Double Bottom Pattern: Neckline and Confirmation

A double bottom is not confirmed merely because a chart resembles the letter W. It is a bullish reversal candidate containing two lows near a predefined price area, an intermediate high that becomes the neckline, and a prior decline that gives the structure something to reverse. The second low completes the visual candidate. A separate neckline rule determines whether the candidate becomes confirmed.

The useful question is therefore not whether double bottoms are “essential” or whether every second low represents institutional accumulation. The useful questions are:

How were the two lows selected?
How similar must their prices be?
How much separation is required?
Which intermediate high defines the neckline?
What event confirms the pattern?
What event invalidates or expires it?
How will outcomes be evaluated without changing the rules afterward?

This guide turns those questions into a repeatable classification and replay workflow.

Educational note: Chart patterns organize historical price observations. They do not guarantee a reversal, an executable price, or a profit. Historical and simulated results can differ materially from live trading because of spread, slippage, liquidity, order timing, commissions, fees, and discretionary decisions.

Key Takeaways

Two similar lows create a candidate, not an automatic bullish signal.
The pattern needs a prior decline, two defined lows, an intermediate high, and a versioned neckline-confirmation rule.
“Similar lows” requires a numerical tolerance; visual resemblance alone is not reproducible.
A lower second low, equal second low, and higher second low can be tested as separate variants.
A neckline wick, completed close, buffered close, follow-through, and successful retest are different confirmation definitions.
The measured move is an evaluation reference, not a promised destination.
Failed, expired, ambiguous, and unconfirmed candidates belong in the dataset rather than being deleted.

Which Double Bottom or Chart Guide Should You Use?

Question	ChartMini owner
What are the main chart-pattern families?	Chart Patterns Cheat Sheet
How are broad HH, HL, LH, LL, trend, and range states read?	Market Structure Guide
How are BOS and CHoCH rules defined and verified?	Market Structure BOS/CHoCH Guide
How is a raw support or resistance level identified?	Support and Resistance Identification Checklist
How are key-level touch, rejection, breakout, retest, and failure decisions recorded?	Key-Level Decision Framework
How are breakout-pattern families compared?	Breakout Patterns Guide
How is pattern-recognition practice structured?	Pattern Recognition With Chart Replay
How is the double bottom itself defined, confirmed, measured, failed, and reviewed?	This page

This boundary keeps the broad chart-pattern query with the visual cheat sheet while this URL owns the exact double bottom specification.

The Minimum Double Bottom Structure

Use four reference points:

P0: the start or reference point of the prior decline;
L1: the first confirmed low;
N: the intermediate high between the lows, used as the neckline reference;
L2: the second confirmed low.

The chronological order is:

P0 → L1 → N → L2

A fifth event, C, records the selected neckline confirmation after L2.

The pattern is incomplete if any of the following is missing:

no measurable prior decline;
no confirmed first low;
no meaningful rally away from L1;
no intermediate high that can be identified before L2 completes;
no second low inside the chosen similarity rule;
no defined confirmation rule.

A chart can contain two visually similar lows and still fail the specification because the structure occurred inside a narrow range, the lows were selected with hindsight, or the neckline was changed after the breakout.

Freeze the Chart Identity First

Pattern classification can change when the underlying chart changes. Record these settings before marking L1:

Field	Example record
Instrument	ES continuous futures contract
Product or contract	Front-month adjusted series
Data provider	Named provider
Price type	Trade, bid, ask, or midpoint
Session	Regular, extended, or combined
Timezone	Exchange timezone
Bar interval	15-minute, 4-hour, daily, or another stated interval
Adjustment	Split-adjusted, dividend-adjusted, back-adjusted, or raw
Candle status	Closed bars only
Tick size	Instrument-specific increment

Changing session boundaries can change the lows and intermediate high. Changing provider or continuous-contract method can also change whether the lows fall inside the selected tolerance.

For reproducible testing, compare patterns only after the chart identity is frozen.

Define the Prior Decline

A reversal pattern needs a decline to reverse. “Price looked weak” is not a testable definition.

Choose one prior-decline version before reviewing outcomes.

Version A: Pivot-sequence decline

Require a predefined sequence of lower highs and lower lows before L1.

Record:

pivot algorithm;
equality tolerance;
number of required confirmed pivots;
whether the pivot confirmation uses future bars;
the last structural high used as P0.

Version B: Net-change decline

Require a minimum decline from P0 to L1:

prior decline % = (P0 - L1) / P0 × 100

The threshold is a research parameter, not a universal rule.

Version C: Range-normalized decline

Express the decline relative to a volatility measure recorded before L1:

normalized decline = (P0 - L1) / volatility unit

The volatility unit might be ATR, standard deviation, or another frozen measure. The period and calculation source must be recorded.

Version D: Support-break decline

Require price to break a previously defined support reference before L1 forms. The support reference must exist before the break; it cannot be drawn after L1 is visible.

Version E: Slope-based decline

Require a negative slope over a fixed lookback, with the regression method and bar count recorded before testing.

Do not combine these methods casually. A pivot-sequence definition can include a chart that a net-change threshold excludes.

Confirm the Lows Without Look-Ahead Bias

A low is not confirmed while the current candle is still forming. A pivot-low rule usually needs bars after the candidate low, which creates a confirmation delay.

Example pivot version:

candidate low is lower than the previous k lows;
candidate low is lower than the next k lows;
L1 or L2 becomes known only after the right-side bars close.

This matters because a replay decision made on the low itself would use information that was not yet available.

Alternative low definitions include:

lowest low within a frozen rolling window;
close-based pivot rather than wick-based pivot;
local minimum separated by a minimum number of bars;
low confirmed by a minimum rebound from the candidate price.

Record the method. Do not select whichever low creates the cleanest W after the outcome is known.

Define How Similar the Two Lows Must Be

The two lows rarely match exactly. TradingView's published Double Bottom pattern reference explicitly allows a slight difference and includes a configurable permissible-deviation input. The important issue is not visual perfection; it is a frozen tolerance.

Absolute-price tolerance

absolute difference = |L2 - L1|

Qualify the pattern when the difference is no greater than a fixed price amount.

This is simple but does not transfer well across instruments with different prices.

Tick-based tolerance

tick difference = |L2 - L1| / tick size

This makes the rule more interpretable for futures and other instruments with fixed minimum increments.

Percentage tolerance

low difference % = |L2 - L1| / L1 × 100

This scales with price but may still behave differently across volatility regimes.

Pattern-height-normalized tolerance

First define the provisional pattern height:

H = N - min(L1, L2)

Then define:

normalized low difference = |L2 - L1| / H

This expresses the low mismatch relative to the structure itself.

Volatility-normalized tolerance

volatility-adjusted difference = |L2 - L1| / volatility unit at L2

The volatility unit must be calculated without future data.

Zone-overlap version

Instead of exact low prices, create a predeclared band around L1. L2 qualifies if its wick, body, or close overlaps that band according to the chosen rule.

These methods produce different candidate sets. Report the tolerance version with every result.

Higher, Equal, and Lower Second Lows Are Different Variants

Separate the candidates into at least three categories:

higher L2: the second low remains above L1 but inside tolerance;
approximately equal L2: the difference falls inside a narrow equality band;
lower L2: the second low undercuts L1 but remains inside the wider pattern tolerance.

A lower L2 is sometimes described as a spring, liquidity sweep, or stop run. The OHLC chart only records that price traded below L1 and later moved higher. It does not reveal who traded, whether stop orders were present, whether an institution accumulated, or why the move occurred.

Treat undercut depth and recovery as measurable variables:

undercut = max(0, L1 - L2)

recovery close distance = close after L2 - L1

Do not require an undercut merely because it creates a compelling narrative.

Set Minimum and Maximum Separation

Two adjacent candles with similar lows may be a short consolidation, tweezer bottom, or ordinary noise rather than the pattern version you intended to study.

Record:

bars from L1 to L2;
calendar time from L1 to L2;
bars from L1 to N;
bars from N to L2;
minimum separation;
maximum separation or expiry.

A daily-chart double bottom separated by three sessions and one separated by nine months may represent different samples. Pool them only if that decision is made before testing.

Define the Neckline Before the Breakout

The neckline is usually the intermediate high between L1 and L2. The exact selection method still needs a rule.

Highest-high neckline

Use the highest traded price between the confirmed lows.

Highest-close neckline

Use the highest completed close between the lows.

Confirmed-pivot neckline

Use a pivot-high algorithm. Record the left/right pivot width and when the pivot became known.

Body-zone neckline

Create a band from the highest open/close region rather than one wick price.

Multi-peak neckline

If the rally between lows forms several similar highs, define a resistance zone with a frozen width rule rather than choosing one convenient point.

The neckline must not move upward after price breaks merely to make the breakout look stronger. Store the original neckline version and any later revision as separate records.

A Candidate Is Not a Confirmed Double Bottom

Use separate labels:

decline detected;
L1 confirmed;
neckline reference confirmed;
L2 candidate detected;
L2 confirmed inside tolerance;
double bottom candidate complete;
confirmation pending;
confirmed;
failed, expired, or ambiguous.

CME's reversal-pattern education describes double bottoms as a low, a retracement, and another attempt at a low that reverses upward. TradingView's pattern status does not mark the structure as formed until a bar closes above the neckline. These conventions support separating visual formation from confirmation.

Five Neckline Confirmation Rules to Test

Do not call all neckline events “the breakout.” Each rule enters the evidence set at a different time and price.

Rule 1: Intrabar Neckline Break

Confirm when the high of a post-L2 bar trades above N.

Advantages:

earliest signal;
captures cases that do not close above the neckline.

Limitations:

more sensitive to brief excursions;
OHLC does not prove that a simulated order would fill at the displayed price;
a candle can trade above N and close back below it.

Record the maximum excursion above N and the closing location.

Rule 2: Completed Close Above the Neckline

Confirm only when a completed bar closes above N.

This aligns with the formed-status convention in TradingView's Double Bottom reference and avoids treating every wick as confirmation.

Record:

breakout-bar open, high, low, and close;
close distance above N;
whether the bar was complete at decision time;
whether the close occurred during the selected session.

Rule 3: Buffered Close

Require:

close > N + buffer

The buffer may be:

fixed price;
ticks;
percentage of N;
percentage of pattern height H;
volatility-normalized distance.

The buffer must be frozen before the evaluation sample. A larger buffer filters marginal closes but confirms later.

Rule 4: Break Plus Follow-Through

Require a close above N, followed within a fixed number of bars by a second event such as:

another close above N;
a higher close;
a higher high without a close back below N;
a minimum extension above the breakout close.

Define the follow-through window and minimum extension explicitly.

Rule 5: Break Plus Successful Retest

Require:

a confirmed neckline break;
a return to a predefined neckline band within a fixed window;
a completed bar that satisfies a frozen hold or rejection rule;
no invalidation before the retest confirms.

Not every confirmed double bottom retests. A retest-based version therefore creates a smaller and later sample than a close-confirmation version.

Neckline Acceptance and Role Reversal

A single close above N does not prove that old resistance has become support.

Possible acceptance variables include:

number of closes above N;
percentage of candle body above N;
time spent above N;
successful retest count;
maximum penetration back below N;
whether a close returned below the original pattern range.

Treat “resistance became support” as a hypothesis with a testable rule, not an automatic consequence of one breakout candle.

Calculate the Measured Objective as a Reference

A common measured-move reference is:

pattern height H = N - B

where B is the selected bottom reference. B may be:

L1;
L2;
the lower of L1 and L2;
the average of the two lows;
the center of the predefined bottom zone.

Then:

measured objective M = N + H

The bottom-reference version must be recorded because it changes M.

The measured objective is not proof that price will travel that distance. Use it as one outcome threshold alongside:

maximum favorable excursion;
maximum adverse excursion;
return after fixed horizons;
time to M;
invalidation before M;
partial excursion as a fraction of H;
net result after stated costs.

TradingView labels its projected level as an expected target while explicitly warning that such target information is educational and not personal investment advice. That is the correct boundary: measurement is a research convention, not a guarantee.

Define Failure, Invalidation, and Expiry

A useful dataset keeps negative and unresolved examples.

Pre-confirmation disqualification

The candidate can be disqualified when:

L2 falls outside the allowed tolerance;
the prior decline no longer qualifies under the frozen rule;
the intermediate rally is too small to create the required pattern height;
L2 appears before the minimum separation;
the candidate exceeds the maximum duration;
the neckline was not identifiable without hindsight.

Candidate failure

Before confirmation, failure may occur when price closes below a predefined candidate-invalidation boundary.

Possible boundaries:

below L2;
below the lower of L1 and L2;
below the bottom-zone edge;
below that boundary by a fixed buffer;
below it for a specified number of closes.

Confirmed-pattern failure

After confirmation, define failure separately. Examples:

close back below N within x bars;
close below the retest band;
close below both bottoms;
adverse threshold reached before the measured objective;
no required follow-through before expiry.

Expiry

A candidate can expire when no confirmation occurs within a fixed number of bars after L2.

A confirmed pattern can also have an evaluation expiry. If neither the objective nor failure threshold occurs by that time, label the outcome unresolved or time-expired instead of forcing a win/loss classification.

Ambiguous

Use an ambiguous label when:

missing candles affect the event order;
a session gap crosses both confirmation and failure references;
data-provider differences change the classification;
adjustment settings materially alter L1, L2, or N;
OHLC cannot determine which threshold occurred first.

OHLC Bars Do Not Reveal Every Intrabar Event

Suppose one candle has:

a high above N;
a low below the candidate failure boundary;
a close above N.

The OHLC record does not reveal whether the high occurred before the low. A bar-based simulation cannot know whether a stop, entry, or target would have triggered first without lower-timeframe or tick data.

Choose one policy:

exclude ambiguous bars;
use a conservative event order;
inspect lower-timeframe data available at that historical moment;
report a range of possible outcomes;
maintain a separate ambiguity category.

Do not silently select the event order that creates the better result.

Volume Is an Optional Feature, Not Proof

Volume may be recorded at:

L1;
the rally from L1 to N;
L2;
the neckline break;
the retest.

Possible variables include:

raw volume;
relative volume versus a frozen lookback;
L2 volume divided by L1 volume;
breakout volume divided by recent median volume;
price change per unit of volume.

Limitations:

centralized exchange volume and decentralized or broker-specific volume are not equivalent;
volume can vary by session and contract rollover;
higher volume does not identify the participant or prove accumulation;
lower L2 volume does not guarantee reversal;
changing the lookback changes the feature.

Compare volume rules as separate versions rather than making one universal requirement.

RSI Divergence Is Also a Separate Feature

Bullish divergence is often defined as price forming an equal or lower L2 while RSI forms a higher low.

A reproducible version records:

RSI period;
price source;
smoothing method;
which bars correspond to the RSI lows;
minimum RSI difference;
whether divergence is required before neckline confirmation.

RSI divergence is not part of the minimum geometric definition. Requiring it creates a filtered subset. Test that subset against the unfiltered candidates rather than assuming divergence proves seller exhaustion.

Distinguish Similar Structures

Structure	Key difference from a double bottom
V-bottom	One sharp low and rebound, without two separated lows and an intermediate neckline
Higher-low reversal	L2 is materially above L1 and may exceed the selected double-bottom tolerance
Triple bottom	Three qualifying lows and at least two intermediate rallies
Rectangle or range	Repeated oscillation between boundaries without a clear prior-decline reversal sequence
Inverse head and shoulders	Three troughs, with the middle trough normally deeper than the shoulders
Rounded bottom	Gradual curved transition rather than two discrete lows
Tweezer bottom	Usually a two-candle candlestick formation rather than a multi-swing chart pattern
Failed breakdown	Price briefly breaks a level and reclaims it, but may lack L1–N–L2 geometry

A structure can satisfy more than one label. Set a precedence rule or permit multiple labels while evaluating each hypothesis separately.

Separate Classification, Confirmation, and Outcome

These are different questions.

Classification

no pattern;
developing candidate;
complete candidate;
disqualified;
ambiguous.

Confirmation

pending;
intrabar break;
close confirmed;
buffer confirmed;
follow-through confirmed;
retest confirmed;
expired unconfirmed.

Outcome

measured objective reached;
partial extension;
neckline failure;
bottom failure;
time-expired;
ambiguous event order;
unavailable data.

A visually valid candidate can remain unconfirmed. A confirmed candidate can later fail. A failed pattern is not evidence that the original classification was wrong; it may be a valid pattern with a negative outcome.

Define Outcome Measures Before Testing

Do not reduce every result to whether the measured objective was touched.

Fixed-horizon return

Measure return after a frozen number of bars from confirmation.

Maximum favorable excursion

Record the largest movement above the selected confirmation price before expiry or failure.

Maximum adverse excursion

Record the largest movement below the confirmation price over the same event window.

Height-normalized excursion

Express movement as a fraction of H so patterns from different price levels can be compared.

Threshold-first outcome

Record whether M or the failure boundary occurred first. Handle same-bar ambiguity explicitly.

Time to event

Record bars from L2 to confirmation and from confirmation to each outcome.

Expectancy after costs

If simulated entries are evaluated, state:

entry version;
assumed fill;
spread;
slippage;
commissions and fees;
sizing method;
overlapping-signal policy.

Do not transfer a success percentage from one market, timeframe, tolerance, confirmation rule, or cost model to every double bottom.

Research Evidence Requires Careful Boundaries

Andrew Lo, Harry Mamaysky, and Jiang Wang developed systematic chart-pattern algorithms and compared conditional return distributions for patterns including double bottoms across a long historical U.S. equity sample. Their research found that some technical indicators contained incremental information in that sample, while also emphasizing that geometric chart interpretation is subjective unless definitions are made systematic.

That does not establish one universal double bottom win rate. The study used a particular algorithm, market, period, and statistical framework. A ChartMini replay sample using different pivots, confirmation rules, costs, or instruments is a separate test.

The practical lesson is methodological: specify the geometry first, preserve every candidate, and evaluate out of sample.

Use Development, Validation, and Evaluation Samples

Development sample

Use this period to choose:

prior-decline version;
low-confirmation algorithm;
low tolerance;
separation limits;
neckline version;
confirmation rule;
failure boundary;
outcome window.

Validation sample

Compare a limited set of frozen alternatives. Reject definitions that work only under one narrow parameter value.

Final evaluation sample

Run the selected rule without changing it. Report all candidates, including unconfirmed, failed, expired, and ambiguous cases.

If the evaluation result leads to a rule change, the changed rule needs a new untouched sample.

Test Robustness Instead of One Perfect Setting

A result is fragile when small parameter changes reverse the conclusion.

Test nearby values for:

pivot width;
low tolerance;
minimum separation;
maximum duration;
required prior decline;
neckline buffer;
follow-through window;
retest band width;
failure buffer;
outcome horizon.

Report a parameter region rather than only the best historical combination. Searching many versions and publishing only the highest result creates data-snooping risk.

Worked Classification Example

Assume a completed daily-chart sequence with:

prior reference P0 = 112.00;
first low L1 = 94.00;
intermediate high N = 101.50;
second low L2 = 93.40;
tick size = 0.10.

1. Prior decline

(112.00 - 94.00) / 112.00 = 16.07%

Whether that qualifies depends on the frozen prior-decline threshold.

2. Pattern height

Using the lower low as B:

H = 101.50 - 93.40 = 8.10

3. Low difference

|93.40 - 94.00| = 0.60

Tick difference:

0.60 / 0.10 = 6 ticks

Height-normalized difference:

0.60 / 8.10 = 7.41%

The candidate qualifies only if the chosen tolerance permits those values.

4. Second-low variant

L2 is 0.60 below L1, so this is a lower-second-low variant.

That fact does not prove a stop run. It is simply the observed price relationship.

5. Close confirmation

Suppose the first completed close above N is 102.10.

Close distance:

102.10 - 101.50 = 0.60

If the rule requires only a close above N, confirmation occurs. If the rule requires a 1.00 buffer, it does not.

6. Measured objective

M = 101.50 + 8.10 = 109.60

Store 109.60 as an evaluation reference. Also record MFE, MAE, failure, costs, and the observation window.

A Repeatable Double Bottom Checklist

Before looking at future candles, record:

instrument and exact product;
data provider;
session and timezone;
bar interval;
adjustment method;
closed-bar requirement;
prior-decline version;
pivot-low rule;
L1 confirmation time;
intermediate-high rule;
neckline value or band;
minimum rally height;
L2 confirmation time;
low-similarity tolerance;
higher/equal/lower L2 category;
minimum and maximum separation;
confirmation rule;
confirmation buffer;
follow-through or retest window;
pre-confirmation invalidation;
post-confirmation failure rule;
measured-objective reference;
evaluation horizon;
ambiguity policy;
costs and assumed fills;
development or evaluation sample label.

A missing field is not automatically fatal, but it weakens reproducibility.

Double Bottom Replay Worksheet

Chart identity

Instrument:
Product or contract:
Data source:
Session:
Timezone:
Bar interval:
Price type:
Adjustment:
Tick size:
Closed bars only: yes / no

Prior decline

Version:
P0 time and price:
L1 time and price:
Decline amount:
Decline percentage:
Structural state before L1:
Rule qualified: yes / no / ambiguous

First low

L1 candidate time:
L1 confirmed time:
Pivot method:
Wick or close basis:
Confirmation delay:

Neckline

N time and price:
Selection method:
Line or zone:
Zone width:
Known before L2 completed: yes / no
Rally height from L1:

Second low

L2 candidate time:
L2 confirmed time:
L2 price:
Difference from L1:
Tick difference:
Percentage difference:
Height-normalized difference:
Higher / equal / lower variant:
Bars between lows:
Tolerance qualified: yes / no

Optional features

L1 volume:
L2 volume:
Relative-volume version:
RSI settings:
RSI at L1:
RSI at L2:
Divergence version:
Support or zone known before L2:
Higher-timeframe state:

Confirmation

Rule version:
Neckline buffer:
Confirmation time:
Confirmation price:
Follow-through window:
Retest required: yes / no
Retest-band definition:

Failure and expiry

Candidate invalidation:
Confirmed failure rule:
Expiry bars:
Ambiguity policy:

Outcome

Bottom reference B:
Pattern height H:
Measured objective M:
Objective reached: yes / no / ambiguous
MFE:
MAE:
Fixed-horizon return:
Bars to confirmation:
Bars to outcome:
Failure event:
Time-expired:
Costs included:

Audit

Candidate recorded before outcome: yes / no
Rule changed afterward: yes / no
Screenshot before confirmation:
Screenshot after outcome:
Development / validation / evaluation sample:
Notes:

How to Practice With ChartMini

ChartMini can be used to hide future candles and review double bottom rules one bar at a time.

A practical exercise:

Select a market, interval, session, and historical sample before viewing the chart.
Advance until a prior decline and L1 are visible.
Pause and record the low-confirmation rule.
Advance until the rally creates N.
Freeze the neckline before L2 completes.
Advance one candle at a time while testing whether L2 falls inside the tolerance.
Classify the candidate before viewing the neckline outcome.
Record the selected confirmation and failure rules.
Continue until confirmation, failure, expiry, or ambiguity occurs.
Record the outcome without deleting unsuccessful examples.

What ChartMini does not reproduce

ChartMini does not reproduce:

live bid/ask spread;
order-book depth;
queue position;
partial fills;
broker latency;
exact intrabar sequence without lower-resolution data;
market impact;
commissions and financing unless added manually;
guaranteed stop or limit fills;
the identity or motive of market participants.

Use replay to improve classification discipline and research records, not to claim that a historical pattern would have produced the same live result.

Common Double Bottom Errors

Calling the second low a completed pattern

The visual candidate may be complete at L2, but the chosen confirmation rule can still be pending.

Selecting the lows after seeing the breakout

This introduces hindsight. Apply the same pivot and tolerance rules before the outcome.

Requiring exact equality

Real lows can differ. Use a frozen tolerance rather than visual perfection.

Calling every lower L2 a liquidity sweep

An undercut is observable. The reason for it is not directly visible in OHLC data.

Moving the neckline after the break

Store the original N. A later alternative is a new version, not a silent correction.

Treating volume or RSI as mandatory proof

They are optional features whose definitions and data sources must be tested.

Treating the measured move as guaranteed

It is one projected reference. Record partial movement, failure, and time expiry.

Mixing timeframes and markets without labels

A five-minute futures sample and a weekly equity sample may have different distributions and costs.

Ignoring unconfirmed candidates

Deleting candidates that never break the neckline creates selection bias.

Assuming replay equals execution

Historical bars do not reproduce actual fills, costs, liquidity, or decision pressure.

Practical Next Step

Choose one market and one interval. Freeze one prior-decline rule, one low-tolerance method, one neckline method, and one confirmation rule. Record at least 30 consecutive candidates—including failures and expiries—before changing the specification.

The objective is not to prove that the double bottom is essential. It is to determine whether a clearly defined version provides useful information in a stated dataset under stated assumptions.

Frequently Asked Questions

What is a double bottom pattern?

A double bottom is a bullish reversal candidate formed by two lows near a predefined price area with an intermediate high between them after a prior decline. The second low creates a candidate, while a chosen neckline-break rule is normally required before the pattern is classified as confirmed.

When is a double bottom confirmed?

Confirmation depends on the rule selected before testing. Common versions include an intrabar move above the neckline, a completed close above it, a buffered close, a break with follow-through, or a breakout followed by a successful retest. These versions should not be mixed within one result set.

How close must the two double bottom lows be?

The lows do not need to match exactly. A reproducible method defines an absolute, percentage, tick-based, or pattern-height-normalized tolerance before reviewing outcomes. The selected tolerance should be recorded because changing it changes which structures qualify.

Does the second low have to undercut the first low?

No. A second low may be higher, approximately equal, or slightly lower if it remains inside the predefined tolerance. A temporary undercut can be recorded as one variant, but it is not required and does not prove that a stop run or institutional accumulation occurred.

How is a double bottom measured objective calculated?

A common reference measures the vertical distance from the selected bottom price to the neckline and projects that distance above the neckline. It is an evaluation benchmark, not a guaranteed target, and results should also record failure, maximum favorable excursion, maximum adverse excursion, costs, and time to outcome.

Can ChartMini prove that double bottoms are profitable?

No. ChartMini can hide future candles and help you classify double bottom candidates, compare confirmation rules, and record historical outcomes. It does not reproduce live spreads, liquidity, order priority, slippage, market impact, commissions, or discretionary execution, so replay results are not proof of live profitability.

Sources and Evidence Boundaries

CME Group: Technical Patterns — Reversals — educational description of double tops and bottoms as reversal structures.
TradingView: Chart Pattern Double Bottom — two-low structure, intermediate-high neckline, permissible deviation, close-based formed status, failed and indeterminable statuses, and projected-target disclaimer.
OANDA: Double Top and Double Bottom Patterns — neckline convention, false-break awareness, retest context, and educational-risk boundary.
Interactive Brokers Campus: Double Bottom — broad bullish reversal terminology.
Lo, Mamaysky, and Wang: Foundations of Technical Analysis — systematic pattern-recognition methodology, subjectivity problem, and sample-dependent empirical evidence involving double bottoms.
CFTC: Commodity Trading Systems Sold on the Internet — limitations of hypothetical performance results and risks of unsupported system-performance claims.

These sources describe conventions and research boundaries. They do not establish one universal low tolerance, confirmation rule, success rate, optimal timeframe, or guaranteed target.

Double Bottom Pattern: Definition, Neckline Confirmation, and Failure Rules