflect passed chart-derived checks
Bearish: symmetric risk geometry
Ranging: no directional evidence means no trade
Blurry: insufficient chart data
Incomplete: missing scale blocks entry
Incorrect labels: unknown identity blocks entry
Unsupported price precision is rejected
Unconfirmed candle blocks BUY NOW
Duplicate signals do not add score
Malformed AI response fails closed
Unsupported candle OHLC is suppressed
Invalid pending-entry geometry is blocked
Missing evidence references cannot pass
Contradictory stated direction blocks entry
Opposing observed structure blocks a high-scoring trade
Cross-group correlated evidence earns points only once
Failed image recheck prevents an actionable setup
Image review must verify every precise trade level
Target ordering and computed R/R are enforced
Missing second timeframe cannot earn confluence
Malformed separate image review fails closed
Invalid risk preference is rejected
Position sizing requires complete contract details
Position sizing rounds down to allowed increments
Insufficient hypothetical outcomes do not imply a win rate
A live end-to-end test requires real uploaded bullish, bearish, ranging, blurry, incomplete and mislabelled screenshots. Failed AI requests leave no generated analysis and display the actual error.