The rules
Zarattini, Aziz and Barbon (2024, revised 2025), the 1x current-band-plus-VWAP version, where VWAP is the session's volume-weighted average price. For each half-hour mark, sigma is the average over the previous 14 sessions of the absolute percentage move from that session's 09:30 open to that mark. The upper band is the higher of today's 09:30 open and yesterday's 16:00 close, times (1 + sigma); the lower band is the lower of the two, times (1 - sigma). Decisions are taken only at HH:00 and HH:30, from 10:00 to 15:30: go long above the upper band, sell short below the lower band, and stay flat inside. A long is closed at a half-hour when price is below the higher of the upper band and VWAP, and a short when price is above the lower of the lower band and VWAP; crossing the opposite band reverses the position, and later re-entry is allowed. Everything is closed at 16:00. Size: shares = the prior day's equity / today's 09:30 open. The headline version instead targets 2% daily volatility with up to 4x leverage.
- Universe
- SPY
- Bars
- 1-minute
- Position size
- 97% of equity in SPY, long or short. Published: 100% of the prior day's equity at the 09:30 open, fixed for the day; the headline version targets 2% daily volatility with up to 4x leverage.
- Rebalance
- Checks at each half-hour from 10:00 to 15:30 (to 12:30 on early-close days), flat at the close. Published: checks from 10:00 to 15:30, flat at the 16:00 close, about 1.8 trades a session.
What the research found
Zarattini, Aziz and Barbon (May 2024, last revised September 2025) tested SPY on one-minute bars from May 2007 to April 2024: at 1x the version exiting on VWAP returned 380% with a Sharpe ratio (return per unit of risk) of 1.24, against 178% (Sharpe 0.61) for the base version, exiting at the opposite band, and SPY's 227% (Sharpe 0.45). The 1x version's average trade made $0.09 a share, about 3.6 basis points (0.036%) at the authors' $247 average SPY price (derived), against the library's 0.10% round trip. The headline version, levered up to 4x, returned 1,985% (Sharpe 1.33); the authors' updates give it +32.2% in 2024 and +1.0% from January to August 2025, and the 1x version 393% to December 2024. The effect behind the rule, the market's earlier intraday move predicting its later one, is documented in the Journal of Financial Economics by Gao, Han, Li and Zhou (2018) and Baltussen, Da, Lammers and Martens (2021); Rosa (Journal of Futures Markets 2022), cited in the paper, finds the overnight return's prediction of the last half-hour gone in later data, though it depends on the signal's strength, and trading only past a threshold earned more than always trading. Brusco's replication (GitHub; SPY from July 2020 to July 2026 and S&P 500 E-mini futures from May 2024; the levered version, at $0.0035 a share commission and no slippage, that is, every fill at the quoted price) found a Sharpe of 1.11 on SPY, against 0.93 for holding it, and -0.07 on the futures at about 0.4 basis points a round trip. On SPY its yearly Sharpe ratios were 1.4 to 2.0 over 2020-2024 and negative from 2025 (-0.27 in 2025 and -1.91 in 2026 to 8 July). Over 2024-26 the base exit beat the VWAP exit on both (Sharpe about 0.7 against about 0), a finding he calls after the fact and statistically inconclusive; over his whole SPY sample the VWAP exit did better (1.11 against 0.86). Paz Serpa (SSRN 2026) matched the paper over 2015-2024 (Sharpe 1.34) and found the levered version's Sharpe fell to 0.39 from May 2024 to March 2026, a statistically significant decline, with a return of 9.4% against 29.5% for holding SPY. VigorTas (GitHub; after costs) found Nasdaq-100 futures profitable over 2010-2026 (t-statistic 2.74; about 2 or more is read as significant), almost all of it after 2018, and S&P 500 futures net negative over 2013-2026 (t -0.20); from publication in June 2024, Nasdaq-100 futures gained and S&P 500 futures lost, neither significantly (t 1.37 and -1.21).
Results
| Measure | Orbital test | SPY, same days |
|---|---|---|
| Window | Oct 30, 2023 to Oct 30, 2026 | Oct 30, 2023 to Oct 30, 2026 |
| Total return | −46.27% | +88.60% |
| CAGR | −18.70% | +23.54% |
| Sharpe | −4.38 | 1.22 |
| Discounted Sharpe | −8.42 | — |
| Max drawdown | −46.7% | −18.7% |
| Win rate | 24.2% | — |
| Profit factor | 1.09 | — |
| Trades | 687 | — |
Hypothetical: a backtest with modeled costs, not traded with real money.
Hypothetical growth of $10,000
R09 Noise-Area Momentum · −46.3% vs SPY +88.6%
Rebuild it in Orbital
Paste this into Orbital Chat. The Strategy Developer builds the workflow and asks before it runs any backtest.
Build a program on 1-minute bars for SPY, using the session clock. Each session, keep the 09:30 bar's open, the prior session's close (the open of its 16:00 closing-print minute, or on early-close days the close of its 13:00 minute) and a VWAP from 09:30: the running sum of (high + low + close) / 3 times volume, divided by the running sum of volume, updated with each bar as it closes, before deciding on that bar. Trade only on the bars that close on each half-hour from 10:00 to 15:30 (to 12:30 on early-close days), apart from the reversal entries and the end-of-day exit below. At each such bar, sigma is the average of the last 14 moves stored for that half-hour, a move being the absolute change from that session's 09:30 open to the close of its half-hour bar, as a fraction (|close / open - 1|); the upper band is (1 + sigma) times the higher of today's open and the prior close, and the lower band is (1 - sigma) times the lower of the two. If flat, buy SPY with 97% of equity when the close is above the upper band, or sell it short with 97% of equity when the close is below the lower band. If long, sell the position when the close is below the higher of the upper band and VWAP, and if that close is also below the lower band, sell SPY short with 97% of equity when the next bar closes. If short, buy to cover when the close is above the lower of the lower band and VWAP, and if that close is also above the upper band, buy SPY with 97% of equity when the next bar closes. Then store today's move for that half-hour. Until 14 moves are stored for a half-hour, only store the move there. Close any position when the bar before the closing-print minute closes (15:59, or 12:59 on early-close days). Use a margin account so it can sell short. Backtest the last 3 years with $100,000.
Is it real?
- Strategies tested together
- 10
- Beta to SPY
- 0.00
- Alpha vs SPY
- −24.9% a year
- Costs paid
- $51,433, 995% of the gain before costs
- Data
- 1m decisions on 1-minute bars, 754 sessions
- Fills
- Next open at market, sized at the fill
Read this before you trust it
Only the 1x version runs; the 1,985% headline needs up to 4x leverage. Each position is 97% of equity at its fill, rather than the prior day's equity fixed for the day. A reversal is an exit at the half-hour and the opposite entry one minute later. The final exit fills at the open of the 16:00 closing-print minute (13:00 on early-close days) rather than in the closing auction. The prior close is that 16:00 minute's open, or after an early-close day the 13:00 minute's close, since that day's closing auction is usually that minute's last trade. The authors' published code, which produced the paper's figures, also enters long only above VWAP and short only below it, and subtracts any dividend from the prior close; the build follows the paper's written rule and enters on the band alone. The first 14 sessions of the window store the moves the bands are built from, so the first trade can come on 17 November 2023. On early-close days the checks stop at 12:30. Shorts need a margin account. Each fill pays 0.05% slippage, where the paper charged $0.0045 a share for commission and slippage over its 1.8 trades a session.
Sources