Research framework · 6J / JPY-USD

6J Mean Reversion Signals: Define the Mean Before You Fade

“6J is stretched” is not a trade. Stretched from what mean, over what window, in what regime, and invalidated where? Mean reversion can be tested only after those choices are frozen. During genuine policy or rate repricing, yesterday’s mean may be the wrong destination entirely.

Candidate means
3
Hypothetical setups
3
6J tick
$6.25
Validated edge
Not claimed
Mean-reversion sequenceGate before signal
1

Define mean

One formula

2

Gate regime

Range, not repricing

3

Trigger

Observable close

4

Kill idea

Price and time

Hypothesis workflowNot a performance claim

Direct answer

Mean Reversion Is a Conditional Hypothesis

The hypothesis is that, under a defined range or stationary-enough regime, an extreme away from a fixed reference will move back toward that reference before invalidation or a time stop. It is not the belief that price “has gone too far” or that every gap fills.

Never average into an undefined trend

A lower entry does not improve a long thesis when the mean has shifted and structure keeps accepting lower prices. Total size, add levels and worst-case invalidation must be fixed before the first order. If they are not, adding is simply increasing risk while evidence moves against you.

Name the destination

Three Different Means, Three Different Questions

Do not swap the reference after entry. If a session VWAP fade fails, changing the target to a 50-bar average does not rescue the original test.

MeanTimestamp-safe definitionMain limitation
Session VWAPCumulative sum of trade price × trade volume divided by cumulative volume from a fixed session startDepends on session boundary and trade feed; it resets and can migrate during a trend
Rolling averageSimple or exponential average of the last N completed bars, using a fixed N and bar intervalLags and follows price; optimizing N on the same sample invites overfit
Range midpoint(pre-confirmed range high + range low) ÷ 2, frozen until the range breaks or expiresThe reference becomes invalid after accepted breakout or structural transition

For VWAP research, use actual trade volume from the selected 6J contract. Tick volume from another feed or spot venue is a different input and must be labeled separately.

The signal does not outrank the regime

Pass a Range Gate Before Looking for an Extreme

Reject the setup when structure or macro conditions suggest price discovery rather than oscillation. The gate itself must use completed data.

Range structure passes

Fixed boundaries have multiple tests, sufficient elapsed bars and no accepted close outside under the prewritten break rule.

Trend filter passes

Higher-timeframe slope, swing sequence or directional-efficiency measure stays inside the strategy’s tested neutral band.

Volatility passes

Current range is neither too small to clear costs nor above the strategy’s tested shock threshold.

Event gate passes

No BOJ decision, major U.S. release or other scheduled event sits inside the blackout. No unscheduled policy headline is moving the market.

Execution passes

Spread, depth and recent slippage remain inside measured limits for the selected expiry.

Mean remains valid

The reference was calculated from available data and has not been invalidated by a session reset, accepted break or time expiry.

Falsifiable examples

Three Hypothetical 6J Mean-Reversion Setups

These are research specifications, not validated signals. Test every rule with realistic bid-ask execution, fees, slippage, roll handling and an untouched out-of-sample period.

Setup 1: session VWAP rejection

Mean: trade-weighted VWAP from the chosen CME session boundary. Regime: 15-minute structure remains a confirmed range; no event blackout; VWAP slope is inside a pretested neutral threshold. Extreme: price closes at least one pretested deviation unit above VWAP, where the unit is either rolling standard deviation or completed-bar ATR fixed before testing. Entry: short only after a one-minute close returns inside the deviation band and the next trade breaks that bar’s low. Invalidation: above the excursion high plus the strategy’s volatility buffer. Target: first touch of VWAP or the nearer internal range level. Time stop: exit after ten one-minute bars if price has not reached halfway to VWAP. Mirror the rules for a long. Skip if the mean is moving sharply toward the extreme.

Setup 2: failed range breakout

Mean: midpoint of a pre-confirmed 30-minute range. Regime: no accepted higher-timeframe trend break and no scheduled event in the window. Entry: after price closes beyond the range, closes back inside within three completed bars, then breaks the return bar toward the midpoint. Invalidation: beyond the failed-break extreme plus a measured buffer. Target: midpoint first; optional remainder only if tested toward the opposite quartile. Time stop: exit if midpoint is not touched within six 5-minute bars. If price closes back beyond the boundary under the acceptance rule, exit rather than widen.

Setup 3: fixed-average overshoot

Mean: 20-bar simple average of completed 15-minute closes. Regime: 4-hour swing state is neutral and the 20-bar average slope is inside the tested band. Extreme: close more than a fixed multiple of 20-bar residual volatility from the mean. Entry: next bar must close back toward the average and inside the extreme threshold. Invalidation: a new extreme close beyond a second threshold or a structural swing break, whichever comes first. Target: the average. Time stop: four completed 15-minute bars. Do not keep a stale trade while the rolling mean chases price.

Every threshold above is intentionally unnamed. Supplying a made-up “best” ATR multiple, slope band or holding period would turn an example into an unsupported recommendation.

Risk from invalidation, not conviction

Size the Structural Stop in Dollars

Standard 6J pays or loses $6.25 per outright tick. The per-contract risk estimate is stop ticks × $6.25 + expected adverse slippage + round-turn costs. Position size is the account’s fixed trade-risk allowance divided by that estimate, rounded down.

Hypothetical sizing example

Suppose the entry-to-invalidation distance is 36 ticks. Gross stop risk is 36 × $6.25 = $225. Add two ticks of modeled adverse stop slippage ($12.50) and $5.40 in fees: $242.90 estimated risk per contract. With a $500 trade-risk cap, two contracts estimate $485.80; three estimate $728.70 and fail the cap. A worse fill can still exceed the estimate.

Structure stop
36 ticks
Gross risk
$225.00
Modeled costs
$17.90
Per contract
$242.90

Costs are hypothetical and must be replaced with account-specific realized data. If one contract exceeds the risk limit, skip standard 6J or evaluate a genuinely smaller instrument after verifying its separate specifications and liquidity.

Confirmation should add information

Use Cross-Market Context Without Double Counting

USD/JPY spot, U.S. and Japanese rates, broad dollar measures and equity volatility can help identify repricing. They do not confirm a trade merely because a transformed chart looks similar to 6J.

ContextUseful questionBoundary
USD/JPY spotDoes synchronized reciprocal spot agree with the broad direction?Account for inverse quote, timestamp, venue and futures basis
U.S.-Japan ratesIs the expected rate differential repricing persistently?A macro relationship is not an intraday entry signal
Dollar basket / crossesIs the move yen-specific or dollar-wide?Shared USD exposure makes many series statistically dependent
Risk assets / volatilityIs deleveraging changing the regime?“Risk-off” is not a deterministic yen rule
6J volume and depthIs the selected expiry showing normal execution conditions?Volume does not reveal trader identity or intent

The most important signal is sometimes stop

When Not to Fade 6J

Mean reversion is structurally vulnerable when new information moves the market’s reference price. Default to no trade in these states unless a separate event strategy has its own evidence.

Policy and intervention

  • BOJ decision, press conference or material guidance surprise.
  • Suspected or confirmed Japanese FX intervention.
  • Official comments producing abnormal spread, range or depth.
  • Do not authenticate intervention from a candle; use MOF records.

Rates and structure

  • Persistent U.S.-Japan rate repricing.
  • Accepted break from the range that defined the mean.
  • Higher-timeframe swing sequence becomes directional.
  • Rolling average accelerates toward price rather than remaining a stable reference.

Also stop when the data or platform state is uncertain, the active contract is rolling, the spread exceeds the model, the daily loss limit is hit or actual slippage leaves the tested range. A strategy cannot recover evidence discipline by trading more often.

Frequently asked questions

6J Mean Reversion Questions

What mean should I use for 6J mean reversion?

There is no universal mean. Session VWAP, a fixed rolling average and a pre-defined range midpoint answer different questions. Choose one before testing, calculate it without future data and do not switch after seeing the trade outcome.

Are 6J mean-reversion signals reliable?

Reliability is not established by this article. A setup needs timestamp-safe historical testing, realistic bid-ask costs, slippage, exact contract rolls, regime labels and out-of-sample validation before any expectancy claim.

When should a trader avoid fading 6J?

Avoid an ordinary fade during untested BOJ or U.S. data windows, suspected or confirmed intervention, persistent rate repricing, accepted range breaks, abnormal spread or depth, and any trend regime that fails the strategy's prewritten gate.

How should a 6J mean-reversion trade be sized?

Measure the distance from entry to structural invalidation in ticks, multiply by $6.25 and add estimated slippage and fees. Divide the account's fixed risk allowance by that per-contract risk and round down; skip the trade if one contract is too large.

Sources, methodology and editorial disclosure

Sources were checked August 12, 2026. All setup prices, thresholds, costs and risk calculations are hypothetical examples. No backtest was performed for this article and no reliability, expectancy, profitability or best-parameter claim is made. A valid study would require timestamp-safe contract data, a disclosed roll method, bid-ask execution, costs, regime labels and out-of-sample results. Grizzly Parrot Trading is not sponsored by the cited institutions.