HISTORICAL HGB BENCHMARK / 09:00 ET EXIT
2023 — 2026911 nights.One complete timeline.
Explore the complete research history.
Follow the live model as new nights unfold.
Historical research benchmark · 832 nights adjusted with Alpaca prices; 79 retain published daily returns. The live journal below begins on 10 September 2026 and uses a different portfolio and price convention.
FORWARD JOURNAL / SINCE 10 SEP 2026
The model, from here.
Four portfolios. One live ranking. Compare the top 1, 3, 5 and 10 stocks, equally weighted each night.
The journal will appear here when data is available.
Waiting for the recorded sessions.
How these results are measured
Saved MASTER exp16_bookloss rankings from the live OKX-panel model, starting with the 10 September 2026 entry session. Top 1, 3, 5 and 10 use the original ranks, equal weights and 100% model exposure, reset each night. These are reference stock portfolios; they do not apply the HGB re-picker or volatility substitutions used by the historical top-1 benchmark.
Each stock uses the bot’s recorded regular-session close and next-session open. Returns include the existing model fee convention: 2.6 basis points on entry and 5 on exit. They exclude slippage and funding and are not Alpaca 15:59 → 09:00 execution returns or actual account profits.
Rankings must have been recorded on the entry day before 15:59 New York time. Missing prices or invalid rankings stay visible as pending or unverified. Cumulative returns stop at the first incomplete recorded session; a missing stock is never replaced with a lower rank. Dates on the chart are entry sessions, so a Friday entry settles in the following market session. Only recorded prediction sessions are covered.
Drawdown starts from the initial portfolio value and is measured at completed nightly exits. New results appear when the bot records its next-session prices. This page checks the journal every minute while visible; the last successful check and latest recorded session are shown separately.
01 / THE HISTORICAL BACKTEST
See the whole picture.
All 911 nights on one timeline. Percentage returns use a linear scale; drawdowns are measured after each nightly exit.
02 / THE RHYTHM OF RETURNS
The gains.
The setbacks.
A strong finish can still include difficult weeks. These are the actual counts across the full backtest.
139 positive / 51 negative
39 positive / 5 negative
03 / EVERY PERIOD, IN DETAIL
Every month. Every week.
* The final week and month are partial. Nights are assigned to their entry dates. Weekly labels are Friday-ending weeks.
04 / About this backtest
The exact original series
This report preserves the original +6,750,218.37% return and −24.7531% maximum drawdown. No new model or execution assumptions were added.
The benchmark combines published close-to-open returns with Alpaca 09:00/09:30 price ratios on 832 nights. The other 79 nights retain their published daily returns. Original costs and price conventions are unchanged. This is a historical benchmark, not a live fill record.
Reading return and drawdown
Every chart uses a linear percentage scale. Cumulative return compounds all previous nights. Monthly and weekly returns compound only that period.
The lower chart always shows account drawdown from the highest equity reached in the full history. Within-period drawdown in the period tiles resets the reference to each period’s starting equity. Both use equity at nightly exits and exclude fluctuations while a position is open.
Grouping uses entry dates, January 3, 2023–August 20, 2026. The final exit is August 21. A Friday entry exits in the following session.
Reconciled to the original total, drawdown and all 15 quarterly returns. The monthly and weekly series compound to the same total.