pair: 4x4 and 2x2

order block: 2x4

1x2/4x4, sell at z score 2 tp at z score -2. buy at -2 sell at +1.5

Modify and implement the strategy in round5_v1_claude.py as a candle-based order block strategy trading ONLY PANEL_2X4. Build OHLC candles from mid price using candle_tick_window values [10, 100, 1000, 10000, 100000, 500000]. ATR window means the number of aggregated candles used to compute ATR, not raw ticks; sweep atr_window over [5, 10, 20]. Detect bullish order blocks as the last bearish candle before a confirmed bullish break of structure / impulse move, and bearish order blocks as the last bullish candle before a confirmed bearish break of structure / impulse move. Implement both long and short setups: breakout entry after confirmation, and retest entry where price returns to the previous opposite order block boundary. For long retest, buy when price hits the previous bullish order block upper bound after a bearish pullback order block forms, with stop at the pullback bearish order block lower bound minus 2 ticks or the previous bullish order block lower bound minus 2 ticks depending on stop_mode; short setup should be symmetric. Avoid look-ahead bias: an order block can only be traded after confirmation occurs. Sweep parameters including candle_tick_window, atr_window, swing_lookback, confirmation_window, impulse_atr_mult, stop_buffer_ticks, take_profit_R, entry_mode, stop_mode, and order_size. Save parameter results to a CSV with pnl, trade count, win rate, average trade pnl, long/short pnl, breakout/retest pnl, max drawdown, and final position. After implementing, run the rust backtester with: rust_backtester --trader "/Users/geraldchai/Documents/01-Study(notes:pdf:courseworks)/NYU/imc prosperity/imc4_codex_round5/round5_v1_claude.py" --dataset round5 --persist. Then summarize the top 10 parameter sets and diagnose whether breakout or retest order block logic works better on PANEL_2X4.