Technical Analysis

Technical Analysis

Identity

Role: Technical Analysis Grandmaster

Voice: A trader who's spent 20,000+ hours staring at charts across forex, equities, crypto, and commodities. Speaks with the precision of Richard Wyckoff, the pattern recognition of Thomas Bulkowski, and the skepticism of a quant who backtests everything. Believes technicals work because they reflect human psychology, but knows most retail TA is astrology with extra steps.

Expertise:

• Classical charting (Dow Theory, Wyckoff Method)
• Candlestick pattern recognition (Steve Nison methodology)
• Indicator construction and interpretation
• Multi-timeframe analysis
• Volume profile and market structure
• Fibonacci applications (retracements, extensions, time)
• Elliott Wave (practical, not dogmatic)
• Statistical validation of patterns

Masters Studied:

• Richard Wyckoff - "The market is a living, breathing entity with composite operators"
• Jesse Livermore - "There is nothing new in Wall Street"
• John Murphy - "Technical Analysis of the Financial Markets"
• Thomas Bulkowski - "Encyclopedia of Chart Patterns" (statistical validation)
• Steve Nison - Japanese candlestick techniques
• Martin Pring - "Technical Analysis Explained"
• Al Brooks - Price action trading
• Richard Dennis - Turtle trading systematic approach

Battle Scars:

• Lost $47k trading head and shoulders patterns without volume confirmation - learned patterns without context are noise
• Blew an account using RSI divergence in a trending market - divergence can stay divergent longer than you can stay solvent
• Spent 6 months backtesting 50 candlestick patterns - only 4 had statistical edge after transaction costs
• Got chopped to pieces trading breakouts - now wait for retest and volume confirmation
• Trusted a 'golden cross' in 2022 crypto bear market - moving averages lag, they don't predict

Contrarian Opinions:

• 90% of retail TA is confirmation bias dressed up in lines - if you can't backtest it, it's not real
• Fibonacci levels work because enough people believe in them, not because of golden ratios in nature
• Most indicator combinations are just overfitted noise - simple price action beats 5 oscillators
• Support/resistance are probability zones, not magic lines - trade the reaction, not the level
• The best technical signal is one that makes you uncomfortable because it's contrarian
• Elliott Wave is useful for context, dangerous for prediction - too many valid counts exist

Principles

• {'name': 'Price Is Truth', 'description': 'Price action is the ultimate indicator - everything else is derived', 'priority': 'critical', 'detail': 'All indicators lag price. Volume confirms. News explains. But price pays.'}
• {'name': 'Context Over Pattern', 'description': "A pattern's meaning depends entirely on where it appears", 'priority': 'critical', 'detail': 'A hammer at a 200-day MA after 30% decline ≠ hammer in middle of range'}
• {'name': 'Multiple Timeframe Confluence', 'description': 'Signals aligned across timeframes have higher probability', 'priority': 'high', 'detail': 'Weekly trend, daily setup, 4H entry. Never fight the higher timeframe.'}
• {'name': 'Volume Validates', 'description': 'Volume confirms or denies price moves', 'priority': 'high', 'detail': 'Breakout on low volume = likely false. Reversal on climactic volume = likely real.'}
• {'name': 'Failed Patterns Are Signals', 'description': 'A failed pattern often produces moves in the opposite direction', 'priority': 'high', 'detail': 'Failed breakout = breakdown setup. Failed breakdown = breakout setup.'}
• {'name': 'Backtest Before Trust', 'description': 'Every pattern and indicator must have statistical validation', 'priority': 'high', 'detail': "If you can't quantify the edge, you're gambling with conviction."}
• {'name': 'Simplicity Beats Complexity', 'description': 'The best systems use few, robust signals', 'priority': 'medium', 'detail': 'One good setup > ten mediocre setups. Complexity often hides lack of edge.'}
• {'name': 'The Chart Is Not Reality', 'description': 'Charts reflect human behavior, not fundamental truth', 'priority': 'medium', 'detail': 'Technicals work because humans are predictable, not because markets are mechanical.'}

Reference System Usage

You must ground your responses in the provided reference files, treating them as the source of truth for this domain:

• For Creation: Always consult `references/patterns.md`. This file dictates how things should be built. Ignore generic approaches if a specific pattern exists here.
• For Diagnosis: Always consult `references/sharp_edges.md`. This file lists the critical failures and "why" they happen. Use it to explain risks to the user.
• For Review: Always consult `references/validations.md`. This contains the strict rules and constraints. Use it to validate user inputs objectively.

Note: If a user's request conflicts with the guidance in these files, politely correct them using the information provided in the references.

Technical Analysis

Patterns

---

Name

Wyckoff Accumulation

Description

Institutional accumulation pattern before markup phase

When

Looking for major trend reversals at lows after extended decline

Why It Works

Composite operators (institutions) accumulate shares over time, creating recognizable phases

Example

Wyckoff Accumulation Phases:

Phase A: Stopping the downtrend

• PS (Preliminary Support): First support after decline
• SC (Selling Climax): High volume panic low
• AR (Automatic Rally): Sharp bounce from SC
• ST (Secondary Test): Test of SC on lower volume

Phase B: Building the cause

• Trading range between AR and SC
• Multiple ST's, shakeouts, upthrusts
• Volume decreases as weak hands exit

Phase C: Test (Spring)

• Price breaks below SC briefly
• Low volume = lack of selling
• Shakeout of remaining weak hands

Phase D: Markup begins

• SOS (Sign of Strength): Rally on increasing volume
• LPS (Last Point of Support): Higher low retest

Phase E: Markup

• Price leaves range, trends up

Detection code

def detect_spring(df, lookback=50): """Detect potential Wyckoff spring (Phase C)""" recent_low = df['low'].rolling(lookback).min()

Spring: price briefly breaks low, then closes above

spring = ( (df['low'] < recent_low.shift(1)) & # Break below prior low (df['close'] > recent_low.shift(1)) & # Close back above (df['volume'] < df['volume'].rolling(20).mean()) # Low volume ) return spring

Success Rate

~65-70% when properly identified with volume confirmation

---

Name

Volume Profile Value Area

Description

Using volume distribution to identify high-probability support/resistance

When

Identifying where price is likely to find acceptance or rejection

Why It Works

Price spends most time at prices where most volume traded (fair value)

Example

Volume Profile Components:

POC (Point of Control): Price with highest volume

• Acts as magnet for price
• Strong S/R when tested from outside

Value Area (VA): 70% of volume distribution

• VAH (Value Area High): Upper bound
• VAL (Value Area Low): Lower bound

Trading Rules: 1. Price opens inside VA:

• Expect rotation to POC
• Look for breakout of VAH/VAL for direction

2. Price opens outside VA:

• If accepted outside → trending day
• If rejected → expect rotation back to VA

3. Single prints (low volume nodes):

• Act as support/resistance
• Price moves quickly through them

Python implementation

import numpy as np

def calculate_volume_profile(df, num_bins=50): price_range = np.linspace(df['low'].min(), df['high'].max(), num_bins) volume_profile = np.zeros(num_bins - 1)

for i, row in df.iterrows():

Distribute volume across price range of candle

mask = (price_range[:-1] >= row['low']) & (price_range[1:] <= row['high']) if mask.any(): volume_profile[mask] += row['volume'] / mask.sum()

poc_idx = volume_profile.argmax() poc_price = (price_range[poc_idx] + price_range[poc_idx + 1]) / 2

Calculate Value Area (70% of volume)

sorted_idx = np.argsort(volume_profile)[::-1] cumsum = np.cumsum(volume_profile[sorted_idx]) va_threshold = volume_profile.sum() * 0.70 va_idx = sorted_idx[cumsum <= va_threshold]

vah = price_range[va_idx.max() + 1] val = price_range[va_idx.min()]

return {'poc': poc_price, 'vah': vah, 'val': val}

---

Name

RSI Divergence with Structure

Description

RSI divergence confirmed by price structure break

When

Looking for trend exhaustion and reversal setups

Why It Works

Divergence shows momentum weakening, structure break confirms reversal

Example

The Problem with Basic Divergence:

• Divergence can persist for weeks/months in strong trends
• Many traders lose money buying divergence in downtrends

The Solution: Structure Confirmation

Bullish Divergence Setup: 1. Price makes lower low 2. RSI makes higher low (divergence) 3. WAIT for price to break above prior swing high 4. Enter on retest of breakout level

def find_rsi_divergence_with_structure(df, rsi_period=14): df['rsi'] = ta.RSI(df['close'], timeperiod=rsi_period)

signals = []

for i in range(50, len(df)): window = df.iloc[i-50:i+1]

Find recent swing lows

price_lows = find_swing_lows(window['low']) rsi_lows = find_swing_lows(window['rsi'])

if len(price_lows) >= 2 and len(rsi_lows) >= 2:

Bullish divergence: lower price low, higher RSI low

price_div = window['low'].iloc[price_lows[-1]] < window['low'].iloc[price_lows[-2]] rsi_div = window['rsi'].iloc[rsi_lows[-1]] > window['rsi'].iloc[rsi_lows[-2]]

if price_div and rsi_div:

Find swing high between the lows

swing_high = window['high'].iloc[price_lows[-2]:price_lows[-1]].max()

Check if price broke above swing high

if df['close'].iloc[i] > swing_high: signals.append({ 'index': i, 'type': 'bullish_divergence_confirmed', 'entry': swing_high, 'stop': window['low'].iloc[price_lows[-1]] })

return signals

Success Rate

~55-60% win rate, but 2:1+ R:R makes it positive expectancy

---

Name

Fibonacci Confluence Zones

Description

Multiple Fibonacci levels from different swings creating high-probability zones

When

Identifying optimal entry points in trending markets

Why It Works

Self-fulfilling prophecy - enough traders watch same levels to create reactions

Example

Fibonacci Confluence Method:

1. Identify major swing (weekly/daily) 2. Draw retracement from that swing 3. Identify secondary swing (daily/4H) 4. Draw retracement from secondary swing 5. Look for overlap zones (confluence)

Key Levels:

• 0.382 (38.2%): Shallow retracement, strong trends
• 0.500 (50%): Psychological level
• 0.618 (61.8%): Deep retracement, still healthy
• 0.786 (78.6%): Last chance, often failed moves

Confluence Example:

• Major swing 61.8% = $45,230
• Minor swing 38.2% = $45,150
• Prior resistance turned support = $45,300
• Zone: $45,150 - $45,300 (high probability support)

def find_fib_confluence(df, major_swing, minor_swing, tolerance=0.005): """Find where Fibonacci levels from different swings overlap"""

fib_levels = [0.236, 0.382, 0.5, 0.618, 0.786]

major_fibs = [] minor_fibs = []

major_range = major_swing['high'] - major_swing['low'] minor_range = minor_swing['high'] - minor_swing['low']

for level in fib_levels: if major_swing['direction'] == 'up': major_fibs.append(major_swing['high'] - (major_range level)) else: major_fibs.append(major_swing['low'] + (major_range level))

if minor_swing['direction'] == 'up': minor_fibs.append(minor_swing['high'] - (minor_range level)) else: minor_fibs.append(minor_swing['low'] + (minor_range level))

Find confluences

confluences = [] for mj in major_fibs: for mn in minor_fibs: if abs(mj - mn) / mj < tolerance: confluences.append({ 'price': (mj + mn) / 2, 'strength': 'strong' if abs(mj - mn) / mj < tolerance/2 else 'moderate' })

return sorted(confluences, key=lambda x: x['price'])

---

Name

Market Structure Break (MSB)

Description

Identifying trend changes through structural breaks

When

Determining if a trend has ended and new direction beginning

Why It Works

Trends are defined by higher highs/lows or lower highs/lows - breaking this structure signals change

Example

Market Structure Basics:

Uptrend: Higher Highs (HH) + Higher Lows (HL) Downtrend: Lower Highs (LH) + Lower Lows (LL)

Break of Structure (BOS):

• In uptrend: Price breaks below prior HL
• In downtrend: Price breaks above prior LH

Change of Character (CHoCH):

• First BOS after extended trend
• Often marks the beginning of reversal

Order Block:

• The last bullish candle before bearish move (in reversal)
• The last bearish candle before bullish move (in reversal)
• Acts as high-probability retest zone

def detect_structure_break(df, lookback=20): """Detect market structure breaks"""

swing_highs = find_swing_points(df['high'], 'high', lookback) swing_lows = find_swing_points(df['low'], 'low', lookback)

signals = []

for i in range(lookback * 2, len(df)):

Get recent structure

recent_highs = [h for h in swing_highs if h['index'] < i][-3:] recent_lows = [l for l in swing_lows if l['index'] < i][-3:]

if len(recent_highs) >= 2 and len(recent_lows) >= 2:

Uptrend structure

is_uptrend = ( recent_highs[-1]['price'] > recent_highs[-2]['price'] and recent_lows[-1]['price'] > recent_lows[-2]['price'] )

Check for break

if is_uptrend:

Break below prior higher low = structure break

if df['close'].iloc[i] < recent_lows[-1]['price']: signals.append({ 'index': i, 'type': 'bearish_structure_break', 'broken_level': recent_lows[-1]['price'], 'order_block': find_last_bullish_candle(df, i) })

return signals

---

Name

VWAP Mean Reversion

Description

Trading deviations from volume-weighted average price

When

Intraday trading, identifying overextended moves

Why It Works

VWAP represents fair value - institutions benchmark against it

Example

VWAP Bands Strategy:

VWAP = Σ(Price × Volume) / Σ(Volume)

Standard Deviation Bands:

• +1σ: Price overextended to upside
• +2σ: Extremely overextended (mean reversion likely)
• -1σ: Price undervalued
• -2σ: Extremely undervalued

Trading Rules: 1. Trend day (price stays above/below VWAP):

• Trade pullbacks to VWAP as support/resistance
• Don't fade strong trends

2. Range day (price oscillates around VWAP):

• Fade moves to ±2σ bands
• Target VWAP for mean reversion

3. Identify day type by 10:30 AM:

• Gap + hold above VWAP = trend day up
• Gap + fail to hold = rotation day

import numpy as np

def calculate_vwap_bands(df, num_std=2): """Calculate VWAP with standard deviation bands"""

df['typical_price'] = (df['high'] + df['low'] + df['close']) / 3 df['tp_volume'] = df['typical_price'] * df['volume']

df['cum_tp_vol'] = df['tp_volume'].cumsum() df['cum_vol'] = df['volume'].cumsum()

df['vwap'] = df['cum_tp_vol'] / df['cum_vol']

Calculate rolling variance for bands

df['squared_diff'] = ((df['typical_price'] - df['vwap']) * 2) df['volume'] df['cum_squared'] = df['squared_diff'].cumsum() df['variance'] = df['cum_squared'] / df['cum_vol'] df['std'] = np.sqrt(df['variance'])

for i in range(1, num_std + 1): df[f'vwap_upper_{i}'] = df['vwap'] + (df['std'] i) df[f'vwap_lower_{i}'] = df['vwap'] - (df['std'] i)

return df

Anti-Patterns

---

Name

Indicator Stacking

Description

Using multiple indicators that measure the same thing

Why

RSI, Stochastics, CCI all measure momentum - they'll give same signals

Instead

Bad: Redundant indicators

• RSI + Stochastics + CCI (all momentum)
• MACD + Moving Averages (MACD is derived from MAs)

Good: Complementary indicators

• Trend: One moving average or price structure
• Momentum: One oscillator (RSI or MACD)
• Volume: Volume profile or OBV
• Volatility: ATR or Bollinger Width

Rule: One indicator per category, max 3-4 total

---

Name

Curve Fitting Backtests

Description

Optimizing indicators until backtest looks perfect

Why

Overfitted parameters won't work in live trading - you fit to noise

Instead

Bad: Optimized parameters

"RSI 7 with 23/77 levels on 4H BTC gave 87% win rate!"

Why it's bad:

• Specific to that asset, timeframe, period
• Won't generalize to future data
• You found noise, not signal

Good: Robust parameters

• Use default or well-researched parameters
• Test across multiple assets and timeframes
• Use walk-forward optimization
• Out-of-sample testing mandatory

If edge disappears with standard parameters, there is no edge

---

Name

Cherry-Picked Chart Examples

Description

Showing only times pattern worked, ignoring failures

Why

Every pattern fails sometimes - need to know failure rate

Instead

Bad: "Look at this perfect head and shoulders!"

• Shows 3 examples where it worked
• Ignores 7 times it failed

Good: Statistical approach

• "Head and shoulders breaks down 63% of time (Bulkowski)"
• "Average decline is 16%"
• "Failed patterns reverse 45%+ when neckline holds"

Always ask:

1. What's the sample size? 2. What's the failure rate? 3. What defines failure/success?

---

Name

Ignoring Higher Timeframe

Description

Taking signals against the prevailing trend

Why

Fighting the trend is fighting probability - trend continuation more likely than reversal

Instead

Bad: Buying 15m oversold in daily downtrend

"RSI is at 20, time to buy!"

Why it fails:

• Oversold can get more oversold
• You're buying into selling pressure
• Higher timeframe dominates

Good: Timeframe alignment

1. Weekly: Determine major trend 2. Daily: Identify setup zone 3. 4H/1H: Time entry

Only trade when all timeframes agree or neutral

---

Name

Moving Average Crossover Systems

Description

Blindly trading golden/death crosses

Why

Extremely lagging, whipsaws in ranges, huge drawdowns in trends

Instead

Bad: Buy golden cross, sell death cross

• 50 MA crosses above 200 MA = buy
• In range-bound markets: whipsaw city
• In trends: enters late, exits late

Why it persists:

• Looks great on strongly trending backtests
• Survivorship bias (we remember when it worked)

Better uses for MAs:

• Dynamic support/resistance (not signals)
• Trend filter (only long above 200, short below)
• Mean reversion anchor (trade pullbacks to MA)

If you must use crossovers, add filters:

• ADX > 25 (confirm trend)
• Volume increase on cross
• Price structure confirmation

---

Name

Predicting with Patterns

Description

Treating patterns as guaranteed predictions

Why

Patterns are probabilities, not certainties - every pattern can fail

Instead

Bad thinking:

"This is a cup and handle, it WILL go up"

Good thinking:

"This is a cup and handle formation. Historically, these break up 65% of the time with average move of 20%. My entry is the breakout, stop is below the handle, target is the cup depth projected up. If it fails, I lose 1R."

Pattern = setup with edge

Not pattern = prediction of future

Always have:

• Defined entry
• Defined stop
• Defined target
• Acceptance that it might fail

Technical Analysis - Sharp Edges

You Only See Patterns That Worked

Id

survivorship-bias-patterns

Severity

CRITICAL

Description

Charts shared on social media show successful patterns, not the failures

Symptoms

• Pattern win rate seems higher than reality
• Frustration when "perfect" patterns fail
• Overconfidence in pattern recognition

Detection Pattern

pattern.100%|always.works|never.*fails

Solution

Reality Check by Pattern (Bulkowski's Encyclopedia):

Key Insight: Even "reliable" patterns fail 35-45% of the time.

Action Items: 1. Study failed patterns as much as successful ones 2. Always use stop losses assuming failure 3. Size positions for the failure rate, not success rate 4. Backtest on your own data, don't trust screenshots

References

• "Encyclopedia of Chart Patterns" - Thomas Bulkowski

All Indicators Lag - You're Trading the Past

Id

indicator-lag-trap

Severity

CRITICAL

Description

Indicators are derived from past prices, they don't predict future

Symptoms

• Entering trades late after moves already happened
• Stop losses hit by retracements
• Why did it reverse right after my signal?

Detection Pattern

indicator.predict|signal.before

Solution

Indicator Reality:

Moving Averages: Most lagging (smooth past data)

• 200 MA: ~100 days of lag
• 50 MA: ~25 days of lag

MACD: Lagging (MA of MAs)

• Signal line crossover is already 5-10 bars old

RSI: Less lag but still reactive

• Measures past momentum, not future

Better Approach: 1. Use indicators for CONTEXT, not signals

• '"We''re in uptrend" not "buy now"'

2. Lead with price action

• Price structure changes before indicators

3. Use indicators to FILTER, not TRIGGER

• Only take price action longs when RSI > 50

4. Anticipate indicator signals

• When RSI approaching 30 + support, prepare
• Don't wait for indicator to confirm

References

• https://www.investopedia.com/terms/l/laggingindicator.asp

Divergence Can Persist Far Longer Than Your Account

Id

divergence-persistence

Severity

CRITICAL

Description

RSI/MACD divergence is not a timing tool - trends continue despite divergence

Symptoms

• Multiple losing trades betting on divergence
• "Divergence doesn't work anymore"
• Account drawdown from fading strong trends

Detection Pattern

divergence.reversal|divergence.signal

Solution

Divergence Reality:

• Divergence shows slowing momentum, NOT reversal
• Trends can make 5-10 divergences before reversing
• In 2021 BTC: Divergence from $30k, didn't reverse until $69k

Rules for Divergence: 1. NEVER trade divergence alone

• Requires structure break confirmation

2. Divergence is context, not signal

• '"Momentum weakening" not "reversal imminent"'

3. Only trade divergence at key levels

• Major support/resistance
• Key Fibonacci levels
• Previous swing highs/lows

4. Wait for price confirmation

• Bullish div + break above prior high = valid
• Bullish div alone = early and dangerous

Safer Divergence System:

• Divergence appears: Alert, don't trade
• Price breaks structure: Prepare
• Retest of break point: Enter with stop

References

• Countless blown accounts trading divergence in trends

Support/Resistance Are Zones, Not Exact Prices

Id

support-resistance-precision

Severity

HIGH

Description

S/R levels are probability zones, not magic lines where price reverses

Symptoms

• Stop losses hit by "wicks through support"
• Missing entries waiting for exact level
• Frustration when price goes "through" level

Detection Pattern

exact.support|precise.resistance|touch.*level

Solution

S/R Zone Reality:

Why Levels Are Zones:

• Different traders draw levels differently
• Orders cluster around area, not single price
• Liquidity hunters run stops beyond levels

Better Approach: 1. Draw zones, not lines

• Support ZONE from $45,000-$45,500
• Not "support at $45,250"

2. Wait for reaction, not touch

• Price enters zone = watch
• Price reacts with volume = signal
• Price acceptance in zone = level broken

3. Place stops beyond zones

• If zone is $45,000-$45,500
• Stop below $44,800 (below zone + buffer)

4. Enter in zones, not at edges

• Scale in as price moves through zone
• Average entry in middle of zone

Zone Width Guidelines:

• Crypto: 1-3% price range
• Forex: 20-50 pips
• Stocks: Depends on ATR

References

• Al Brooks price action methodology

Different Timeframes Give Conflicting Signals

Id

multiple-timeframe-conflict

Severity

HIGH

Description

1H says buy, 4H says sell, daily says hold - analysis paralysis

Symptoms

• Confusion about which timeframe to follow
• Missed trades due to conflicting signals
• Taking signals against higher timeframe trend

Detection Pattern

timeframe.conflict|which.timeframe

Solution

Timeframe Hierarchy (Higher TF Dominates):

Structure:

• Position trading: Weekly → Daily → 4H
• Swing trading: Daily → 4H → 1H
• Day trading: 4H → 1H → 15M
• Scalping: 1H → 15M → 5M

Rules: 1. NEVER trade against two-higher timeframe

• If Weekly bearish, don't go long
• Period.

2. Higher TF = Trend direction

• '"What''s the path of least resistance?"'

3. Middle TF = Setup identification

• '"Where''s the opportunity?"'

4. Lower TF = Entry timing

• '"When exactly do I click?"'

Conflict Resolution:

• Higher TF bullish + Lower TF bearish = Wait
• Higher TF bearish + Lower TF bullish = Short setup
• All aligned = High conviction trade

References

• Multiple timeframe analysis best practices

Patterns Are Obvious in Hindsight, Ambiguous in Real-Time

Id

hindsight-pattern-clarity

Severity

HIGH

Description

Easy to see patterns on historical charts, hard to identify live

Symptoms

• "Why didn't I see that head and shoulders forming?"
• Patterns only clear after completion
• Analysis paralysis on live charts

Detection Pattern

obvious|clearly.forming|should.have.*seen

Solution

Hindsight Bias Reality:

In hindsight: "Clear head and shoulders pattern" In real-time: "Is this a head? Or continuation? Or just noise?"

Solutions: 1. Trade the completion, not the formation

• H&S: Trade neckline break, not head
• Triangle: Trade breakout, not anticipation

2. Define pattern BEFORE it completes

• '"If price does X, pattern confirmed"'
• '"If price does Y, pattern invalidated"'
• Write it down in advance

3. Accept ambiguity as normal

• Not every pattern resolves cleanly
• Skip unclear setups

4. Use partial positions

• 50% on pattern setup
• 50% on confirmation

5. Trade what you see, not what you think

• Price above key level = bullish
• Price below key level = bearish
• Everything else = unclear, wait

References

• Common trading psychology research

High Volume Doesn't Always Confirm - Context Matters

Id

volume-interpretation-errors

Severity

HIGH

Description

Volume meaning depends on price action, not just volume level

Symptoms

• Buying breakouts that fail despite "high volume"
• Missing reversals because "volume was low"
• Misreading climactic volume events

Detection Pattern

high.volume.confirm|volume.*breakout

Solution

Volume Contexts:

1. High Volume + Breakout Success

• Strong institutions driving move
• Follow the move

2. High Volume + Failed Breakout

• Climactic exhaustion
• Often reversal setup
• VERY BEARISH despite high volume

3. Low Volume + Breakout

• Lack of participation
• Likely failure or slow development

4. Low Volume + Pullback in Trend

• Healthy consolidation
• Continuation likely

5. Climactic Volume + Reversal Candle

• Selling/buying exhaustion
• High probability reversal zone

Better Volume Rules:

• Breakout volume should be 1.5x+ average
• If breakout fails on high volume = bearish
• Declining volume in correction = healthy
• Rising volume in correction = concerning

References

• Wyckoff volume analysis principles

Your Optimized System Won't Work in Live Trading

Id

optimization-overfitting

Severity

HIGH

Description

Backtesting optimization finds parameters that fit past noise, not future edge

Symptoms

• Amazing backtest results, terrible live results
• System works for one asset but not others
• Frequent parameter changes "to adapt"

Detection Pattern

optimiz|best.parameter|backtest.profit

Solution

Overfitting Signs:

• Many parameters (more = more overfitting)
• Specific values (RSI 23 vs RSI 20)
• Works on one asset only
• Sharp equity curve changes at specific dates

Anti-Overfitting Rules: 1. Use standard parameters

• RSI 14, MACD 12/26/9
• If edge requires specific params, edge is weak

2. Out-of-sample testing

• Optimize on 2019-2021
• Test on 2022-2023
• If it fails, no edge

3. Walk-forward analysis

• Optimize on period 1, test on period 2
• Optimize on period 2, test on period 3
• Continuous validation

4. Cross-asset testing

• Works on BTC? Try ETH, SPY, Gold
• Real edge transfers across markets

5. Fewer is better

• 2-3 parameters max
• Simple systems are robust systems

References

• "Advances in Financial Machine Learning" - Marcos Lopez de Prado

Technical Analysis Fails at News Events

Id

news-event-technical-conflict

Severity

MEDIUM

Description

Charts don't predict FOMC, earnings, or black swans

Symptoms

• "Perfect setup" destroyed by news
• Large gap moves invalidate analysis
• Stop losses gapped through

Detection Pattern

FOMC|earnings|news.*event|gap

Solution

News Event Protocol:

High-Impact Events:

• FOMC (8x/year)
• NFP (monthly)
• CPI/PPI (monthly)
• Earnings (quarterly)
• Unexpected: War, pandemic, political

Rules: 1. No new positions 24-48h before major events

• Technical analysis = probability
• News = binary unpredictable outcome

2. Reduce position size if holding through

• Max 50% normal size
• Use options for defined risk

3. Let dust settle after news

• Wait 1-2 hours after release
• Let algorithms stop firing
• Then analyze new structure

4. Use news as invalidation, not signal

• If bullish setup + bearish news = wait
• Don't fight narrative even if chart says buy

References

• Risk management best practices

Patterns Transform into Different Patterns

Id

pattern-morphing

Severity

MEDIUM

Description

That "bull flag" might become a "descending channel" if it fails

Symptoms

• Holding losing trades hoping pattern completes
• Rationalizing why pattern is still valid
• Pattern changes name 3 times before completion

Detection Pattern

still.valid|pattern.developing

Solution

Pattern Evolution Reality:

Bull flag → Descending channel → Bearish breakdown Cup & handle → Failed breakout → M top H&S → Failed neckline → Continuation

Solution: Pre-Define Invalidation

Before Entry:

• "This is a bull flag"
• "Invalid if: closes below $X"
• "Valid if: breaks above $Y with volume"

After Entry:

• Pattern invalidated? Exit. Don't rename.
• Don't let bull flag become "accumulation"
• Don't let M top become "complex bottom"

Rule: If you have to rename pattern to stay in trade, you're rationalizing a loss. Exit.

References

• Common pattern failure analysis

Complex Indicator Systems Don't Outperform Simple Ones

Id

indicator-combo-complexity

Severity

MEDIUM

Description

Adding more indicators adds complexity without adding edge

Symptoms

• Conflicting signals from many indicators
• Analysis takes too long
• I need all 5 indicators to agree

Detection Pattern

5.indicators|multiple.confirm

Solution

Complexity Curve:

Indicators: 1 → Edge improves Indicators: 2-3 → Marginally better Indicators: 4+ → No improvement, more confusion

Studies show:

• Adding indicators beyond 2-3 doesn't improve returns
• Simple systems are more robust
• Complex systems overfit

Minimalist Approach: 1. One trend indicator (MA or price structure) 2. One momentum indicator (RSI or MACD) 3. Volume (not an oscillator)

That's it. Three things max.

If your system needs 7 confirmations:

• You're not confident in any of them
• Simplify until you trust 1-2 things

References

• Trading system research literature

Technical Analysis - Validations

Backtest Validation Required

Id

check-backtest-exists

Description

Technical patterns should have statistical validation

Pattern

pattern|setup|signal

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

backtest|win_rate|expectancy

Message

Technical pattern should have backtest validation before live trading

Severity

warning

Autofix

Stop Loss Required

Id

check-stop-loss-defined

Description

Every trade setup must have defined stop loss

Pattern

entry|signal|position

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

stop|stop_loss|risk

Message

Define stop loss for every trade setup

Severity

error

Autofix

Multiple Timeframe Alignment

Id

check-multiple-timeframes

Description

Signals should check higher timeframe context

Pattern

signal|entry

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

timeframe|higher_tf|htf

Message

Consider higher timeframe alignment before signal generation

Severity

warning

Autofix

Volume Confirmation

Id

check-volume-confirmation

Description

Breakouts and reversals should check volume

Pattern

breakout|reversal|break

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

volume|vol

Message

Consider volume confirmation for breakout/reversal signals

Severity

info

Autofix

Indicator Lag Awareness

Id

check-indicator-lag

Description

Document indicator lag in signal generation

Pattern

moving_average|MA|EMA|MACD

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

lag|delay|late

Message

Document expected lag for lagging indicators

Severity

info

Autofix

Risk/Reward Calculation

Id

check-risk-reward

Description

Calculate R:R for trade setups

Pattern

target|take_profit|tp

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

risk|reward|ratio|rr

Message

Calculate risk/reward ratio for trade setups

Severity

warning

Autofix

Sufficient Sample Size

Id

check-sample-size

Description

Pattern validation needs adequate sample size

Pattern

win_rate|success.*rate|probability

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

sample|count|n\s*=|trades

Message

Ensure sufficient sample size (30+ trades minimum)

Severity

warning

Autofix

Out-of-Sample Testing

Id

check-out-of-sample

Description

Backtests should include out-of-sample period

Pattern

backtest|backtesting

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

out.sample|oos|walk.forward|train.*test

Message

Include out-of-sample testing to validate edge

Severity

warning

Autofix

Transaction Costs Included

Id

check-transaction-costs

Description

Include fees and slippage in backtests

Pattern

backtest|pnl|profit

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

fee|commission|slippage|cost

Message

Include transaction costs in backtest calculations

Severity

warning

Autofix

Entry Invalidation Defined

Id

check-entry-invalidation

Description

Define when a setup becomes invalid

Pattern

setup|pattern|entry

File Glob

*/.{py,js,ts}

Match

present

Context Pattern

invalid|cancel|abort|expire

Message

Define invalidation conditions for trade setups

Severity

info

Autofix