CURRENT-MODEL RETROSPECTIVE / IN-SAMPLE HINDSIGHT. Simulated replay outcomes, not executed Kraken PnL or out-of-sample performance.
UTC calendar-day outcomes calculated from the same current-model selection as Equity. No historical live-authority fallback.
One resolved replay outcome per selected historical episode. Select a row to open Trade View.
Trade details come from the selected retrospective row, not the retired Classic Dashboard. Choose a row under Trades.
Optional Binance spot candles are independent market data, not proof of the original replay's TP/SL sequence.
Legacy-only paired comparison on identical resolved episodes. NOT an out-of-sample comparison.
Unresolved current-model outcomes are excluded. The 1,666 missing old live-authority decisions are a separate contract.
Retrospective in-sample bucket evidence, distinct from the live-execution strategy. Context families overlap; 14d/30d windows are not additive.
Research group counts cannot be interpreted as distinct trades.
Legacy reason-based recency evidence: preserved independently of the new contextual bucket definitions. Low n is flagged.
Historical fitted-policy selection, not realized exchange PnL. Only one selected policy per episode; family statistics still overlap.
Observed live five-timeframe votes and BTC pressure dynamics. These values are not necessarily conditions used by the selected research bucket.
Historical V2 and native live pressure are different data sources. Uncovered intervals are not filled with neutral states.
Recommendation audit only; NOT evidence of filled Kraken orders. The recorded legacy tier and contextual candidate remain separate.
Optional fixed historical paired-comparison report. This view is NOT recalculated from the daily release.