Worst one-minute returns
Jump to ±12 hours around a loss. Equity and drawdown retain their full-history values.
BUSHIDO / RESEARCH TOOLS
SYNTHETIC DATAReturns lab.
A closer look at co-movement, sudden losses, and chart performance.
Preparing explorer…
LOADED ROWS—
DATA FETCH + DECODE—
FULL-DATA CALCULATION—
LAST CHART UPDATE—
VISIBLE / DRAWN VERTICES—
Explore the series
Drag a time region to zoom across all charts. Double-click to reset. Hover for values.
Measure your browser
Start with 100K rows, then choose 1M. Adaptive mode sends each bucket’s extrema to the chart and restores detail as you zoom. Raw mode sends every visible observation; 1M rows means up to 5M plotted vertices and can take considerably longer.
Correlation and worst losses always use all loaded observations. Timings are measured on this device; rendering includes display preparation and Plotly update, not GPU completion or a formal FPS benchmark.
Hosted synthetic data loads as binary chunks. This tests download, computation, and plotting—not Parquet decoding or DuckDB query speed. Matching Parquet files are available below.