PulseForce Momentum Strategy Algorithm Deep Dive

1. Overview

Momentum strategies are among the most common and practical trend-enhancement approaches in quantitative trading.
They are based on one core assumption:

“The strong stay strong, the weak stay weak.”
Assets that have performed strongly in the recent past are more likely to continue rising; assets that have performed poorly are more likely to remain weak.

In PulseForce, the built-in momentum strategy is not a naive model that only looks at recent returns. Instead, it integrates:

• EMA crossovers (ema_fast / ema_slow) – to confirm the direction of short-term and medium-term trends
• Recent high breakout (recent_high_window) – to confirm whether price has truly broken above recent resistance levels
• RSI momentum signal (rsi_period / rsi_buy_threshold / rsi_sell_threshold) – to measure the strength of buying vs. selling pressure
• Volume expansion (volume_window) – to confirm that the up move is driven by real capital, not “thin-volume spikes”
• CatBoost AI filter (optional) – an AI model that filters out low-quality signals on top of the base rules

At the same time, the strategy manages risk and position sizing through:

• Hard take-profit / stop-loss (force_take_profit / force_stop_loss)
• Daily take-profit / stop-loss (daily_take_profit / daily_stop_loss)
• Dynamic capital allocation (allowable_funds_*)

and supports HyperOpt-based automatic search for optimal parameter combinations within PulseForce.

---

2. Brief history and evolution of momentum

The momentum effect has very solid academic foundations and long-term empirical support:

• In 1993, the classic paper by Jegadeesh & Titman systematically demonstrated that
  buying past winners and selling past losers can generate significant excess returns.
• Subsequent research found that the momentum effect exists across:
  equities, futures, FX, commodities, indices, and even cryptocurrencies.

In live trading, momentum strategies are widely used for:

• Short- and medium-term trend trading
• Sector / industry rotation
• Multi-asset long/short portfolios
• Trend filters in higher-frequency systems

PulseForce’s Momentum strategy is an enhanced version built on top of these classic ideas, with additional multi-indicator confirmation and AI-based signal filtering.

---

3. What problems does this Momentum strategy aim to solve?

The momentum strategy in PulseForce is designed to address several real-world challenges:

1. Return-only momentum is unstable; we need multi-dimensional confirmation of “true momentum”.
   → Use EMA crossovers + recent high breakout + RSI + volume together to form a joint decision.

2. How to avoid “fake breakouts” and “fake pumps”?
   → Price must break above a recent high and volume must expand.

3. How to get in early when the trend accelerates, instead of entering too late?
   → Use recent breakouts + elevated RSI as “trend acceleration” signals.

4. How to reduce invalid signals and drawdowns?
   → Add a CatBoost AI model on top of base rules to filter some signals, and use multi-layer TP/SL logic to manage exits.

---

4. Core indicators and signal logic

4.1 Core technical indicators

In the momentum strategy, the key indicators (all computed on 1-minute bars) are:

Indicator	Meaning
ema_fast	Fast EMA, more sensitive, represents short-term momentum
ema_slow	Slow EMA, represents medium- to long-term trend direction
recent_high	Highest price over a period (recent_high_window), used to confirm breakout
rsi	Relative Strength Index, used to measure short-term buying strength
volume_mean	Average volume over a period, used to detect volume expansion

Corresponding indicator calculations in code (excerpt):

df["rsi"] = ta.RSI(df, timeperiod=rsi_p).fillna(0)
df["ema_fast"] = ta.EMA(df, timeperiod=ema_f)
df["ema_slow"] = ta.EMA(df, timeperiod=ema_s)

df["recent_high"] = df["high"].rolling(window=rh_w, min_periods=1).max()
df["volume_mean"] = df["volume"].rolling(window=vol_w, min_periods=1).mean()

4.2 Trend labels: short trend / long trend / daily trend

Based on EMAs, the strategy additionally labels:

• short_trend: determined by the relationship between ema_fast and ema_slow
• long_trend: computed from a combination of longer-period EMAs
• daily_short_trend / daily_trend: daily-level trends derived by resampling 1m data to 1D

These trend fields can be used for:

• Features for future AI models
• Extensions to risk control modules
• More intuitive trend-state visualization in the UI

---

5. Core buy/sell logic (Runner + AI filter)

5.1 Buy signal: breakout + strong momentum + volume expansion + EMA bull trend + AI filter

In populate_buy_trend, the logic used by the Runner can be summarized as:

# 1) Base momentum condition in the Runner
base_cond = (
    (df["close"] > df["recent_high"].shift(1)) &   # Breaks above the previous recent high
    (df["rsi"] > rsi_buy_th) &                     # RSI above buy threshold
    (df["volume"] > df["volume_mean"]) &           # Volume above average (expansion)
    (df["ema_fast"] > df["ema_slow"]) &            # EMA bullish alignment
    (df["volume"] > 0)                             # Real traded volume
)

df.loc[:, "buy"] = 0

# 2) If AI filter is disabled, use base condition directly
if not self.ai_manager or not self.ai_manager.plugins:
    df.loc[base_cond, "buy"] = 1
    return df

# 3) With AI filter enabled, only evaluate AI scores for rows where base_cond is true
ai_mask = df.loc[base_cond].apply(_ai_pass, axis=1)
df.loc[base_cond, "buy"] = ai_mask.astype(int)

In plain English, the buy logic of this Momentum strategy can be described as:

“When price breaks above a recent high + RSI is strong + volume expands + EMAs are bullish, we consider this a strong momentum breakout. If the AI model also judges this signal to be high quality, then an official buy signal is generated.”

---

5.2 Sell signal: momentum decay + SL/TP combination

The strategy does not simply exit on an EMA death cross. Instead, it defines a “momentum decay signal layer” _has_sell_signal:

def _has_sell_signal(self, row: pd.Series) -> bool:
    rsi_sell_th = float(self._val(self.rsi_sell_threshold))
    c = float(row.get("close", np.nan))
    ema_f = float(row.get("ema_fast", np.inf))
    rsi = float(row.get("rsi", 50))
    return (c < ema_f) and (rsi < rsi_sell_th)

Meaning:

• Close price falls below ema_fast → short-term momentum is weakening
• RSI falls below rsi_sell_threshold → overbought / strong phase has ended

The actual exit logic is implemented in custom_exit:

# ① Hard stop-loss / hard take-profit (independent of momentum signal)
if current_profit <= -self.force_stop_loss.value:
    return "hard_sl"
if current_profit >= self.force_take_profit.value:
    return "hard_tp"

# ② No sell signal → even if daily SL/TP thresholds are hit, do nothing
if not self._has_sell_signal(row):
    return None

# ③ Only when a sell signal exists, check daily SL/TP
th = self._pick_trend_thresholds(row)  # {"sl": daily_stop_loss, "tp": daily_take_profit}
sl = float(th["sl"])
tp = float(th["tp"])

if current_profit <= -sl:
    return "sl_signal"
if current_profit >= tp:
    return "tp_signal"

In summary, this exit logic:

1. Provides a “hard TP/SL safety net” to protect capital in extreme scenarios.
2. Does not exit casually in normal conditions; it first waits for a momentum decay signal.
3. Only exits when “momentum decay + daily SL/TP threshold” are both triggered.

Compared to simple “sell at stop-loss” or “exit on EMA cross”, this is more nuanced and allows the strategy to hold positions as long as the trend persists.

---

6. PulseForce parameter configuration and HyperOpt

The PulseForce app (available on Apple App Store and Google Play) provides visual configuration interfaces for live tasks, backtests, and hyper-parameter optimization, making the strategy fully parameterized.

Backtest parameter configuration screen:

Hyper-parameter optimization configuration screen:

6.1 Core parameter group

This group defines the core structure of the momentum signal, and all parameters support HyperOpt:

Key	Name	Role	Optimizable	Suggested initial search range (otp_init_min ~ otp_init_max)
ema_fast	Fast EMA Period	Fast EMA period; smaller values react more quickly	✅	15 ~ 30
ema_slow	Slow EMA Period	Slow EMA period, smoothing medium- to long-term trend	✅	50 ~ 90
recent_high_window	Recent High Window	Lookback window used to compute “recent high”	✅	20 ~ 45
rsi_period	RSI Period	Period for RSI calculation	✅	10 ~ 25
rsi_buy_threshold	RSI Buy Threshold	Minimum RSI required to trigger a buy signal	✅	60 ~ 75
rsi_sell_threshold	RSI Sell Threshold	Maximum RSI at which a sell signal is considered (momentum decay level)	✅	20 ~ 45
volume_window	Volume MA Window	Window length used to calculate average volume	✅	10 ~ 30

These parameters map directly to the HyperOpt search space in code:

ema_fast           # IntParameter(15, 30, ...)
ema_slow           # IntParameter(50, 90, ...)
recent_high_window # IntParameter(20, 45, ...)
rsi_period         # IntParameter(10, 25, ...)
rsi_buy_threshold  # DecimalParameter(60.0, 75.0, ...)
rsi_sell_threshold # DecimalParameter(20.0, 45.0, ...)
volume_window      # IntParameter(10, 30, ...)

6.2 Capital management parameters

In PulseForce, these parameters mainly control maximum usable capital and trend-based allocation, and are currently user-configurable but not included in HyperOpt by default:

Parameter	Meaning
max_funds_allowed_using	Maximum capital that this strategy is allowed to deploy
allowable_funds_neutral	Capital ratio used in neutral market conditions
allowable_funds_uptrend	Capital ratio used in uptrend conditions
allowable_funds_downtrend	Capital ratio used in downtrend conditions

You can adjust these values in the PulseForce app according to your risk preferences. For example:

• In strong trending markets, you might set allowable_funds_uptrend to 1.0
• In uncertain markets, you can lower allowable_funds_neutral and allowable_funds_downtrend

---

6.3 Stop-loss parameters

This group is the core of risk control, and all parameters support HyperOpt:

Key	Name	Description	Optimizable	Search range
force_stop_loss	Force Stop-Loss Ratio	Hard exit when loss reaches this ratio, regardless of signal	✅	0.005 ~ 0.30 (config), or 0.01 ~ 0.1 (HyperOpt)
daily_stop_loss	Daily Stop-Loss Ratio	Only when a sell signal appears, trigger stop-loss if loss exceeds this	✅	0.005 ~ 0.30 (config), or 0.01 ~ 0.1 (HyperOpt)

Note:

• Force stop-loss is a permanently active hard defense.
• Daily stop-loss is only evaluated when _has_sell_signal is true.

---

6.4 Take-profit parameters

These also support HyperOpt:

Key	Name	Description	Optimizable	Search range
force_take_profit	Force Take-Profit Ratio	Hard exit when profit reaches this ratio, regardless of signal	✅	0.005 ~ 0.30 (config), or 0.01 ~ 0.1 (code)
daily_take_profit	Daily Take-Profit Ratio	When a sell signal appears, exit if profit exceeds this threshold	✅	0.005 ~ 0.30 (config), or 0.01 ~ 0.1 (code)

Overall:

• The Force parameters act as a hard logic layer for black-swan protection / locking extreme gains.
• The Daily parameters are a softer logic layer for finding reasonable exit points within a trend.

Within the PulseForce HyperOpt environment, these four parameters can be optimized per symbol to find more suitable ranges.

---

7. Use cases and characteristics

7.1 Suitable scenarios

• Assets with significant volatility and clear trends (e.g., US growth stocks, popular ETFs, cryptocurrencies)
• Short-term trend trading on 1-minute or 5-minute bars
• Symbols with stable liquidity that frequently generate breakout moves
• Research/live environments where you want to use HyperOpt to search for “best parameter combinations”

7.2 Less suitable scenarios

• Symbols stuck in long-term narrow ranges without clear breakouts or trends
• Extremely illiquid stocks (where prices jump in large ticks)
• Ultra-short-term, news-driven markets that require much faster reaction logic

---

8. Strengths and limitations

Strengths

• Combines EMA crossover + breakout + RSI + volume → “true momentum” with multi-factor confirmation
• Adds AI filter (CatBoostMomentumPlugin) to reduce low-quality signals
• More nuanced exit logic: momentum decay + SL/TP combined decision
• Full HyperOpt support, easy to tune per symbol
• Deep integration with PulseForce’s UI / backtesting / HyperOpt modules

Limitations

• Compared to simpler pure momentum models, this strategy is more complex and requires higher data and backtesting quality
• In long-lasting choppy markets, breakouts and strong momentum events are rare, resulting in fewer trades
• The AI filter depends heavily on model quality; if trained poorly, it may filter out good signals instead

---

9. Summary

The built-in momentum strategy in PulseForce is an enhanced trend-following strategy constructed on top of classic momentum ideas, with added multi-indicator confirmation and AI filtering:

• Uses fast/slow EMA to determine trend direction
• Uses recent high breakout + strong RSI + volume expansion to confirm momentum quality
• Uses an AI model (CatBoostMomentumPlugin) to review each base buy signal
• Uses hard/daily take-profit and stop-loss to manage risk
• Uses HyperOpt hyper-parameter optimization to search for optimal parameter combinations per symbol

For users who want to build a “smarter momentum strategy” inside PulseForce, this is a solid and extensible building block.

To learn more about strategy details and the latest features, please visit:
👉 PulseForce official website: https://pulse.hiforce.ai