Fibonacci Corrections: Depth and Reaction Testing

A Fibonacci retracement cannot predict where a market correction will stop. What it can do is convert a frozen directional price leg into a consistent set of percentage references. Those references make it easier to measure how deep a countertrend move has become and to test whether different kinds of reactions occurred near predeclared levels.

The central question is not:

Which Fibonacci level will hold?

It is:

Given one fixed leg, one fixed anchor version, and one fixed reaction rule, what happened as the correction moved through each depth range?

Key takeaways:

Separate the original directional leg from the later correction.
Freeze endpoints before revealing the outcome whenever possible.
Measure correction depth continuously rather than treating 38.2%, 50%, or 61.8% as magic barriers.
Define approach, touch, rejection, acceptance, continuation, failure, invalidation, expiry, and ambiguity before testing.
Distinguish a completed reference leg from a developing leg whose endpoint can still move.
Preserve corrections that exceed 100%, never reach a listed level, or remain unresolved.
Keep chart behavior separate from trade execution and profitability claims.

For exact anchor selection, upward/downward price calculations, wick-versus-close variants, tolerance zones, and drawing versioning, use the Fibonacci retracement anchor and level rules. For the broad introduction, common levels, and general strategy families, use the Fibonacci retracement trading guide. This page owns the narrower correction-depth and reaction-testing workflow.

What Counts as a Correction?

A correction is a move against a previously defined directional leg. It is a relative label, not a guarantee that the earlier direction will resume.

For an upward reference leg:

the leg runs from a frozen low to a frozen high;
the correction is the subsequent movement downward from that high;
correction depth increases as price moves back toward the starting low.

For a downward reference leg:

the leg runs from a frozen high to a frozen low;
the correction is the subsequent movement upward from that low;
correction depth increases as price moves back toward the starting high.

A correction can later be classified as:

a shallow pullback followed by continuation;
a deep pullback followed by continuation;
a full retracement of the reference leg;
a reversal beyond the origin;
a sideways pause that never reaches a selected level;
an unresolved move when the evaluation horizon ends.

Do not label every countertrend candle as a separate correction. Freeze the minimum structure needed to define the move:

reference leg identity;
correction start timestamp;
correction endpoint policy;
maximum allowed gap between leg completion and correction start;
whether sideways consolidation belongs to the correction;
whether a new directional extreme resets the reference leg.

The market structure guide owns the broader definition of swing highs, swing lows, trends, ranges, and transitions. Fibonacci depth should be applied only after the structure version is defined.

Measure Correction Depth as a Continuous Variable

The familiar horizontal lines are discrete labels on a continuous measurement.

For an upward leg from low L to high H, with current correction price P:

Correction depth = (H − P) / (H − L)

For a downward leg from high H to low L, with current correction price P:

Correction depth = (P − L) / (H − L)

The result can be expressed as a decimal or percentage.

Example: upward leg

Suppose a frozen leg moves from 100 to 140.

Current correction price	Depth calculation	Correction depth
132	`(140 − 132) / 40`	20%
124	`(140 − 124) / 40`	40%
120	`(140 − 120) / 40`	50%
112	`(140 − 112) / 40`	70%
100	`(140 − 100) / 40`	100%
96	`(140 − 96) / 40`	110%

This table does not say where price should reverse. It only describes how much of the reference leg has been retraced.

Why continuous depth matters

A candle that reaches 60.9% and one that reaches 62.1% should not automatically be treated as completely different market events merely because one falls on each side of a 61.8% line.

Store the actual depth along with the named level or band. That enables later analysis of:

maximum correction depth;
depth at first reaction;
depth at first closed-bar confirmation;
depth before continuation;
depth before full invalidation;
distribution of outcomes by depth bins;
sensitivity to line-versus-zone rules.

Fibonacci Levels Are Reference Labels

TradingView describes Fibonacci retracement as a drawing tool that places percentage levels between two selected extreme points. Fidelity similarly presents the levels as possible support and resistance areas. Neither description proves that a specific ratio must stop price.

Common displayed levels include:

0%;
23.6%;
38.2%;
50%;
61.8%;
78.6%;
100%.

Platforms may also include:

14.6%;
65%;
70.5%;
76.4%;
88.6%;
custom levels;
extensions beyond 100%.

The 50% level is widely displayed even though it is not derived from the Fibonacci sequence in the same way as 38.2% or 61.8%. Treat every enabled level as part of the drawing version.

Do not rank levels by folklore

Avoid universal statements such as:

38.2% is always a healthy pullback;
50% is fair value;
61.8% is the last line of defense;
78.6% means the trend is over;
the golden zone is the best entry area.

These labels may be used as hypotheses, but they need independent rules and evidence.

A proper test asks:

How often did the correction enter the band?
What was the first closed-bar state after entry?
How often did price continue in the original direction within the evaluation horizon?
How often did price cross the leg origin?
Did the result survive another market, period, provider, and anchor version?

Freeze the Reference Leg

The correction-depth number is only as stable as the leg being measured.

Record:

symbol and provider;
chart type;
interval;
session and timezone;
adjusted or unadjusted history;
futures contract and roll method;
reference direction;
start anchor;
end anchor;
anchor method;
endpoint confirmation timestamp;
drawing version ID.

Completed leg

A completed leg uses an endpoint that was knowable under the chosen confirmation rule.

Examples:

confirmed swing high after a fixed number of bars;
highest high before a defined structure break;
event-window high at a fixed end timestamp;
session high after the session closes;
fixed-range endpoint selected by a rule.

Developing leg

A developing leg uses the latest extreme while the directional move is still unfolding. Every new high in an upward leg or new low in a downward leg changes the denominator and moves all retracement levels.

A developing-level study must retain timestamped versions. It cannot use the final high or low from the completed chart to judge decisions made before that extreme existed.

Re-anchoring policy

Define when the drawing may be replaced:

new directional extreme before the correction begins;
new extreme during a correction;
confirmed continuation after a partial pullback;
full retracement beyond the origin;
session or calendar boundary;
structure transition;
fixed maximum age.

Never silently drag the endpoint after seeing which version produces the cleanest reaction.

Line, Zone, and Depth-Bin Versions

A Fibonacci reaction can be tested with three different geometries.

Exact line

A line version requires:

exact calculated price;
tick-size rounding;
whether bid, ask, last trade, high-low range, or close counts;
tolerance for floating-point differences;
first-touch versus all-touch policy.

Tolerance zone

A zone version requires a width rule such as:

fixed number of ticks;
fixed percentage of price;
percentage of the reference leg;
ATR-normalized width;
half-distance to neighboring enabled levels;
platform display precision.

Record whether the level is at the center, upper edge, or lower edge of the zone.

Depth bin

A bin version groups continuous depth into ranges, for example:

below 23.6%;
23.6% to below 38.2%;
38.2% to below 50%;
50% to below 61.8%;
61.8% to below 78.6%;
78.6% to below 100%;
100% or deeper.

This avoids pretending that price must react at an exact line. It is often better suited to research questions about correction severity.

Build a Correction State Machine

The state machine prevents a later bounce from being retroactively described as an obvious Fibonacci reaction.

1. Reference leg qualified

The leg meets the frozen structure, size, duration, and endpoint requirements.

2. Correction candidate

Price begins moving against the reference direction under a defined rule.

Possible definitions:

first lower close after an upward leg;
break of a minor swing;
minimum countertrend movement;
minimum number of countertrend bars;
entry into the first enabled depth band.

3. Approach

Price is near a level or band but has not entered it.

Record the approach side and distance. A level approached from above in an upward-leg correction is not the same event as a later revisit from below.

4. Touch or entry

Define whether touch means:

intrabar high-low overlap;
closing price reaches the line;
candle body enters the zone;
lower-timeframe trade reaches the level;
first entry into a depth bin.

5. Reaction candidate

A reaction candidate is an event after touch, such as:

close away from the level;
local pivot formation;
close back outside the zone;
break of a countertrend line;
break of a minor correction swing;
minimum favorable movement from the deepest point.

It is not yet continuation.

6. Continuation confirmed

Continuation must be defined independently from the Fibonacci level.

Possible versions:

original leg high or low is exceeded on a close;
price travels a fixed fraction of the reference leg in the original direction;
market structure forms a new directional extreme;
price reaches a predefined horizon without invalidation;
a higher-timeframe bar confirms the original direction after completion.

7. Acceptance through the level

A level can be classified as accepted when price:

closes beyond it;
spends a minimum number of bars beyond it;
reaches the next depth band;
retests from the deeper side and remains there;
exceeds a buffer beyond the level.

Acceptance through 61.8%, for example, is not the same as invalidating the entire reference leg.

8. Reaction failure

A reaction candidate fails when its predefined rejection or continuation condition is reversed.

Examples:

price closes back through the reaction zone;
the deepest correction point is exceeded;
continuation confirmation never occurs before expiry;
the reference origin is crossed;
a new structure state replaces the hypothesis.

9. Leg invalidated

A simple invalidation version marks the leg invalid when correction depth exceeds 100%.

Other versions may use:

close beyond the origin;
intrabar breach beyond the origin;
buffer beyond the origin;
two consecutive closes;
higher-timeframe confirmation.

10. Expired

A correction can expire without continuation or invalidation after:

a fixed number of bars;
a session close;
a calendar period;
a new reference leg forms;
price remains inside a range too long;
data becomes incomparable after a contract roll or corporate action.

11. Ambiguous

Use an ambiguous class when:

one OHLC bar hits both reaction and failure boundaries;
the exact intrabar sequence is unknown;
gaps skip over one or more levels;
endpoint confirmation and level touch occur on the same bar;
data corrections change the extreme;
the required lower-timeframe data is unavailable.

A Correction Can Skip Levels

Markets do not have to visit retracement levels in order.

A gap can move from a 20% correction to a 55% correction without trading at the intermediate calculated prices. A single wide bar can cross several levels.

Freeze how skipped levels are recorded:

touched because the bar range crossed them;
skipped because no trade was documented at those prices;
entered on the open at the deepest band;
ambiguous without tick or lower-timeframe data;
first eligible level only;
all crossed levels, with one shared event timestamp.

Do not count one gap as several independent Fibonacci reactions.

Maximum Depth, Closing Depth, and First-Reaction Depth

Different depth metrics answer different questions.

Metric	Definition	Use
Intrabar maximum depth	Deepest high/low penetration	Measures full excursion but can be sensitive to spikes
Closing maximum depth	Deepest closed-bar value	Reduces wick sensitivity but ignores intrabar risk
First-touch depth	Depth when the first enabled level is touched	Measures early interaction
First-reaction depth	Deepest point before the first qualified reaction	Connects the level to the initial reversal event
Pre-continuation depth	Maximum depth before continuation confirmation	Describes correction severity before resumed direction
Final evaluation depth	Depth at expiry or study end	Records unresolved cases

Store more than one metric if the research question requires it. Do not switch between wick depth and closing depth after seeing which produces better results.

Continuation Is Not the Same as Profitability

A correction can continue in the original direction and still produce a losing trade under a particular execution plan.

Examples:

price reacts but gaps through the stop before entry;
a limit order is never filled;
spread and slippage make the entry worse;
continuation occurs only after a deeper adverse excursion;
the target is too far for the evaluation horizon;
the reaction is visible only on a completed candle, delaying entry;
the position size is incompatible with the stop distance.

Separate two research layers.

Chart-behavior layer

Measure:

correction depth;
level or zone entered;
reaction timing;
continuation timing;
adverse and favorable movement;
invalidation;
expiry;
ambiguity.

Trade-simulation layer

Add:

order type;
entry timestamp;
spread;
slippage;
commissions;
stop and target;
same-bar event policy;
gaps and halts;
partial fills;
position sizing;
borrowing or funding costs.

The CFTC warns that hypothetical trading results have inherent limitations and can differ from actual execution. Do not convert a chart reaction rate into an expected account return.

Context Variables to Record

A Fibonacci depth can be interpreted differently across market conditions, but those conditions need objective labels.

Trend structure

Record whether the reference leg occurred in:

established trend;
first breakout from a range;
late-stage extended trend;
broad sideways market;
countertrend move within a higher-timeframe trend;
post-gap repricing;
event-driven spike.

Leg characteristics

Record:

leg size in price, percentage, ATR, or another normalized unit;
number of bars;
overlap between bars;
gap contribution;
volume context if available;
acceleration or deceleration rule;
distance from higher-timeframe structure.

Correction characteristics

Record:

bars to first level;
bars to maximum depth;
correction slope;
overlap and range compression;
gap presence;
retracement speed relative to the leg;
whether the correction formed a recognizable range;
whether a new event occurred during the correction.

Multi-timeframe availability

A higher-timeframe swing endpoint may not be confirmed until its bar closes and any pivot-confirmation rule completes. Do not use the final weekly or daily anchor on an earlier intraday decision unless it was already available.

Use the multiple-timeframe analysis guide for broader information-availability rules.

Correction Versus Reversal

Fibonacci depth cannot determine by itself whether a move is a temporary correction or a lasting reversal.

A study needs an external classification rule, such as:

reference origin breached;
prior major swing broken;
higher-timeframe structure changed;
continuation failed within a fixed horizon;
trend definition changed;
a new opposite-direction leg qualified.

Avoid circular definitions such as:

It was a correction because price continued.

and then claiming:

Fibonacci predicted the continuation.

Instead, define the correction candidate first and classify the later outcome independently.

The pullback trading guide owns the broader pullback-versus-reversal framework and trade-planning context.

Fibonacci and Support or Resistance

A Fibonacci level is calculated from two selected anchors. It is not the same as a raw historical price level.

Keep separate labels for:

Fibonacci calculated line;
swing high or low;
prior range boundary;
repeated closing level;
gap edge;
VWAP or moving average;
supply or demand zone;
volume-profile level.

When several references are near each other, define the distance threshold before calling them confluence. Overlapping derived levels are not automatically independent evidence.

For raw price-level identification, use the support and resistance checklist. For later approach, touch, rejection, breakout, and retest decisions, use the support and resistance trading framework.

Robustness Tests

A Fibonacci correction rule should not depend on one precise chart configuration.

Test reasonable variants of:

wick, body, close, and confirmed-pivot anchors;
developing versus completed endpoint;
line versus tolerance zone;
continuous depth versus named levels;
intrabar versus closing depth;
one market versus several markets;
one interval versus several intervals;
regular versus extended session;
adjusted versus unadjusted stock data;
individual versus continuous futures contracts;
spot versus derivative crypto data;
development, validation, and final evaluation periods.

A result that disappears after a small change in anchor, zone width, or sample may be unstable.

Reproducible Correction Worksheet

Use one row per reference leg and correction.

Field	What to record
Version ID	Immutable rule name
Symbol/provider	Exact market and feed
Chart identity	Interval, session, timezone, chart type
Adjustment policy	Stock adjustment or futures roll method
Reference direction	Upward or downward leg
Start/end anchors	Exact price and timestamp
Anchor method	Wick, close, body, confirmed pivot, fixed event
Endpoint availability	Timestamp when the endpoint became usable
Leg size	Price, percentage, ATR or normalized measure
Correction start	Exact rule and timestamp
Level set	Enabled ratios and custom levels
Geometry	Line, zone or depth bin
Zone width	Exact formula
Depth metric	Intrabar, close, first reaction, pre-continuation
Approach	Distance rule
Touch	Exact interaction rule
Reaction	Exact independent event
Acceptance	Exact rule for moving through a level
Continuation	Exact original-direction confirmation
Failure	Exact reversal of the candidate
Invalidation	Rule at or beyond the leg origin
Expiry	Maximum bars, session, structure or calendar rule
Ambiguity	Same-bar, gap and missing-data policy
Context	Trend, range, event, gap, volatility
Maximum depth	Continuous value and deepest band
Outcome	Continued, invalidated, expired, ambiguous, unresolved
Execution layer	None or separately versioned simulation assumptions

Replay Workflow

Select one chart identity and one correction-rule version.
Freeze the reference leg and confirm that both anchors were available.
Calculate or export the selected Fibonacci levels externally.
Hide future candles.
Record the reference leg, levels, and correction state before advancing.
Move forward one bar at a time.
Update continuous correction depth without moving the original anchors unless the re-anchoring rule triggers.
Record approach, touch, reaction, acceptance, continuation, failure, invalidation, expiry, or ambiguity.
Preserve every non-reaction and unresolved correction.
Review results only after completing the frozen sample.

The general backtesting guide owns sample design, overfitting, cost assumptions, and broader evaluation. The pattern-recognition replay guide covers deliberate chart-reading practice.

Common Errors

Saying Fibonacci predicts corrections

The tool measures a selected price range. It does not identify the future endpoint of a correction.

Choosing anchors after seeing the bounce

This changes every percentage level and creates hindsight fit.

Treating 61.8% as universally strongest

A named ratio has no universal reliability without a defined sample and rule.

Calling a wick a successful reaction

A wick is only a reaction if the test version defines it that way.

Ignoring price beyond 100%

Full retracements and reversals must remain in the dataset.

Redrawing after each new correction low or high

That changes the denominator and makes earlier states incomparable.

Counting several nearby lines as independent confirmation

Fibonacci ratios from the same anchors are mathematically related.

Mixing retracements and extensions

Retracement levels lie within or describe the original leg; extensions use a different construction and research question.

Using final higher-timeframe anchors too early

This leaks information unavailable at the historical decision time.

Optimizing zone width on the same sample

A wide zone can make nearly any nearby reaction appear successful.

Excluding corrections that never bounce

Failed and unresolved cases are necessary for a credible denominator.

Confusing chart continuation with trade returns

Execution, costs, stops, fills, and timing can produce different results.

What ChartMini Can and Cannot Do

ChartMini can support lightweight candle-by-candle replay after the analyst records externally calculated reference levels.

ChartMini does not currently:

draw Fibonacci retracements or extensions;
select swing anchors;
calculate correction percentages;
preserve developing Fibonacci drawings automatically;
import TradingView or broker drawing objects;
reconstruct intrabar order from OHLC bars;
model spreads, slippage, commissions, gaps, halts, or partial fills;
predict whether a correction will end;
prove that a Fibonacci strategy is profitable.

Use ChartMini for hidden-future observation and rule logging, not as evidence that the Fibonacci tool itself forecasts price.

Practical Next Steps

Choose one reference-leg and endpoint version.
Use the anchor guide to calculate the levels consistently.
Store continuous correction depth, not only the nearest named ratio.
Define line, zone, or depth-bin geometry.
Write reaction, continuation, failure, invalidation, expiry, and ambiguity rules.
Test completed and developing endpoints separately.
Keep full retracements, skipped levels, gaps, and unresolved cases.
Separate chart outcomes from trade simulation.
Recheck the result across reasonable anchors, markets, intervals, and samples.

Frequently Asked Questions

Can Fibonacci retracement predict where a correction will end?

No. Fibonacci retracement converts a frozen price leg into reference percentages. It can help describe correction depth and organize candidate reaction zones, but it cannot determine in advance which level will hold or whether the original trend will continue.

How do you measure the depth of a market correction?

Freeze the directional leg and its two endpoints, then express the countertrend move as a percentage of that leg. The result depends on the chosen anchors, wick-versus-close policy, endpoint confirmation, chart identity, and the timestamp at which the measurement is made.

Is a 61.8 percent retracement stronger than other levels?

Not universally. A 61.8 percent line is one calculated reference among several. Its usefulness depends on the market, sample, swing definition, context, confirmation rule, failure rule, and whether the same method survives out-of-sample testing.

What happens when price retraces more than 100 percent?

A move beyond the origin of the measured leg means the original leg has been fully retraced under that anchor version. The drawing can be marked invalidated, completed, or replaced according to a predeclared rule rather than silently re-anchored after the fact.

Should Fibonacci reactions use a line or a zone?

Either can be tested, but the choice must be fixed before reviewing outcomes. A line requires an exact touch rule, while a zone requires a width formula, entry boundary, exit boundary, and overlap policy. Widening a zone after seeing price react creates hindsight bias.

Can ChartMini draw Fibonacci retracement levels?

No. ChartMini is a lightweight candle-by-candle replay tool and does not calculate or draw Fibonacci retracement or extension levels. You can record externally calculated levels before replay and then classify later price interactions without seeing future candles.

Fibonacci Retracement Corrections: Measure Depth and Test Reactions