Mean Reversion Trading Strategy for Futures: How It Works on ES and NQ

Mean reversion bets that price stretched away from a reference value will drift back toward it. The idea is easy to state and easy to lose money on. The market sometimes does not come back that day, and one stretched move that keeps going can cost more than several small winners earned. This guide covers how to define stretched, how to size the trade, and how to check whether the setup works on your own data.

Why price might revert, and why it might not

Common explanations include large participants working orders around VWAP, overreaction to news, options hedging near popular strikes, and resting orders stacked at obvious levels. These are plausible descriptions. They are not proven causes, and none of them guarantees a repeatable edge in ES or NQ. Treat reversion as a hypothesis you test, not a property of the market you can count on.

Defining stretched

There are two common ways to measure how far price has moved from a reference.

VWAP deviation bands

Use VWAP as the mean and standard deviation bands around it as the ruler. Price at two bands above VWAP is stretched relative to that session. Band width depends on your platform's settings and tends to widen as the session builds, so the same two-band touch means different point distances at 10:00 and at 2:00.

ATR distance

Measure the move from a reference, such as the session open or a prior level, in multiples of ATR. A move of 1.5 to 2 times ATR is large for the current volatility. The advantage is that it adjusts when volatility changes. The tradeoff is that you must choose the ATR period and timeframe, and results change with those choices.

A hypothetical worked example

Suppose ES VWAP is 5,200.00 and one band is 4 points wide. Price reaches 5,208, two bands above VWAP. You short at 5,208, put a stop at 5,211 (3 points, or $150 per ES contract, $15 per MES) and target VWAP at 5,200 (8 points, or $400 per ES, $40 per MES). Reward to risk is about 2.67 to 1.

This is a hypothetical calculation, not a result. It shows the win rate you would need, not the win rate you will get. It also assumes the target fills, and VWAP moves during the trade, so the real distance to the mean shifts. Stop orders can slip past 5,211 in a fast market.

Trend days are the failure mode

On a day when price travels one way and never returns, each counter-trend entry gets stopped and the next one gets stretched further. The usual filters are the overnight gap, scheduled news, VWAP slope, and whether price is being accepted outside the prior day's range. None of these is proven to separate range days from trend days in advance, and the regime can change midday. Test each filter by comparing results with and without it, and count how many days it removes. A filter that skips most days leaves you with too few trades to judge anything.

Sizing for the bad day

Reversion setups often produce more small winners and fewer large losers, but that shape is something to measure in your own log, not assume. Size so the worst realistic loss, stop distance plus slippage, fits inside your daily limit. A simple risk-of-ruin formula assumes independent trades with fixed outcomes. Variable dollar risk and drawdown limits need a simulation, and even that is only as good as the data fed into it.

A manual version you can test

1. Write down your definition of range conditions before the open, using a gap limit, a news check and VWAP slope.
2. Define stretched as a number, such as two bands from VWAP or 1.5 times ATR from the open.
3. Define the exhaustion signal you require, if any, in a way you can check on a chart afterward.
4. Place the stop beyond the extreme, sized so the dollar risk is within your limit.
5. Target the mean area, and write the exit rule for the case where price accelerates through VWAP the wrong way.
6. Log every instance, including the ones you skipped.

Thirty trades is not a statistically proven edge, and 50 simulated trades is not live proof. A simulated fill also tells you nothing about queue position or slippage in the real order book.

Where the Marty Bot fits

The YMI Marty Bot is a NinjaTrader strategy built around mean reversion on ES. Its current logic and variants are documented on the Marty Bot specs page, which is the place to read how it behaves rather than relying on summaries. Bot access is part of YMI Pro at $795 per month with no trial. Intro and VIP do not include bots. Also check any prop firm's rules before running automation. Apex explicitly prohibits automated or algorithmic preset bots that place, modify or cancel trades without personal real-time review, per its prohibited activities page and user agreement. Do not treat the Marty Bot, or any bot, as Apex compatible or as a way to pass an Apex evaluation.

Related Reading

• VWAP Indicator Guide covers the reference most reversion setups are built on.
• ATR Indicator Guide covers measuring extension and sizing stops.

If you want bot access, see YMI Pro, a paid membership with no trial.