<- all tokdocs

The Noise Band SPY Intraday Momentum Strategy: Does the Backtest Hold Up?

By , founder of Mavric and Apso

View on TikTok ->

The Noise Band is an intraday momentum strategy that trades breakouts from a volatility band drawn around the market open, and it comes from a published 2024 paper by Barbon, Aziz and Zarattini that reported a 19.6% annualized return on SPY. A TikTok from @thequantbuilder presents it as the one survivor of 50 strategies he tested, turning $10,000 into $150,000 since 2018. The video's own charts are labeled "SCHEMATIC" and "ILLUSTRATIVE BACKTEST," and an independent out-of-sample test found the edge weakened sharply after publication.

This is a review of a trading claim, not investment advice.

What is the Noise Band strategy?

It is the "Noise Area" rule from Swiss Finance Institute Research Paper N°24-97, "Beat the Market: An Effective Intraday Momentum Strategy for S&P500 ETF (SPY)," by A. Barbon, A. Aziz and C. Zarattini, dated 27 November 2024. The abstract says it "uses a well-defined 'Noise Area' to filter out unproductive trades and identify intraday zones of abnormal demand and supply imbalances."

A review by Quant Macro describes the mechanics:

  • Build a volatility cone around the open from historical intraday returns, which captures the typical shape of intraday volatility.
  • Enter when price breaks beyond the band times a factor.
  • Exit on a VWAP-based trailing stop, or at the close. Nothing is held overnight.

The video's animation matches: a widening band from the open, a trailing stop, and "you're flat before the day ends." The video lists it as "Strategy 27 of 50" but never names the paper or its authors.

What returns did the original paper report?

The SFI abstract reports that from 2007 to early 2024 the portfolio "achieved a total return of 1,985% (net of costs), an annualized return of 19.6%, and a Sharpe Ratio of 1.33."

The video's claim is much higher. Turning $10,000 into $150,000 from 2018 is 15x, which works out to about 40% a year over eight years, or about 36% a year through October 2026. That is roughly double the paper's rate. The video gives no position size, borrowing, contract, commission or slippage, so there is no way to see how the curve was built.

Did the Noise Band work after it was published?

Not by the evidence so far. Sheimy Paz Serpa's "Out-of-Sample Evaluation of an Intraday Momentum Strategy for the S&P 500 ETF (SPY)," dated June 2026, replicated the rule on one-minute data. Larry Swedroe's review on 22 October 2026 says the in-sample replication matched the original almost exactly, at 19.8% annualized and a Sharpe of 1.34. The out-of-sample period, May 2024 through March 2026, "told a very different story than the in-sample ones."

The video says "the edge kept coming back." That describes the backtest era, not the two years after publication.

How often does the Noise Band lose?

The video says "Half the trades lose." The Quant Macro review puts the win rate at about 43%, so closer to 57% of trades lose. The strategy relies on a few large winners paying for many small losses.

Are the video's validation tests real?

The video says it tried to kill the strategy with random-entry tests, six blind walk-forward tests and tick-level runs in MetaTrader on Nasdaq and S&P. Each matching chart carries a small grey label: "SCHEMATIC / TESTED AGAINST RANDOM ENTRIES," "SCHEMATIC / SIX BLIND WALK-FORWARD TESTS" and "SCHEMATIC / TICK-BY-TICK NASDAQ + S&P." The $10,000 to $150,000 chart, with "~50% LOSING TRADES" and "30% DRAWDOWN," is labeled "ILLUSTRATIVE BACKTEST."

No strategy tester report, trade count, Sharpe ratio or parameter set appears anywhere. The charts are drawings of what such tests might look like.

Can MetaTrader backtest every tick since 2018?

It can try. The MetaTrader 5 help describes an "Every tick based on real ticks" mode that uses tick data accumulated by your broker. It also says that where real ticks are missing, the tester simulates prices instead. Broker tick history often does not go back to 2018, so a credible claim would name the broker, symbol and modeling quality. The video names none of them.

How many published trading strategies fail?

The video says 49 of 50 failed, a 98% rate, with no scoring method. The research literature puts it lower. The NBER paper "Replicating Anomalies" by Kewei Hou, Chen Xue and Lu Zhang found 286 of 447 anomalies (64%) insignificant at the 5% level, and 380 (85%) at a t-value of three. McLean and Pontiff's "Does Academic Research Destroy Stock Return Predictability?", which examined 97 predictors from 79 papers, is the standard reference on how returns decay after publication.

Some entries on the video's list have recent evidence against them. For Opening Range Breakout, item 01, Mulham Fetna's pre-registered study of 225 settings on nine US futures markets found none survived costs, with the grid earning $1.57 million gross and losing $6.49 million at $25 per round turn. On Micro E-Mini Nasdaq 100 futures, Mathias Mesfin's arXiv study tested fourteen families of intraday momentum signals over 947 trading days from 2021 to 2025 and found none met all five success criteria after realistic costs.

The video's list shows items 01 through 31 and 48 through 50. Items 32 through 47 never appear on screen.

Bottom line

The Noise Band is a real, published strategy with a strong backtest from 2007 to early 2024. Its first out-of-sample test after publication found it underperformed. The video's 15x result has no shown backtest behind it, and its own charts call themselves schematic.

Sources

Published October 25, 2026. Based on a TikTok by @thequantbuilder. Full video notes (markdown)