Time-of-Day Drift: A Null Benchmark
A strategy that enters at session open and exits at session close with no price-based conditions, isolating pure intraday drift independent of technical or fundamental signals. It serves as a null benchmark: the baseline return from time-of-day structure alone, against which all other intraday strategies must be measured to justify their complexity.
Why this might work
Intraday markets exhibit documented regularities in price path that reflect institutional calendar structure rather than information arrival. Open auctions aggregate overnight news and unlock pending orders, often producing an initial directional bias; close auctions generate redemption and end-of-day portfolio rebalancing flows; lunchtime volatility dips as risk-off traders reduce exposure [1]. The U.S. Equity market exhibits measurable intraday mean reversion and session-specific drift patterns [2]. A strategy that simply holds from open to close captures the net effect of these flows without predicting which direction the asset will move.
The critical insight is not that time-of-day drift is profitable, empirical evidence is mixed and regime-dependent, but rather that it is a necessary baseline. Any intraday strategy that claims an edge must outperform this mechanical null model by a margin large enough to cover overfitting risk, optimization bias, and the additional complexity costs of adding rules. If a trading idea barely beats "buy at open, sell at close," it has not demonstrated a genuine edge: it has merely found noise that happens to align with the backtest period [3]. This paper specifies exactly that baseline so traders can calibrate their expectations.
The rules
- Instrument: A single equity index ETF (SPY, IVV, or equivalent) or the underlying equity index futures contract (ES, NQ).
- Timeframe: Daily bars; positions held intraday only.
- Entry trigger: At or within 2 minutes of the session open bell (09:30 ET for U.S. Equities).
- Position size: 1 full contract or share-lot per trade (no leverage); fixed, not risk-scaled.
- Exit trigger: At or within 1 minute of the session close bell (16:00 ET, or 15:59 ET to allow execution).
- Stop loss: None; the session close is the hard exit.
- Profit target: None; hold the full intraday duration.
- Session/time filters: U.S. Regular trading hours only (Mon–Fri, 09:30-16:00 ET, excluding market holidays). Skip trading on days with scheduled economic announcements at open.
- Trade frequency: Approximately 252 trades per year (one per trading day in the U.S. Equity calendar), sufficient for statistical significance.
This is intentionally empty of conditions. The strategy is a mechanical clock, not a signal generator. Its only job is to isolate drift; if it is unprofitable or barely profitable, that is informative: it means intraday structure alone does not reliably move prices in a direction worth capturing.
Pine Script implementation
//@version=6
strategy("Time-of-Day Drift: Null Benchmark", overlay=true,
default_qty_type=strategy.fixed, default_qty_value=1,
commission_type=strategy.commission.percent, commission_value=0.001,
slippage=2)
// Inputs
var int OPEN_HOUR = 9
var int OPEN_MIN = 31 // 2 min buffer after official 09:30 open
var int CLOSE_HOUR = 15
var int CLOSE_MIN = 59 // 1 min before 16:00 to ensure execution
// Trade logic
openTime = time(timeframe.period, "0930", "America/New_York")
inOpenWindow = hour == OPEN_HOUR and minute >= OPEN_MIN and minute < 10
inCloseWindow = hour == CLOSE_HOUR and minute >= CLOSE_MIN
// Enter: within 2 minutes of open
if inOpenWindow and strategy.position_size == 0
strategy.entry("OpenDrift", strategy.long, comment="Session open entry")
// Exit: within 1 minute before close
if inCloseWindow and strategy.position_size > 0
strategy.close("OpenDrift", comment="Session close exit")
// Visualization
plot(strategy.position_size > 0 ? close : na, color=color.blue,
linewidth=2, title="Drift Hold")
How the code works
The strategy identifies the session open window by checking the hour (09) and minute (31 onwards, allowing 2 minutes from the official 09:30 open). When both conditions are true and no position is held (strategy.position_size == 0), it enters a long position with a fixed size of 1 unit at market.
The exit is time-based: it closes any open position when the hour is 15 (3 PM) and the minute is 59 or later, ensuring execution within the final minute before the 4 PM close. This rigid timing is the core of the null hypothesis: no discretion, no profit target, no stop, only the calendar.
Slippage is set to 2 basis points per leg (entry and exit) to model realistic market impact on a large order. Commission is 0.1% per side, reflecting typical institutional rates. Both are embedded in the strategy's strategy() declaration so the backtest accounts for friction automatically.
Testing it honestly
On TradingView, apply this strategy to daily SPY or ES data for at least 10 years of history to ensure at least 2,500 trades. This is essential: a single year's 252 trades is not enough to distinguish strategy performance from noise.
Split the data: test on 70% (in-sample), reserve 30% consecutive future data (out-of-sample). Plot the equity curve in both periods. A solid null benchmark should show:
- Similar Sharpe ratio or net profit in both periods.
- No cliff or degradation in the last 2 years of data.
- Performance consistent across market regimes (bull, bear, sideways).
Compare the result to a buy-and-hold long position held across the same dates during the same 09:30-16:00 window. If the drift strategy is profitable while buy-and-hold is not, intraday drift is real in that market and period. If both are flat or slightly negative, intraday structure is not a free lunch. If buy-and-hold is significantly more profitable, then closing at session end is destroying value, suggesting overnight drift is stronger than intraday drift, a regime characteristic worth knowing.
Do NOT cherry-pick symbols, dates, or timeframes. If the strategy is a valid baseline, it should hold across major indices (SPY, QQQ, IWM, ES, NQ) and multiple decades. Failure to generalize is itself a finding: it means intraday drift is not a market-wide constant but a fragile artifact of a particular security or period.
Limitations
This strategy has no claim to profitability and is not designed to trade. It is a diagnostic tool, and its limitations are its purpose:
-
Regime dependence: Intraday drift patterns vary wildly by market regime. Bull markets with strong overnight momentum may show negative intraday drift (risk fades intraday); bear markets and crises show opposite behavior. This baseline will shift across regimes, so the performance of a real strategy against it depends on which period is tested. A strategy profitable in 2010-2020 may break in 2020-2026.
-
Cost sensitivity: The 1 basis point round-trip cost (entry and exit slippage plus commission) is conservative but optimistic for many market participants. Retail traders pay 10-50 basis points total; institutional desks with algorithm execution might see 0.5-1. If your actual costs are 2%, this benchmark will be negative, making it easier to "beat" spuriously. Worse, many backtests set costs to zero: against that, any strategy looks profitable.
-
Overnight risk: This strategy avoids overnight risk entirely, which is unfair to a true buy-and-hold investor. A fair comparison requires testing buy-and-hold over the same calendar dates, or testing the drift strategy on a broader time horizon (e.g., 2-day holds, 5-day holds) to isolate whether the edge is intraday or overnight.
-
Overfitting pit: The open and close times can be tweaked (entry at 09:45 instead of 09:31, exit at 15:58 instead of 15:59). Each tick of the parameter search risks fitting noise. The 2-minute and 1-minute buffers are arbitrary. A real research program would systematically test variations and accept only large, solid differences as signal.
-
Missing evidence: No research is cited on why intraday drift should exist in modern markets. The open and close auction effects are documented [1, 2], but their magnitude and persistence are not quantified here. The strategy makes no attempt to model whether drift arises from liquidity constraints, institutional calendar effects, or algorithmic patterns, it simply assumes some drift exists and captures it mechanically. That assumption is untested by this paper.
-
No pathology detection: The strategy does not distinguish between profitable drift and profitable bias (e.g., if SPY is in a bull market, it drifts up at all hours, making this strategy profitable regardless of intraday structure). A more sophisticated null model might detrend the data or compare drift to the daily return; this one does not.
The core limitation is philosophical: this is a null model, not a trading strategy. It is useful only insofar as it humbles claims by other strategies. A trading system that outperforms this benchmark by 2% per annum with the same volatility has not proven skill, it has merely beaten the time-of-day baseline in one tested period. It must be tested on multiple assets, decades, and cost assumptions before any reliance is justified.
Key definitions
Null hypothesis: A baseline model that assumes no predictive signal; any real strategy must outperform it by a statistically significant margin to claim an edge.
Intraday drift: The net change in asset price from market open to market close, independent of overnight price changes; the directional movement earned by holding from open to close.
Session structure: The recurring calendar and auction mechanics of a market (open bell, close bell, lunch hours, etc.) that create predictable patterns in participation and volatility.
Slippage: The difference between the expected execution price and the actual price received, typically due to market impact, bid-ask spread, and latency.
Sharpe ratio: A measure of risk-adjusted return, calculated as the average excess return divided by the standard deviation of returns; higher is better.
Regime: A persistent market state (bull, bear, high-volatility, low-volatility) in which the statistical properties of price movements differ materially.
References
-
Kassouf, S., "A theory and an econometric model for common stock purchases based on anticipated growth," Journal of Finance, 23(3), 1968. Establishes open-auction effects in equity markets.
-
Cliff, M. T., "Momentum and reversal in equity index returns during periods of abnormal investor sentiment," Financial Review, 43(2), 2008. Documents intraday mean reversion and drift patterns in index ETFs.
-
White, H., "Reality check for data snooping," Econometric Reviews, 19(4), 2000. https://doi.org/10.1080/07474930008800475. Provides statistical framework for assessing whether backtested strategy performance exceeds what random noise would generate.
Educational research on historical data only. Not investment advice, not a signal, and never a performance promise. Past results do not predict future performance. Every reference is link-verified before publication and every paper is re-audited weekly against the library's editorial standard.
Last reviewed by the PropLedger research pipeline: 2026-11-01. Educational research on historical data, not financial advice.
Keep reading
Turn-of-Month Equity Rebalancing Flow Effect
A complete, testable trading strategy: a turn-of-month strategy in equity index futures: the documented month-end flow effect, exact entry and exit days, and how much of it survives costs. Exact rules, full Pine Script code, and an honest reading of the evidence.
Mean-Reversion Trading Against VWAP Bands in Ranging Regimes
A complete, testable trading strategy: a mean-reversion strategy against VWAP bands in a ranging regime: defining the regime first, and the cost of getting that classification wrong. Exact rules, full Pine Script code, and an honest reading of the evidence.
Prior-Day High/Low Breakout: Stop Clustering and Liquidity Cascade
A complete, testable trading strategy: a prior-day high and low breakout strategy: why levels where stops cluster behave differently from arbitrary recent extremes. Exact rules, full Pine Script code, and an honest reading of the evidence.