ensemble statistical forecast ?

[jonshank62]

ensemble statistical forecast is a way of using multiple indicators to achieve a score . I have currently started doing this by hand before every trade.
I like to use 5 different indicators grade each one -.5 = short , 0 = natural , +.5 =long add them all up and get a score.
my particular favorites are

200pma
20pma
true momentum
on balance volume
breakout range

Is their a way to make a algorithm for this ?

 

[useThinkScript]

ensemble statistical forecast is a way of using multiple indicators to achieve a score . I have currently started doing this by hand before every trade.
I like to use 5 different indicators grade each one -.5 = short , 0 = natural , +.5 =long add them all up and get a score.
my particular favorites are

200pma
20pma
true momentum
on balance volume
breakout range

Is their a way to make a algorithm for this ?

def score1 = if close > Average(close, 200) then 0.5 else -0.5;
def score2 = if Average(close, 20) > Average(close, 20)[1] then 0.5 else -0.5 ;
def score3 = if Momentum(10, close) > 0 then 0.5 else -0.5;
def score4 = if onbalanceVolume() > onbalanceVolume()[1] then 0.5 else -0.5 ;
def score5 = if close > Highest(high, 20)[1] then 0.5 else
             if close < Lowest(low, 20)[1] then -0.5 else 0;

def finalScore = score1 + score2 + score3 + score4 + score5;

declare lower;

plot zero = 0 ;
zero.SetDefaultColor(color.gray);

plot Score = finalScore;
Score.SetLineWeight(3);
Score.SetPaintingStrategy(PaintingStrategy.SQUARED_HISTOGRAM);
Score.AssignValueColor(if finalScore > 0 then Color.GREEN else
if finalScore < 0 then Color.RED else Color.gray);

 

[antwerks]

def score1 = if close > Average(close, 200) then 0.5 else -0.5;
def score2 = if Average(close, 20) > Average(close, 20)[1] then 0.5 else -0.5 ;
def score3 = if Momentum(10, close) > 0 then 0.5 else -0.5;
def score4 = if onbalanceVolume() > onbalanceVolume()[1] then 0.5 else -0.5 ;
def score5 = if close > Highest(high, 20)[1] then 0.5 else
             if close < Lowest(low, 20)[1] then -0.5 else 0;

def finalScore = score1 + score2 + score3 + score4 + score5;

declare lower;

plot zero = 0 ;
zero.SetDefaultColor(color.gray);

plot Score = finalScore;
Score.SetLineWeight(3);
Score.SetPaintingStrategy(PaintingStrategy.SQUARED_HISTOGRAM);
Score.AssignValueColor(if finalScore > 0 then Color.GREEN else
if finalScore < 0 then Color.RED else Color.gray);

should they be weighted?

 

[useThinkScript]

should they be weighted?

The actual definitions were kept simplistic, as it is assumed the poster would be inserting his unspecified secret sauce.

Given that the 200avg is a factor along with other lagging indicators; it can be assumed that this score is being applied to swing trades.

For swing trading, a weighting system such as this would be appropriate:

input w1 = 3.0; # 200 SMA (macro trend)
input w2 = 1.5; # 20 SMA slope (micro trend)
input w3 = 1.0; # Momentum
input w4 = 0.75; # OBV
input w5 = 2.5; # Breakout range

Giving the most weight to the 200avg trend while allowing the short-term slope minor influence.
Momentum is only for direction confirmation, and OBV is only supportive, so light-weighted.
And lastly Breakouts are the other heavyweights.

A VIP version would probably also include smoothing.

#   WEIGHTED ENSEMBLE SYSTEM
#   Five‑Factor Market Bias Model

declare lower;
# These defaults reflect real-world behavioral importance:
# 200 SMA (macro trend) > Breakout > 20 SMA > Momentum > OBV

input w1 = 3.0;     # Weight for 200 SMA trend regime
input w2 = 1.5;     # Weight for 20 SMA slope
input w3 = 1.0;     # Weight for Momentum
input w4 = 0.75;    # Weight for OBV direction
input w5 = 2.5;     # Weight for Breakout Range

def score1 =
    if close > Average(close, 200) then 0.5 else -0.5;

def score2 =
    if Average(close, 20) > Average(close, 20)[1] then 0.5 else -0.5;

def score3 =
    if Momentum(10, close) > 0 then 0.5 else -0.5;

def score4 =
    if OnBalanceVolume() > OnBalanceVolume()[1] then 0.5 else -0.5;

def score5 =
    if close > Highest(high, 20)[1] then 0.5 else
    if close < Lowest(low, 20)[1] then -0.5 else 0;

def finalScore =
      score1 * w1
    + score2 * w2
    + score3 * w3
    + score4 * w4
    + score5 * w5;

plot ZeroLine = 0;
ZeroLine.SetDefaultColor(Color.GRAY);

plot Score = finalScore;
Score.SetLineWeight(3);
Score.SetPaintingStrategy(PaintingStrategy.SQUARED_HISTOGRAM);

Score.AssignValueColor(
    if finalScore > 0 then Color.GREEN else
    if finalScore < 0 then Color.RED else
    Color.GRAY
);


input showLabel = yes;
AddLabel(showLabel,
         "Weighted Score: " + Round(finalScore, 2),
         if finalScore > 0 then Color.GREEN else
         if finalScore < 0 then Color.RED else
         Color.GRAY);

 

[Adeodatus]

Then a variation of load pressure hypothesis, main script is Convex position sizing curve (the "Edge"), added Risk/Cap mathematics, direction label, and time frame alignment multiplier See it applied to different time frames.

# ================================ #
# WEIGHTED ENSEMBLE → POSITION SIZING WITH HTF ALIGNMENT + OPTIONS MAPPING #
# AdeodatusTravelLink Series #
# ================================ #

declare lower;

# ----------------
# WEIGHTS
# ----------------
input w1 = 3.0;     # 200 SMA regime
input w2 = 1.5;     # 20 SMA slope
input w3 = 1.0;     # Momentum
input w4 = 0.75;    # OBV
input w5 = 2.5;     # Breakout

# ----------------
# CORE SCORES
# ----------------
def score1 = if close > Average(close, 200) then 0.5 else -0.5;
def score2 = if Average(close, 20) > Average(close, 20)[1] then 0.5 else -0.5;
def score3 = if Momentum(10, close) > 0 then 0.5 else -0.5;
def score4 = if OnBalanceVolume() > OnBalanceVolume()[1] then 0.5 else -0.5;

def score5 =
    if close > Highest(high, 20)[1] then 0.5
    else if close < Lowest(low, 20)[1] then -0.5
    else 0;

# ----------------
# FINAL SCORE
# ----------------
def finalScore =
      score1 * w1
    + score2 * w2
    + score3 * w3
    + score4 * w4
    + score5 * w5;

# ----------------
# NORMALIZATION
# ----------------
def maxScore = 4.375;
def normScore = finalScore / maxScore;

# ----------------
# DEAD ZONE
# ----------------
input deadZone = 0.15;

def activeScore =
    if AbsValue(normScore) < deadZone then 0
    else normScore;

# ----------------
# CONVEX SIZING CURVE
# ----------------
input curvePower = 1.8;

def sizeCurve =
    Sign(activeScore) *
    Power(AbsValue(activeScore), curvePower);

# ----------------
# TIMEFRAME ALIGNMENT (HTF)
# Daily 200 SMA vs current price
# ----------------
def dailyClose = close(period = AggregationPeriod.DAY);
def daily200 = Average(dailyClose, 200);

def htfAligned =
    if dailyClose > daily200 then 1 else 0;

def tfMultiplier =
    if htfAligned then 1.10 else 0.75;

# ----------------
# FINAL POSITION SIZE
# ----------------
input maxPosition = 1.0;

def rawSize = AbsValue(sizeCurve) * tfMultiplier;

def positionSize =
    Min(maxPosition, rawSize);

def direction =
    if sizeCurve > 0 then 1
    else if sizeCurve < 0 then -1
    else 0;

# ----------------
# OPTIONS DELTA TRANSLATION
# ----------------
def deltaLow =
    if positionSize < 0.30 then 0.25
    else if positionSize < 0.60 then 0.35
    else 0.50;

def deltaHigh =
    if positionSize < 0.30 then 0.30
    else if positionSize < 0.60 then 0.45
    else 0.65;

# ----------------
# PLOTS
# ----------------
plot ZeroLine = 0;
ZeroLine.SetDefaultColor(Color.GRAY);

plot BiasScore = finalScore;
BiasScore.SetPaintingStrategy(PaintingStrategy.SQUARED_HISTOGRAM);
BiasScore.SetLineWeight(3);
BiasScore.AssignValueColor(
    if finalScore > 0 then Color.GREEN
    else if finalScore < 0 then Color.RED
    else Color.GRAY
);

plot Size = positionSize;
Size.SetLineWeight(3);
Size.SetDefaultColor(Color.CYAN);

# ----------------
# LABELS
# ----------------
AddLabel(yes,
    "Bias Score: " + Round(finalScore, 2),
    if finalScore > 0 then Color.GREEN
    else if finalScore < 0 then Color.RED
    else Color.GRAY
);

AddLabel(yes,
    "Direction: " +
    (if direction == 1 then "LONG"
     else if direction == -1 then "SHORT"
     else "FLAT")
     + " | Size: " + AsPercent(positionSize),
    if direction == 1 then Color.GREEN
    else if direction == -1 then Color.RED
    else Color.GRAY
);

AddLabel(yes,
    "Options Delta Guide: " +
    deltaLow + " – " + deltaHigh,
    Color.YELLOW
);

AddLabel(yes,
    "HTF Alignment: " +
    (if htfAligned then "BULLISH" else "BEARISH"),
    if htfAligned then Color.GREEN else Color.RED
);

 

[jonshank62]

RSI instead of momentum seems to give better results

tweaking the grading might help for instance I consider the break out zone as 0 just above that is 5 and just above that is 4 and above that is 3 and so on thank you for the basic script finding the secret sauce could prove harder.

I changed momentum to macd and added a w6 rsi much more sensitive actually might work on lower time frames for day trading.

# WEIGHTED ENSEMBLE SYSTEM
# Six‑Factor Market Bias Model

declare lower;
# These defaults reflect real-world behavioral importance:
# 200 SMA (macro trend) > Breakout > 20 SMA > Macd > OBV >Rsi

input w1 = 2.0; # Weight for 200 SMA trend regime
input w2 = 1.5; # Weight for 20 SMA slope
input w3 = 1.0; # Weight for Macd
input w4 = 2.0; # Weight for OBV direction
input w5 = 2.0; # Weight for Breakout Range
input w6 = 4.0; # Weight for Relative Stregnth

def score1 =
if close > Average(close, 200) then 0.5 else -0.5;

def score2 =
if Average(close, 20) > Average(close, 20)[1] then 0.5 else -0.5;

def score3 =
if MACD() > 0 then 0.5 else -0.5;

def score4 =
if OnBalanceVolume() > OnBalanceVolume()[1] then 0.5 else -0.5;

def score5 =
if close > Highest(high, 20)[1] then 0.5 else
if close < Lowest(low, 20)[1] then -0.5 else 0;

def score6 =
if RSI(10, close) > 0 then 0.5 else -0.5;

def finalScore =
score1 * w1
+ score2 * w2
+ score3 * w3
+ score4 * w4
+ score5 * w5
+ score5 * w6;

plot ZeroLine = 0;
ZeroLine.SetDefaultColor(Color.GRAY);

plot Score = finalScore;
Score.SetLineWeight(3);
Score.SetPaintingStrategy(PaintingStrategy.SQUARED_HISTOGRAM);

Score.AssignValueColor(
if finalScore > 0 then Color.GREEN else
if finalScore < 0 then Color.RED else
Color.GRAY
);


input showLabel = yes;
AddLabel(showLabel,
"Weighted Score: " + Round(finalScore, 2),
if finalScore > 0 then Color.GREEN else
if finalScore < 0 then Color.RED else
Color.GRAY);

 

[antwerks]

RSI instead of momentum seems to give better results

tweaking the grading might help for instance I consider the break out zone as 0 just above that is 5 and just above that is 4 and above that is 3 and so on thank you for the basic script finding the secret sauce could prove harder.

I changed momentum to macd and added a w6 rsi much more sensitive actually might work on lower time frames for day trading.

# WEIGHTED ENSEMBLE SYSTEM
# Six‑Factor Market Bias Model

declare lower;
# These defaults reflect real-world behavioral importance:
# 200 SMA (macro trend) > Breakout > 20 SMA > Macd > OBV >Rsi

input w1 = 2.0; # Weight for 200 SMA trend regime
input w2 = 1.5; # Weight for 20 SMA slope
input w3 = 1.0; # Weight for Macd
input w4 = 2.0; # Weight for OBV direction
input w5 = 2.0; # Weight for Breakout Range
input w6 = 4.0; # Weight for Relative Stregnth

def score1 =
if close > Average(close, 200) then 0.5 else -0.5;

def score2 =
if Average(close, 20) > Average(close, 20)[1] then 0.5 else -0.5;

def score3 =
if MACD() > 0 then 0.5 else -0.5;

def score4 =
if OnBalanceVolume() > OnBalanceVolume()[1] then 0.5 else -0.5;

def score5 =
if close > Highest(high, 20)[1] then 0.5 else
if close < Lowest(low, 20)[1] then -0.5 else 0;

def score6 =
if RSI(10, close) > 0 then 0.5 else -0.5;

def finalScore =
score1 * w1
+ score2 * w2
+ score3 * w3
+ score4 * w4
+ score5 * w5
+ score5 * w6;

plot ZeroLine = 0;
ZeroLine.SetDefaultColor(Color.GRAY);

plot Score = finalScore;
Score.SetLineWeight(3);
Score.SetPaintingStrategy(PaintingStrategy.SQUARED_HISTOGRAM);

Score.AssignValueColor(
if finalScore > 0 then Color.GREEN else
if finalScore < 0 then Color.RED else
Color.GRAY
);


input showLabel = yes;
AddLabel(showLabel,
"Weighted Score: " + Round(finalScore, 2),
if finalScore > 0 then Color.GREEN else
if finalScore < 0 then Color.RED else
Color.GRAY);

Be careful cause RSI and momentum are NOT interchangeable. They may look like they act the same but they are telling you two different things!!!! RSI is a filter, momentum is an actionable tattle tail!!!

 

[antwerks]

RSI = where you are in a bounded pressure cycle
Momentum = how fast price is actually moving right now

They answer two different questions.
Below is the clean, trader-useful way to think about it.

RSI (Relative Strength Index)- What RSI really measures
RSI is a normalized oscillator built from:

• average up closes
• vs
• average down closes

over a fixed window. So conceptually:

RSI measures buying vs selling pressure, not speed.

It is bounded:

• 0 to 100

What RSI is good at:

• overbought / oversold regimes
• compression / reversion environments
• divergence detection

In other words:
RSI is a regime / condition indicator

The most important limitation (for you), RSI does not care how big the bars are. A series of tiny green candles can push RSI up the same way a series of large impulse candles can. So RSI can look strong while price is barely moving.

For my style I already trade:

• breakouts
• structure breaks
• momentum continuation

So RSI for you should mainly be used as:

• a filter
• not a trigger

Example you already lean toward RSI > 55–60 = bullish regime, RSI < 45 = bearish regime

That’s exactly how you should use it.

Momentum

Now here’s the key distinction. “Momentum” indicators usually measure some form of:


price(t) − price(t−n)

or a smoothed / normalized version of that.

Conceptually Momentum measures velocity of price change. It is not bounded.

What momentum is good at and what Momentum is very good for:

• detecting impulses
• catching early acceleration
• confirming breakouts

In physics terms (which fits your process-engineering mindset):
RSI = pressure balance
Momentum = flow rate

The big advantage for your trading, Momentum reacts immediately when:

• displacement expands
• bars start to stretch
• range increases

That is exactly what you want on:

• 1-5 minute
• breakout
• expansion legs

The real difference in one line

RSI answers “Is buying pressure dominant lately?”

Momentum answers “Is price actually moving faster right now?”

Why people get confused is that both go up in uptrends.
So visually they look similar. But internally:

Feature	RSI	Momentum
Bounded	yes (0–100)	no
Uses up vs down closes	yes	no
Uses size of move	indirectly	directly
Good for regimes
Good for breakouts
Good for divergences		sometimes

---

This is the key for the above system

From everything I've built:

• PCM
• TDFI
• SuperTrend + slope
• volume-weighted momentum
• my Doppler idea

I am already building velocity and force models

That means:
momentum belongs in the signal engine
RSI belongs in the context engine

How I would wire them in my style pipeline

Very clean and very “Antwerks style”:
Step 1 – regime filter


RSI(14) > 55

Only then…

Step 2 – entry / continuation trigger

Use:

• your momentum line crossing its zero / baseline
• or your acceleration / deceleration state

One subtle but important trap

RSI can stay overbought for a long time in strong trends. If you try to fade:


RSI > 70
in a breakout system…You will kill your best trades.


RSI overbought is

a warning in range markets
a feature in trend markets

Where momentum beats RSI (especially for the above system)
You trade:

• opening ranges
• structure reclaim
• CHoCH / IDM
• fast rotations

Momentum will turn before RSI in:

• breakout bars
• displacement candles
• squeeze releases

RSI lags because it averages up/down pressure.

In my words (process analogy), this fits a flow / throttling analogy perfectly:

• RSI = pressure balance in the system
• Momentum = actual flow through the pipe

High pressure with no flow → compression
High flow → breakout

Practical takeaway for these charts, if you only keep one as a trigger:
keep momentum

If you only keep one as a filter:
keep RSI

Final rule for you

Do not choose between them, use them like this:

RSI tells you if you should even be looking long or short
Momentum tells you when the move is actually happening

That separation will clean up a lot of the “almost good” signals you’ve been fighting in your 1-minute entries.