A lower-band touch is not an entry instruction
Price below a lower Bollinger Band describes an observation relative to a moving average and dispersion estimate. It does not prove that price is too low, that a reversal is due, or that a loss can be avoided by waiting. This example requires a completed return inside the lower band and a separate flat-average context. A stock moving persistently downward can remain near or beyond its lower band while the whole band structure follows it lower.
The strategy is a daily long-side educational example. It targets the middle-band value recorded at confirmation rather than assuming that the middle will remain at the same price or that price must reach it. Freezing that target prevents a falling average from quietly changing the objective after the entry is established.
Specify the band calculation
This version uses the simple average of 20 completed closes as the middle band. Upper and lower bands sit two population standard deviations above and below that average, calculated from the same 20 closes. Population and sample standard deviations differ, and platform defaults can differ as well. Save the convention rather than assuming that every “20, 2” label represents an identical series.
Each prior close is compared with its own prior band values. Yesterday's close below today's lower band is not a substitute for yesterday's close below yesterday's lower band. The moving window changes both its mean and its dispersion, so a valid event comparison requires both observations from their respective calculation dates.
Define the flat-average filter
The absolute change in SMA20 over five completed sessions, divided by current ATR14, must be no greater than 0.5. With current middle band at $52, its five-session-earlier value at $51.80, and ATR at $1, the normalized change is 0.20. It passes the example filter. If the earlier middle was $53, the normalized change is one and fails, even if the last candle appears to bounce.
This filter provides a measurable interpretation of a relatively flat average. It does not independently prove a stable trading range or eliminate trend risk. It also depends on ATR, so a sudden volatility increase can change the normalized slope. Treat it as an input to the stated experiment rather than a universal description of market regime.
Check the re-entry event
The prior close must be below the prior lower band. The current close must be above the current lower band, below the current middle band, and above the prior close. These conditions isolate a specific recovery into the lower half of the bands. A close already above the middle band is not accepted as a late version of the same entry because the original target reference would no longer lie above it.
The next-session proposed entry must also be below the fixed middle-band target. A gap toward or above that target can invalidate the transaction geometry even when the confirmation bar was correctly identified. The example preserves the signal and canceled proposal separately instead of moving the target upward after seeing the gap.
A fictional lower-band recovery
Suppose yesterday closed at $48.70 below its $48.90 lower band. Today the bands are $49, $52, and $55, and the close is $49.50. The current low is $48.80. With the flat-average example above, the numeric context and recovery conditions pass. A hypothetical next-session entry at $50 has a stop reference at $48.70, using the confirmation low minus 0.1 ATR.
The fixed target remains $52. Gross loss distance is $1.30, while target distance is $2. Costs and adverse friction reduce the modeled ratio. If entry changes to $51.70, only $0.30 remains to the target while the stop stays tied to the original confirmation low. A valid-looking recovery does not preserve the economics of an earlier proposed entry.
Now set the current close to $48.90. The recovery back inside the current lower band fails. Alternatively, retain the close but make the normalized middle-band change 0.8: the context filter fails. These failures should remain distinct in the worksheet. A favorable hypothetical target result cannot repair either missing condition.
Manage moving indicators and fixed instructions
The target uses the middle value recorded at confirmation. The closing invalidation uses each later bar's own lower band. Those are different update policies and should be labeled explicitly. Without that distinction, a tester can select whichever moving or fixed reference produces the more favorable exit on each trade.
The protective instruction is checked first, followed by the fixed target, a completed close below the lower band, and a five-session time limit. A falling market can gap through the stop, and a daily bar can contain both stop and target. Neither event is resolved simply by drawing the bands more smoothly.
What to practice before interpreting outcomes
Use the tool to change prior-band relationships independently of current ones. Then test the exact 0.5 normalized-slope boundary and an entry near the middle target. Confirm that the displayed result changes for the correct reason. Finally leave the indicator-setting confirmation unknown and observe that the worksheet does not pretend to verify the platform calculation.
The lesson is about defining a mean-reversion hypothesis with observable conditions and preserving its target policy. It is not a claim that a two-standard-deviation band provides a particular coverage probability for market prices or that a band recovery establishes a profitable edge.
Versioned rule card
EDU-2026-11-v1. These illustrative settings define one educational experiment. They are not optimized parameters, a verified edge, or personal trade instructions.
Entry condition 1: Daily bands use SMA20 plus/minus two population standard deviations of the same 20 completed closes.
Entry condition 2: Absolute SMA20 change over five sessions divided by current ATR14 at most 0.5.
Entry condition 3: Prior close below prior lower band; current close back above current lower band but below current middle band.
Entry condition 4: Current close above prior close.
Entry condition 5: Model next-session entry only below the middle-band target fixed at confirmation.
Exit rule 1: Modeled stop equals confirmation low minus 0.1 ATR14.
Exit rule 2: Target is the middle band recorded at confirmation; do not silently move it.
Exit rule 3: Protective exit first; otherwise target, then a completed close below lower band, then five-session time exit.
From a rule match to an executable paper plan
A strategy condition and an executable order are different objects. First record the completed observation that satisfied the condition. Then write the earliest moment at which that observation could have been known. A rule requiring the final daily close cannot normally be evaluated early enough to assume a fill at that same closing price. The next session may open above, below, or through the proposed reference. Intraday examples have the same sequencing issue at the boundary between a completed five-minute bar and the next bar. The hypothetical entry used in the exercise is a separate input for exactly this reason.
Write the order instruction before calculating a result. A market instruction emphasizes execution while allowing price uncertainty. A limit instruction imposes a price boundary while accepting nonexecution or partial execution. A stop reference is an activation condition, not an insurance contract. A stop-limit instruction can activate without filling. If a gap skips the reference, the chart does not prove that the position could have been closed at that reference. A realistic paper record therefore preserves the planned price, actual assumed fill, quantity filled, remaining quantity, and the reason for any cancellation. These distinctions are explained further in the site's Trading Basics lessons.
Quantity, costs, and a worked execution budget
For a hypothetical long position, planned price risk per share is entry minus stop. For a hypothetical short, it is stop minus entry. Both expressions must be strictly positive. Suppose a long entry is $50, a stop reference is $48.50, the total adverse-friction assumption is $0.05 per share across the round trip, and fixed round-trip charges are $2. With a modeled $100 loss budget, the arithmetic quantity is floor((100 − 2) / (1.50 + 0.05)), or 63 shares. This models $99.65 of loss at the assumed exit, rather than proving an actual maximum loss.
A separate cash check matters. If only $2,000 is allocated to the share purchase and charges, floor((2000 − 2) / 50) permits 39 shares in this simplified cash calculation. The smaller quantity governs the long example. With those 39 shares, a hypothetical $53 target and the same friction produce 39 × (3 − 0.05) − 2 = $113.05 of modeled net gain. The modeled stop loss is 39 × 1.55 + 2 = $62.45. Their ratio is approximately 1.81. That ratio describes two assumed outcomes; it says nothing about how likely either outcome is.
The exercise uses whole shares and a single fixed-friction assumption to keep the arithmetic visible. Borrow fees, financing, taxes, changing spreads, partial fills, market impact, and a stop skipped by a gap require additional treatment. In particular, the long cash-cap input is not a margin or buying-power model for a short position. Review aggregate exposure and correlated positions using the site's Risk Management learning path before treating any individual-position calculation as a complete portfolio plan.
Exit priority and ambiguous bars
The listed exits describe distinct reasons for leaving. The protective reference is checked first under the illustrative plan; a target or closing-bar condition is considered afterward. However, a daily or five-minute OHLC bar may contain both the stop and target without revealing which was reached first. Do not choose the favorable sequence simply because it improves a result. Use finer trustworthy observations, flag the sequence as unresolved, or apply a disclosed conservative convention consistently. A closing-bar exit cannot be assumed to have prevented an earlier intrabar stop touch.
If a target is fixed at confirmation, preserve that original value. If a strategy explicitly updates a trailing reference, record the update time and make the new value effective only afterward. A revised stop should never be applied retroactively to the bar that generated it. A time exit also needs an execution assumption: submitting an order at a chosen time does not establish that the observed chart price was the actual available fill. Record the difference instead of silently replacing the plan with hindsight.
Review evidence without manufacturing a track record
Start a paper test by freezing the rule version, instrument universe, session definition, time zone, data source, corporate-action treatment, and cost assumptions. Include every eligible observation, including failed conditions and unfilled orders. A collection of attractive screenshots is not a complete sample. Missing data should remain missing rather than being converted into a passing condition. The examples here illustrate mechanics and arithmetic; none of the chart paths establishes a historical win rate or expected future return.
If testing historical data later, separate development observations from a subsequent evaluation period. Record every parameter combination tried so that an attractive result is not presented as an independent discovery after many hidden trials. Include delisted instruments when the universe requires them, avoid using future information to select past leaders, and check how split and dividend adjustments affect price and volume comparisons. These details can change which bars qualify even when the indicator names remain identical. The Technical Analysis lessons provide related material on indicators, reference windows, and disciplined testing.
Using the dedicated practice exercise
Open the exercise linked to this strategy and first review the fictional defaults. Change one assumption at a time. The tool evaluates the supplied numbers and explicitly entered review states; it does not retrieve an independent chart or verify a person's judgment. “Meets entered example conditions” means only that the entered conditions satisfy this particular rule version. “Unresolved” identifies required information that has not been confirmed. A failed condition remains a failed condition even when the hypothetical target result is attractive.
After reviewing the result, preserve the inputs and write which observation changed the conclusion. Repeat with a near-boundary example, a missing requirement, and an adverse execution assumption. Reset restores the fictional example. The worksheet does not send orders, connect a broker, or save a trading account. A repeatable process includes decisions to wait, reject a setup, or investigate an unknown; it is not a requirement to produce a trade every session. The Trading Psychology lessons explain how to review those decisions independently of profit and loss.
Indicator background
For indicator definitions, see Fidelity’s Technical Indicator Guide. The particular thresholds and combinations on this page are our illustrative rule version, not a strategy endorsed or validated by that reference.