Back to HFT article

Interactive appendix

Polymarket HFT feature explorer

A real recorded L1 slug viewer wrapped into the site: feature overlays, smoothing controls, window selection, and the exact kind of inspection that sat behind the HFT dataset work.

slug=sol_177181020058,613 rows57 columns
Plotly
Recorded L1
Feature debugging
Prepared datasets
HFT
Polymarket HFT feature explorer cover

Interactive appendix

A real HTML export from the HFT repo, now hosted inside the portfolio so the recorded L1 feature work can be inspected directly.

Why this explorer matters

This is the concrete layer between 'I built HFT features' and 'you can actually inspect what that means on real recorded data.'

The HFT work was not just about training LightGBM on a prepared table. A big part of the job was checking whether individual features were stable, whether smoothing helped or hurt, how Binance and Polymarket moved together, and how much of the signal survived once timestamps were treated honestly.

This viewer comes from the real repo export, not from a portfolio mock. It makes the dataset story more credible because you can inspect the feature surface itself instead of only reading the summary around it.

Explore the recorded L1 slug

Use the controls to compare raw series, smoothing choices, and aligned HFT features on one real slug from the recorder pipeline.

L1 recorded slug
Series overlays
Hampel / EWMA / rolling median
Feature inspection

This explorer is intentionally embedded as part of the website rather than linked as an external file, so the feature-debugging layer becomes part of the portfolio narrative.

Next move

Want the narrative behind the chart?

The HFT article explains why recorder quality, local receipt timestamps, prepared_l1 and prepared_l2, and execution realism mattered more than raw classifier glamour.