Markov Candle Prediction For ThinkOrSwim

[antwerks]

What pattern(settings) should be used for Mega stocks? Thank you

Human pattern recognition is powerful but also highly prone to false positives. In markets, this is often called "apophenia" — seeing meaningful signals in noise. Institutions know this risk and instead rely on statistically testable patterns with repeatable edge.
Here are the go-to patterns and models institutional and quant traders favor, ranked by statistical robustness:
Statistically Reliable Patterns Used by Institutions

1. Mean Reversion Around VWAP / Moving Averages

• Used by: Market makers, HFTs, intraday funds
• Tools: VWAP, standard deviation bands (like Bollinger), regression channels
• Why: VWAP reflects where large volume has traded — price reversion to VWAP often happens due to institutional anchoring.
• Statistical Model: Ornstein-Uhlenbeck processes (stochastic mean-reverting)
• (e.g., RSI, Bollinger Bands, Stochastics)

2. Breakouts from Consolidation Zones

• Used by: Trend followers, volatility traders
• Statistical basis:
  • Low volatility regimes often precede high volatility
  • Tested using ATR squeeze, Keltner vs. Bollinger bandwidth
• Quant Confirmation: Price > range high + increased volume + volatility expansion
• Popular Tools: Donchian channels, Keltner, ADX, ATR ramps
• Bollinger Band Width, ADX- Volume Spike, TTM Squeeze- Momentum, RSI, VWAP - Supply and demand zones, Price closes outside range

3. Momentum with Confirmation

• Used by: Quant equity, macro funds
• Core principle: Assets that perform well continue to outperform (short-term or medium-term)
• Quant version:
  • RSI > 70 + slope positive
  • Price > 20d MA & 50d MA, rising
• Key Add-on: Volume confirmation or price persistence

Signal Type	Momentum Indicator	Confirmation Pair
Trend + Momentum	MACD	ADX > 20
Breakout Momentum	RSI > 60	Volume Spike + OBV↑
Acceleration Check	ROC	Bollinger Band Width↑
Pullback Entry	Stoch RSI Cross↑	Price > 21 EMA

4. Divergence in Price vs. Momentum

• Used by: Discretionary institutional traders, macro traders
• Statistical type: Leading indicator of exhaustion
• Best used with:
  • RSI or ROC vs price
  • Volume divergence (OBV flat/down while price rises)
• Caveat: Best in combination with range-bound or mean-reverting conditions
• Recommended Setup for Divergence Detection
  Combine 1 or 2 momentum indicators with simple price structure tools:
  • RSI + MACD Histogram
  • ROC + Trendlines
  • AO + Price Swings
  • Stoch RSI + Support/Resistance

5. Order Flow Imbalances

• Used by: High-frequency trading, execution algos
• Concept: Institutional traders watch limit order book flow → imbalance leads price
• Data Required: Level 2, DOM
• Quant Proxy: Tick imbalance, delta (buy vs. sell pressure), CVD (cumulative volume delta)
• Volume Profile indicator, Delta Volume/Bid-Ask Deltas, Cumulative Delta, VWAP & Standard Deviations and Market Structure with Volume Surge

6. Volatility Expansion / Compression

• Used by: Options desks, gamma scalpers
• Quant tool:
  • IV Rank/Percentile
  • HV vs. IV comparison
  • Bollinger Band width
• Why it works: Options are priced on expected volatility. When reality diverges from pricing, opportunity arises.

Bonus: Repeated Institutional “Footprint” Patterns
These aren’t shapes like “cup and handle” — they are conditions that often signal institutional behavior:

Pattern Condition	Implication
Price flat, volume rising	Accumulation or distribution
Sharp drop, no volume	Lack of conviction / likely bounce
High OI + gamma at strike	Option dealer positioning
Multiple failed breakouts	Liquidity hunt / stop runs

So what should you look for?
If you want to think like an institution, prioritize conditions, not shapes.

Instead of:​

• “I think this is a head and shoulders”, Think:
  • “Price is failing to make higher highs while RSI is fading and volume is declining — that’s a statistically testable reversal setup.”

 

[tellyt]

i don't know what to make of this. how would you change the settings for an ETF like TQQQ, for instance, and a daily time preference instead of 5m and the like?

This is a Probability pattern matching system. The best way to use it it to keep that defaults and then move down to a timeframe that has at least 30 matches. So for TQQQ it looks like you can use the 15 min or lower. This indicator is better for trying to trade stocks or index that have ZeroDTE options . On a 15 min chart you should get a high pattern match and then make a short term option play . Without a high number of matches this is a useless indicator!

 

[tellyt]

Human pattern recognition is powerful but also highly prone to false positives. In markets, this is often called "apophenia" — seeing meaningful signals in noise. Institutions know this risk and instead rely on statistically testable patterns with repeatable edge.
Here are the go-to patterns and models institutional and quant traders favor, ranked by statistical robustness:
Statistically Reliable Patterns Used by Institutions

1. Mean Reversion Around VWAP / Moving Averages

• Used by: Market makers, HFTs, intraday funds
• Tools: VWAP, standard deviation bands (like Bollinger), regression channels
• Why: VWAP reflects where large volume has traded — price reversion to VWAP often happens due to institutional anchoring.
• Statistical Model: Ornstein-Uhlenbeck processes (stochastic mean-reverting)
• (e.g., RSI, Bollinger Bands, Stochastics)

2. Breakouts from Consolidation Zones

• Used by: Trend followers, volatility traders
• Statistical basis:
  • Low volatility regimes often precede high volatility
  • Tested using ATR squeeze, Keltner vs. Bollinger bandwidth
• Quant Confirmation: Price > range high + increased volume + volatility expansion
• Popular Tools: Donchian channels, Keltner, ADX, ATR ramps
• Bollinger Band Width, ADX- Volume Spike, TTM Squeeze- Momentum, RSI, VWAP - Supply and demand zones, Price closes outside range

3. Momentum with Confirmation

• Used by: Quant equity, macro funds
• Core principle: Assets that perform well continue to outperform (short-term or medium-term)
• Quant version:
  • RSI > 70 + slope positive
  • Price > 20d MA & 50d MA, rising
• Key Add-on: Volume confirmation or price persistence

Signal Type	Momentum Indicator	Confirmation Pair
Trend + Momentum	MACD	ADX > 20
Breakout Momentum	RSI > 60	Volume Spike + OBV↑
Acceleration Check	ROC	Bollinger Band Width↑
Pullback Entry	Stoch RSI Cross↑	Price > 21 EMA

4. Divergence in Price vs. Momentum

• Used by: Discretionary institutional traders, macro traders
• Statistical type: Leading indicator of exhaustion
• Best used with:
  • RSI or ROC vs price
  • Volume divergence (OBV flat/down while price rises)
• Caveat: Best in combination with range-bound or mean-reverting conditions
• Recommended Setup for Divergence Detection
  Combine 1 or 2 momentum indicators with simple price structure tools:
  • RSI + MACD Histogram
  • ROC + Trendlines
  • AO + Price Swings
  • Stoch RSI + Support/Resistance

5. Order Flow Imbalances

• Used by: High-frequency trading, execution algos
• Concept: Institutional traders watch limit order book flow → imbalance leads price
• Data Required: Level 2, DOM
• Quant Proxy: Tick imbalance, delta (buy vs. sell pressure), CVD (cumulative volume delta)
• Volume Profile indicator, Delta Volume/Bid-Ask Deltas, Cumulative Delta, VWAP & Standard Deviations and Market Structure with Volume Surge

6. Volatility Expansion / Compression

• Used by: Options desks, gamma scalpers
• Quant tool:
  • IV Rank/Percentile
  • HV vs. IV comparison
  • Bollinger Band width
• Why it works: Options are priced on expected volatility. When reality diverges from pricing, opportunity arises.

Bonus: Repeated Institutional “Footprint” Patterns
These aren’t shapes like “cup and handle” — they are conditions that often signal institutional behavior:

Pattern Condition	Implication
Price flat, volume rising	Accumulation or distribution
Sharp drop, no volume	Lack of conviction / likely bounce
High OI + gamma at strike	Option dealer positioning
Multiple failed breakouts	Liquidity hunt / stop runs

So what should you look for?
If you want to think like an institution, prioritize conditions, not shapes.

Instead of:​

• “I think this is a head and shoulders”, Think:
  • “Price is failing to make higher highs while RSI is fading and volume is declining — that’s a statistically testable reversal setup.”

I agree with your review but here is a different view of your word choice
Apophenia (/æpoʊˈfiːniə/) is the tendency to perceive meaningful connections between unrelated things.[2]

• Your brain evolved to find patterns for survival. Thousands of years ago, spotting the pattern of rustling leaves might mean the difference between detecting a predator or becoming dinner. This served us well in the wild

• But in trading, this same survival mechanism works against you.

• Think about it: your brain processes visual information incredibly fast, but it's also incredibly biased. It looks for familiar shapes and tries to fit new information into patterns you already Know

Here's a quick test: Look at any chart for 30 seconds. Your brain will immediately start connecting dots, drawing invisible trend lines, and spotting "obvious" patterns. Head and shoulders here, double bottom there, clear breakout forming………

 

[useThinkScript]

• Your brain evolved to find patterns for survival. Thousands of years ago, spotting the pattern of rustling leaves might mean the difference between detecting a predator or becoming dinner. This served us well in the wild

• But in trading, this same survival mechanism works against you.

@tellyt is correct.
Patterns are training wheels. They help beginning retail traders see something before they understand the underpinnings of stock market behavior.

@antwerks provides an excellent summation of the best pattern recognition. But pattern trading is not the path to long‑term profitability. To be successful in life, we learn to look beyond correlation and see causality. To be successful in trading, the same evolution is required.

Trader Development

Level	What They See	Result
Novice	Static patterns and chart shapes (flags, wedges, breakouts, double bottoms)	Random outcomes driven by apophenia; trades feel inconsistent because shapes don’t explain strength, intent, or catalyst pressure
Emerging	Behavior: who is in control, clean versus chaotic movement, strength versus exhaustion, liquidity behavior	More consistent setups; begins filtering out noise and recognizing when a move has real sponsorship
Experienced	Causality: catalyst strength, structural drift, trapped participants, liquidity vacuums, institutional intent	Predictive edge; trades align with the forces that cause moves, not the shadows they leave behind

Social media, the news, and the people at the water cooler take the lazy route and follow the brain’s natural predilection for seeing patterns. The talking heads of the stock market do the same. Pattern hunting is easy to teach, easy to package, and easy to sell.

Patterns are not what is causing price movement.
Behavior creates moves. Behavior presents in many patterns and sometimes without any.
Causality sustains moves.

 

[antwerks]

@tellyt is correct.
Patterns are training wheels. They help beginning retail traders see something before they understand the underpinnings of stock market behavior.

@antwerks provides an excellent summation of the best pattern recognition. But pattern trading is not the path to long‑term profitability. To be successful in life, we learn to look beyond correlation and see causality. To be successful in trading, the same evolution is required.

Trader Development

Level	What They See	Result
Novice	Static patterns and chart shapes (flags, wedges, breakouts, double bottoms)	Random outcomes driven by apophenia; trades feel inconsistent because shapes don’t explain strength, intent, or catalyst pressure
Emerging	Behavior: who is in control, clean versus chaotic movement, strength versus exhaustion, liquidity behavior	More consistent setups; begins filtering out noise and recognizing when a move has real sponsorship
Experienced	Causality: catalyst strength, structural drift, trapped participants, liquidity vacuums, institutional intent	Predictive edge; trades align with the forces that cause moves, not the shadows they leave behind

Social media, the news, and the people at the water cooler take the lazy route and follow the brain’s natural predilection for seeing patterns. The talking heads of the stock market do the same. Pattern hunting is easy to teach, easy to package, and easy to sell.

Patterns are not what is causing price movement.
Behavior creates moves. Behavior presents in many patterns and sometimes without any.
Causality sustains moves.

Determining causality in stock patterns using other indicators is complex, markets are chaotic and influenced by many factors. However, you can approximate causal inference (not pure causality, but strong conditional relationships) by layering data intelligently. Here's how traders approach it:

1. Sequence-Based Confirmation (Before - After Logic)
Use indicators that lead price action vs. those that confirm it.
Example:

• Volume spike → Breakout
  A surge in volume before price breakout implies accumulation (potential cause).
• Divergence → Reversal
  RSI bearish divergence often precedes a top; this sequence strengthens causality.

2. Multi-Domain Analysis (Price × Momentum × Volatility)
Layer different indicator domains to explain why something moved.

Domain	Indicator Examples	Causal Insight Example
Price	Trendlines, Price Patterns	Structure breaks show market intent
Momentum	RSI, MACD, ROC	Weakening momentum explains reversal setups
Volume	OBV, VWAP, Acc/Dist, Delta	Volume confirms institutional support/resistance
Volatility	ATR, Bollinger Bands, VIX	Expansion shows fuel for breakouts

3. Event-Based Conditioning
Causality can sometimes be inferred around known catalysts, such as:

• Earnings → Price gap
  You can correlate earnings surprises with price and volume shifts.
• FOMC decisions, CPI
  Map economic calendar events to spikes in volatility or reversals.

4. Statistical & Algorithmic Tools
Advanced users explore causal analysis with statistical tools:

• Granger Causality (used in time-series analysis): Tests if one variable predicts another.
• Bayesian networks / conditional probability trees
• Machine learning: Lasso regression or causal forests to determine which indicators matter before a price move.

Practical Tactics (Used by Quant-Inspired Traders)
Correlate Events: Build a table of pattern occurrences with indicator states 5–10 bars before it.

1. Backtest Pattern Triggering Conditions: Does the pattern only play out when certain indicators align?
2. Build Indicator Stacks: Require "cause indicators" to align before the pattern confirms.

Example Use Case:
Bullish Flag Causality Layering:

• Rising OBV → Suggests institutional buying before pattern
• Tight Bollinger Bands → Low volatility, sets up explosion
• Increasing MACD Histogram → Confirms upside momentum

This stack suggests a pattern that isn’t random—it has energy behind it.
CAUSALITY BULLISH SCANNER

Signal Type	Logic Detected
Volume Pressure	OBV rising (smart money flow)
Momentum	Price acceleration (ROC)
Volatility	Compression → potential energy
Price Confirm	Breakout beyond prior range

Here are 2 scanners that try to bridge that Gap;

# === Causality-Based Bullish Scanner ===
# Looks for volume pressure, momentum build, volatility squeeze BEFORE breakout

# === Inputs ===
input length = 14;
input squeezeLength = 20;

# === 1. Volume Pressure (Cause)
def OBV = TotalSum(if close > close[1] then volume else if close < close[1] then -volume else 0);
def obvRising = OBV > OBV[1] and OBV[1] > OBV[2];

# === 2. Momentum Leading (Cause)
def ROC = (close - close[length]) / close[length] * 100;
def rocUp = ROC > ROC[1];

# === 3. Volatility Squeeze (Fuel)
def stdDev = StDev(close, squeezeLength);
def bollWidth = BollingerBands().UpperBand - BollingerBands().LowerBand;
def lowVol = bollWidth < bollWidth[squeezeLength];

# === 4. Price Pattern Confirmation (Effect)
def breakout = close > Highest(close[squeezeLength], squeezeLength);

# === Final Causal Score
def causeScore = (obvRising and rocUp and lowVol);
def effect = breakout;

# === SCANNER TRIGGER
plot scan = causeScore and effect;

Explanation of Logic for the MACD_OBV_VOLATILITY_SQUEEZE_BO_SCANNER:

• MACD rising: Shows increasing bullish momentum.
• OBV rising: Confirms volume support behind the move.
• Low Bollinger Width: Detects volatility compression or "squeeze."
• Breakout from range: Shows price has resolved the compression upward.

declare lower;

input squeezeLength = 20;
input bbMult = 2.0;

# === MACD Signal
def macdLine = MACD().Value;
def macdSignal = MACD().Avg;
def macdRising = macdLine > macdSignal and macdLine > macdLine[1];

# === OBV Direction
def obv = TotalSum(Sign(close - close[1]) * volume);
def obvRising = obv > obv[1];

# === Bollinger Band Width (Squeeze)
def basis = Average(close, squeezeLength);
def dev = bbMult * StDev(close, squeezeLength);
def upper = basis + dev;
def lower = basis - dev;
def bollWidth = upper - lower;
def lowVol = bollWidth < Lowest(bollWidth, squeezeLength);

# === Breakout Condition
def breakout = close > Highest(close[squeezeLength], squeezeLength);

# === All Conditions Align
def bullishSetup = macdRising and obvRising and lowVol and breakout;

# === Scanner Output
plot scan = bullishSetup;

 

[joe0909]

cool. But....
Right now, your script is already very clean conceptually:
You have:

• A bullish continuation probability (green)
• A bearish continuation probability (red)

Those are conditional probabilities, not momentum. They answer: “When this trend state exists, how often does price continue on the next bar?”
That’s statistical edge, not force.
What you don’t explicitly show, “Is the current move accelerating or decelerating right now?”
That’s where a momentum plot would add value—not redundancy.
Think of your system as three layers:

1. Trend definition
   → VWAP + HMA slope
2. Statistical continuation probability
   → Your green/red probability lines
3. Real-time force
   → Momentum (what price is doing now)

Right now, you have layers 1 and 2. A third plot would give you layer 3.
That turns this from a probability model into a decision engine.
Here’s how the interpretation would change with momentum added:
Without momentum:
You see:

• Bull prob = 68%
• Bear prob = 22%

That tells you “Statistically, bulls usually continue from here.”
But it doesn’t tell you, “Are they actually pushing right now?”
With momentum:
If momentum is rising while bull probability is high:
= High confidence continuation setup
If momentum is falling while bull probability is high:
= Warning: probability is favorable, but force is missing
That’s professional-level filtering right there, lol.

Your current model is structural probability.
Momentum would be dynamic confirmation.
They are not the same thing and they complement each other perfectly.

If you were to add a third plot, I would not use RSI or MACD.
You already have price structure and slope. Keep it native.
Something like:

momentum = close - close[3];
Or smoother:

def mom = ExpAverage(close - close[1], 3);
Then plot:

• Positive → bullish force
• Negative → bearish force
• Crossing zero → momentum shift

Now you get:

Condition	Interpretation
High Bull Prob + Positive Momentum	Strong continuation
High Bull Prob + Flat Momentum	Caution
High Bull Prob + Negative Momentum	Failure risk
Low Prob + Strong Momentum	Breakout potential
High Bear Prob + Negative Momentum	Strong downside continuation

---

What you’ve built already is statistically elegant.
Adding momentum would make it trading actionable.
Right now it tells you, “What usually happens.”
With momentum added, it tells you, “What is happening and how likely it is to continue.”
That combination is rare and very powerful.

## by Strehsum 1/24/2026
# ============================================
# Probability-Based Trend Continuation Model
# with Momentum Confirmation (5-Min)
# added Momentum angle antwerks 01/24/2026
# ============================================

declare lower;

# ---------- INPUTS ----------
input hmaLength    = 21;
input lookbackBars = 200;
input momLength    = 3;      # Momentum smoothing
input momLookback  = 3;      # Raw momentum lookback

# ---------- CORE REFERENCES ----------
def v = vwap();
def h = HullMovingAvg(close, hmaLength);
def hSlope = h - h[1];

# ---------- TREND STATES ----------
def bullTrend = close > v and hSlope > 0;
def bearTrend = close < v and hSlope < 0;

# ---------- CONTINUATION CONDITIONS ----------
def bullContinue = bullTrend[1] and close > close[1];
def bearContinue = bearTrend[1] and close < close[1];

# ---------- ROLLING COUNTS ----------
def bullTrendCount = Sum(bullTrend[1], lookbackBars);
def bullContCount  = Sum(bullContinue, lookbackBars);

def bearTrendCount = Sum(bearTrend[1], lookbackBars);
def bearContCount  = Sum(bearContinue, lookbackBars);

# ---------- PROBABILITIES ----------
def bullProb = if bullTrendCount > 0 then bullContCount / bullTrendCount else 0;
def bearProb = if bearTrendCount > 0 then bearContCount / bearTrendCount else 0;

# ---------- MOMENTUM ENGINE ----------
# Raw momentum: how far price moved over last N bars
def rawMomentum = close - close[momLookback];

# Smoothed momentum: reduces noise
def momentum = ExpAverage(rawMomentum, momLength);

# ---------- PROBABILITY PLOTS ----------
plot BullContinuationProb = bullProb;
BullContinuationProb.SetDefaultColor(Color.GREEN);
BullContinuationProb.SetLineWeight(2);

plot BearContinuationProb = bearProb;
BearContinuationProb.SetDefaultColor(Color.RED);
BearContinuationProb.SetLineWeight(2);

# ---------- MOMENTUM PLOT ----------
plot MomentumLine = momentum;
MomentumLine.SetLineWeight(2);
MomentumLine.AssignValueColor(
    if momentum > 0 then Color.CYAN
    else if momentum < 0 then Color.MAGENTA
    else Color.GRAY
);

# Zero line for momentum balance
plot MomZero = 0;
MomZero.SetDefaultColor(Color.DARK_GRAY);
MomZero.SetLineWeight(1);

# ---------- INTERPRETIVE LABELS ----------
AddLabel(yes,
    "Bull Trend Continuation: " + AsPercent(bullProb),
    Color.GREEN
);

AddLabel(yes,
    "Bear Trend Continuation: " + AsPercent(bearProb),
    Color.RED
);

AddLabel(yes,
    if momentum > 0 then "Momentum: Bullish"
    else if momentum < 0 then "Momentum: Bearish"
    else "Momentum: Neutral",
    if momentum > 0 then Color.CYAN
    else if momentum < 0 then Color.MAGENTA
    else Color.GRAY
);

# ---------- OPTIONAL: HIGH-CONFIDENCE ZONES ----------
# These don't plot, but you can use them for scans or signals later

def HighBullConfidence = bullProb > 0.6 and momentum > 0;
def HighBearConfidence = bearProb > 0.6 and momentum < 0;

https://imgur.com/5xnWW2v

These two in combination with option flow are looking very promising. Been testing on the hourly, so large AH jumps/dumps can be dangerous, but otherwise winners seem to have a high chance of big returns and losers are cut fairly quickly.

 

[JK1]

Can this study be modified to adapted to various time frames?