Premarket Momentum and Relative Volume

Methodology

Pre-registered: both volume definitions, the bins and the tests are fixed before any outcome is examined.

1. Two RVOL measures, both backward-looking. Production relative volume (RVOL) is today's volume divided by average daily volume, with no time-of-day adjustment. The replay rebuilds it as volume from 04:00 to the trigger minute over the average 04:00–20:00 volume of the previous 20 completed sessions. Because premarket volume accumulates through the morning, a trigger at 09:15 shows a higher production RVOL than an equally active one at 07:00, so the study adds a time-of-day RVOL: volume from 04:00 to the trigger minute over the average for the same clock window in those 20 sessions.
2. Triggers. Default PREMARKET_BREAKOUT rules on completed 1-minute bars only: previous-close reference, armed from +3%, a trigger when a bar's close reaches +5% after an earlier bar finished below it, skipped above +8%, first trigger per stock and day. The RVOL filter is off by default, so the trigger set does not depend on RVOL, and both measures are frozen at the trigger minute.
3. Bins. Quintiles of each measure, with cut points set on the first 70% of trading days (in-sample) and applied unchanged to the last 30% (out-of-sample).
4. Outcomes and costs. Entry at the next bar's open; costs as in Momentum Lab's backtest (0.2% spread, 0.05% slippage, no commission), with doubled costs as a sensitivity.
5. Confounders. RVOL tends to rise with news, with the size of the move and, in its production form, with time of day. Results are shown within trigger-time bands (04:00–07:00, 07:00–08:00, 08:00–09:00, 09:00–09:30) and in a logistic model that adds move size at the trigger, log market cap, share price and SPY's premarket change.
6. Tests. Confirmatory, run once out-of-sample and two-sided at the 5% level: the success-rate difference between the top and bottom quintiles of time-of-day RVOL, with a 95% interval from a bootstrap over trading days. Secondary and Holm-adjusted: the same contrast for production RVOL, a trend across quintiles, and the adjusted model. Time-band, market-cap and half-year results are exploratory. Success rates carry Wilson intervals, and net R is reported as a full distribution per quintile.
7. Live check. Live RVOL exists only where IBKR supplies the data. Aggregated, anonymised records from those accounts compare IBKR's value with the replay's for the same trigger and report outcomes by production-RVOL quintile, paper and live apart; no figure drawing on fewer than five accounts is published.
8. Reproducibility. Parameter version, RVOL formulas, quintile cut points, cost settings, data checksums and seeds are published with the results.

How success is defined

Primary outcome. Entry is at the open of the 1-minute bar after the trigger. A trigger succeeds when price reaches 3% above the fill before 3% below it, within 60 minutes of clock time; 3% is the default hard-stop distance, so the bracket is ±1R. A bar that spans both counts as a loss, and a trigger that reaches neither in time closes at the last price inside the window as a non-success.

Compared quantity. The success rate in the top quintile of time-of-day RVOL minus that in the bottom quintile. The same difference in mean net R is reported beside it.

Net R. (Exit price − fill) ÷ (3% of the fill), after the assumed spread and slippage. Losses beyond −1R, from gaps at the 09:30 open or after halts, are kept.

Secondary outcomes.

• Production RVOL: the same contrasts using the production definition, to show how much the time-of-day adjustment changes the answer.
• Follow-through to the open: the return from the fill to the 09:30 opening price, for triggers before 09:00.
• Strategy outcome: the default exit model's net return (0.5% trail on the last price, 3% hard stop, no take-profit, flat 2 minutes before the close), with the adverse extreme of each bar assumed to come first.
• Excursions: maximum favourable and adverse excursion within the 60 minutes, in R.

Pattern and strategy criteria

Momentum Lab's Premarket Momentum Breakout (PREMARKET_BREAKOUT) at default settings; every parameter is described in the Momentum Lab guide.

• Trigger logic: move measured against the previous close; LONG; armed from +3%; trigger at +5% with 1 confirmation tick, only after the price has been seen below +5%; skipped above +8% (chase limit of 3 points); re-armed below +4%; active in the premarket session, 04:00–09:30 New York time.
• Filters: price ≥ $1; premarket volume ≥ 50,000 shares; premarket dollar volume (premarket shares × price) ≥ $250,000; spread ≤ 1.0% of the mid; halt detection on; market cap of at least $500M under the default BALANCED preset. The market-cap filter cannot be switched off, and an unknown cap never passes.
• Relative volume in production: RVOL = today's volume ÷ average daily volume. It is not adjusted for time of day, it is available only with IBKR data, and the RVOL filter is off by default. Momentum Lab's in-app backtest does not replay RVOL, which is why this study rebuilds it from bars.
• AI filter: optional and off by default. It is a fixed weighted formula in which relative volume carries the largest weight (1.6), and it is not applied in this study.
• Default exits (secondary outcome only): 0.5% trail on the last price; 3% hard stop from the average fill; no take-profit; CONTINUE_REGULAR session policy; flat 2 minutes before the close.

The relative volume in the premarket guide explains why early-session RVOL readings need care.

Exclusions

• Short volume history: stocks with fewer than 20 completed sessions before the trigger, such as recent IPOs and new listings, and stocks whose 20-session window contains a split or other corporate action that changes the share count.
• No point-in-time market cap, or below $500M: funds and ETFs, foreign issuers filing 20-F or 40-F, secondary listings, and stocks without a 10-K or 10-Q within 200 days. The $100M floor applies in the sensitivity set.
• Default filters failed at the trigger minute: price below $1, premarket volume below 50,000 shares, or premarket dollar volume below $250,000.
• Halts: stock-days with a halt or volatility pause in the outcome window, identified in bars by a regular-session gap of 5 minutes or more, and in records by the engine's own halt detection where the data source supports it.
• Repeat triggers: only the first trigger per stock and day.
• Live records with incomparable volume: records whose market data came from a single-exchange feed, such as Alpaca's IEX feed (premarket only from 08:00), are left out of every volume measure.
• Small cells: no figure drawing on fewer than five accounts is published.

Limitations

• Attention, not cause. High relative volume usually reflects news, earnings or share offerings that also move price. The adjusted model controls only for what it measures, so an association, if one is found, will not show that volume itself causes follow-through.
• Denominator choice. The 20-session average is this study's choice. IBKR's average daily volume may use a different window, so the live comparison reports how far the two disagree.
• Volume conventions. Feeds report premarket prints, odd lots and late trades differently, and bar vendors do not all treat them alike.
• Live coverage is IBKR-only. RVOL exists only for accounts on IBKR data, so the live check covers a subset of users who may trade differently from others. Paper fills are kinder than real fills.
• Timing is approximate. The live engine checks quotes about once a second with confirmation ticks; the replay uses 1-minute closing prices and enters one bar later.
• Universe differences. The live engine sees only its movers lists (50 gainers and 50 losers per data source, refreshed every 15 seconds); the replay sees all qualifying stocks.
• Quintiles are relative. Cut points are fixed in-sample, but what counts as top-quintile RVOL in one period may be ordinary in another.
• Period dependence. Premarket participation can change over time, so results from the sample period may not carry over. Market-cap effects are examined separately in momentum breakouts by market capitalization, and the wider risks are covered in premarket liquidity and spreads.