Persia Strategy In-Depth Analysis

Chapter 1: Strategy Overview and Design Philosophy

1.1 Strategy Background

Persia is an innovative hyperparameter optimization strategy framework created by developer Mablue (Masoud Azizi). The name "Persia" comes from the ancient Persian Empire, suggesting it embodies wisdom from its Persian developer. Unlike traditional strategies with fixed technical indicators and trading logic, Persia adopts a unique "formula-based" design, allowing the system to automatically search for the optimal trading strategy combination.

1.2 Core Design Philosophy

The core design philosophy of this strategy is "let the machine discover the optimal strategy". The strategy doesn't preset fixed buy and sell logic, but instead provides a framework containing multiple technical indicators, time periods, and mathematical formulas, automatically finding optimal combinations through Freqtrade's Hyperopt functionality.

Core design elements include:

Flexible Indicator Combinations: Supports SMA, EMA, TEMA, DEMA four moving average types
Flexible Time Periods: Supports multiple periods from 3 to 189
Rich Formula System: Predefines 10 mathematical relationship formulas
Fully Adjustable Parameters: Both buy and sell parameters support hyperparameter optimization

1.3 Strategy Positioning

Persia is positioned as a strategy discovery tool rather than a finished strategy. It's suitable for:

Finding optimal strategies for specific market environments
Exploring effectiveness of different indicator combinations
Researching optimal time period configurations
Learning quantitative strategy development methods

Chapter 2: Technical Indicator System

2.1 Moving Average Types

The strategy supports four mainstream moving averages, each with unique characteristics:

SMA (Simple Moving Average)

Calculation: All prices weighted equally averaged
Characteristics: Stable, strong lag
Application: Trend confirmation

EMA (Exponential Moving Average)

Calculation: Recent prices have higher weights
Characteristics: Fast response, high sensitivity
Application: Trend change detection

TEMA (Triple Exponential Moving Average)

Calculation: Triple exponential smoothing
Characteristics: Extremely low lag, rapid response
Application: Short-term trading signals

DEMA (Double Exponential Moving Average)

Calculation: Double exponential smoothing
Characteristics: Between EMA and TEMA
Application: Medium-short-term trend tracking

2.2 Time Period System

The strategy supports extremely rich time period configurations:

COSTUMETF = [3, 7, 9, 21, 27, 63, 81, 189]

Period Interpretation:

3 periods: Extremely short-term, fastest response, most noise
7 periods: Short-term baseline, commonly used for intraday trading
9 periods: Classic short-term period
21 periods: Near monthly period (approximately one month of trading days)
27 periods: Medium-long-term transition
63 periods: Approximately one quarter of trading days
81 periods: Long-term trend
189 periods: Super long-term trend (approximately one year of trading days)

2.3 Formula System

The strategy predefines 10 mathematical relationship formulas describing relationships between two indicators:

Number	Formula	Meaning
1	A>B	Indicator A greater than indicator B
2	A<B	Indicator A less than indicator B
3	A*R==B	Indicator A times coefficient R equals indicator B
4	A==R	Indicator A equals specific value R
5	A!=R	Indicator A not equal to specific value R
6	A/B>R	Ratio of two indicators greater than R
7	B/A>R	Ratio of two indicators greater than R (reverse)
8	0<=A<=1	Indicator A between 0 and 1
9	0<=B<=1	Indicator B between 0 and 1

Formula Application: Each buy/sell condition can choose one formula, where A and B represent two indicator values, and R represents an adjustable real number parameter.

2.4 Real Number Parameter System

The strategy defines real number parameter value ranges:

REALSRANGE = [-2, 2]
DECIMALS = 1

Real number parameter R can take values between -2.0 and 2.0, with precision of one decimal place. This parameter serves as threshold or coefficient in formulas.

Chapter 3: Parameter Structure Details

3.1 Condition Quantity Configuration

The strategy supports up to 100 independent buy and sell conditions:

CONDITIONS = 3  # Currently enables 3 conditions

Each condition includes these parameters:

formula: Selected formula type
indicator: Indicator A's type
crossed: Indicator B's type (for crossover judgment)
timeframe: Indicator A's time period
crossed_timeframe: Indicator B's time period
real: Real number parameter R's value

3.2 Buy Parameter Example

Currently configured buy parameters:

buy_params = {
    "formula0": "A!=R",
    "formula1": "0<=B<=1",
    "formula2": "B/A>R",
    "indicator0": "TEMA",
    "indicator1": "DEMA",
    "indicator2": "DEMA",
    "timeframe0": 9,
    "timeframe1": 3,
    "timeframe2": 27,
    "crossed0": "TEMA",
    "crossed1": "EMA",
    "crossed2": "DEMA",
    "crossed_timeframe0": 81,
    "crossed_timeframe1": 27,
    "crossed_timeframe2": 63,
    "real0": 0.5,
    "real1": 1.8,
    "real2": -1.4,
}

3.3 Sell Parameter Example

Currently configured sell parameters:

sell_params = {
    "sell_formula0": "A==R",
    "sell_formula1": "B/A>R",
    "sell_formula2": "A!=R",
    "sell_indicator0": "EMA",
    "sell_indicator1": "DEMA",
    "sell_indicator2": "TEMA",
    "sell_timeframe0": 21,
    "sell_timeframe1": 21,
    "sell_timeframe2": 3,
    "sell_crossed0": "SMA",
    "sell_crossed1": "TEMA",
    "sell_crossed2": "EMA",
    "sell_crossed_timeframe0": 27,
    "sell_crossed_timeframe1": 27,
    "sell_crossed_timeframe2": 3,
    "sell_real0": 1.5,
    "sell_real1": 1.4,
    "sell_real2": 1.6,
}

Chapter 4: Signal Generation Mechanism

4.1 Indicator Calculation Logic

The strategy calculates all indicator combinations in populate_indicators:

for indicator in indicators:  # SMA, EMA, TEMA, DEMA
    for tf_idx in timeframes:  # 3, 7, 9, 21, 27, 63, 81, 189
        dataframe[f'{indicator}-{tf_idx}'] = getattr(ta, indicator)(dataframe, timeperiod=tf_idx)

This generates 4×8 = 32 indicator columns, such as:

TEMA-9: 9-period TEMA
EMA-27: 27-period EMA
SMA-189: 189-period SMA

4.2 Buy Signal Generation

Buy signals are generated through condition combinations:

for i in range(CONDITIONS):
    # Get parameters
    indicator = getattr(self, "indicator"+str(i)).value
    crossed = getattr(self, "crossed"+str(i)).value
    timeframe = getattr(self, "timeframe"+str(i)).value
    crossed_timeframe = getattr(self, "crossed_timeframe"+str(i)).value
    formula = getattr(self, "formula"+str(i)).value
    
    # Build data
    A = dataframe[f'{indicator}-{timeframe}']
    B = dataframe[f'{crossed}-{crossed_timeframe}']
    R = pd.Series([getattr(self, "real"+str(i)).value] * len(A))
    
    # Calculate condition
    df = pd.DataFrame({'A': A, 'B': B, 'R': R})
    conditions.append(df.eval(formula))

# All conditions AND combined
dataframe.loc[reduce(lambda x, y: x & y, conditions), 'buy'] = 1

4.3 Current Buy Conditions Interpretation

Based on current parameters, the three buy conditions are:

Condition 0:

Formula: A!=R (TEMA-9 not equal to 0.5)
A = TEMA-9
R = 0.5

Condition 1:

Formula: 0<=B<=1 (DEMA-3 value between 0 and 1)
B = EMA-27

Condition 2:

Formula: B/A>R (DEMA-63 / DEMA-27 > -1.4)
A = DEMA-27
B = DEMA-63
R = -1.4

Combined Logic: All three conditions must be satisfied simultaneously to trigger buy.

4.4 Sell Signal Generation

Sell logic is similar to buy, using sell_ prefixed parameters:

for i in range(CONDITIONS):
    indicator = getattr(self, "sell_indicator"+str(i)).value
    crossed = getattr(self, "sell_crossed"+str(i)).value
    # ... similar to buy logic

Chapter 5: Risk Management Configuration

5.1 ROI Configuration

Strategy's ROI configuration is relatively aggressive:

minimal_roi = {
    "0": 0.273,    # Immediately: 27.3%
    "26": 0.084,   # After 26 minutes: 8.4%
    "79": 0.033,   # After 79 minutes: 3.3%
    "187": 0       # After 187 minutes: 0% (forced exit)
}

Configuration Analysis:

Initial ROI set at 27.3%, very high profit target
After 26 minutes drops to 8.4%, still relatively high
After 79 minutes drops to 3.3%, becoming realistic
After 187 minutes set to 0%, forced close

This configuration indicates the strategy pursues high profits while setting a final exit mechanism.

5.2 Stop Loss Configuration

stoploss = -0.19

Interpretation: 19% stop loss is very wide, giving strategy enough room for fluctuation. Typically used for high volatility markets or strategies pursuing major trends.

5.3 Timeframe

timeframe = '5m'

5-minute timeframe is standard configuration for medium-frequency trading, balancing signal frequency and noise filtering.

Chapter 6: Hyperparameter Optimization Mechanism

6.1 Hyperopt Basics

Freqtrade's Hyperopt functionality can automatically search for optimal parameter combinations. Persia strategy is fully designed for this:

freqtrade hyperopt --hyperopt-loss SharpeHyperOptLoss --strategy Persia

6.2 Parameter Space

The strategy defines an enormous parameter space:

Formula Selection: 10 choices per condition
Indicator Selection: 4 choices per condition
Time Period: 8 choices per condition
Real Number Parameter: 40 values per condition (-2.0 to 2.0, step 0.1)

Single condition parameter space approximately: 10 × 4 × 8 × 4 × 8 × 40 ≈ 400,000 combinations Three conditions combined space is astronomical.

6.3 Optimization Suggestions

Parameter Space Control:

Reduce CONDITIONS number (like setting to 1-2)
Limit time period selection range
Narrow real number parameter range

Loss Function Selection:

SharpeHyperOptLoss: Optimize Sharpe ratio
WinRatioLoss: Optimize win rate
ProfitDrawDownHyperOptLoss: Balance profit and drawdown

Validation Methods:

Use Walk-Forward validation
Reserve test set
Check parameter boundary issues

Chapter 7: Strategy Feature Analysis

7.1 High Flexibility

Persia's biggest feature is flexibility:

Can simulate moving average crossover strategies
Can simulate mean reversion strategies
Can simulate trend following strategies
Strategy type entirely determined by parameters

7.2 Auto Machine Learning Characteristics

Strategy has AutoML (Automated Machine Learning) philosophy:

Automatically searches for optimal parameters
No need for human preset logic
Results may exceed human intuition

7.3 Overfitting Risk

High flexibility brings overfitting risk:

Huge parameter space, easy to find "perfect match to historical data" parameters
Good historical performance doesn't equal future effectiveness
Requires strict validation

Chapter 8: Strategy Advantages and Limitations

8.1 Strategy Advantages

Fully Customizable: Parameter combinations cover almost all moving average strategies
Automatic Optimization: Automatically discover optimal combinations through Hyperopt
Learning Value: Helps understand effects of different indicators and time periods
Adaptability: Can optimize different parameters for different markets

8.2 Strategy Limitations

Extremely High Overfitting Risk: Huge parameter space easily leads to overfitting
Black Box Nature: Optimization results difficult to interpret
Large Computational Resource Needs: Requires massive computation for parameter search
Lacks Protection Mechanisms: No built-in stop loss protection or fund management
Unknown Stability: Optimized parameter stability needs verification

8.3 Applicable Scenarios

Suitable for:

Strategy research and development
Exploring specific market patterns
Learning quantitative trading principles

Not suitable for:

Direct live trading (risk too high)
Beginner use
Scenarios pursuing stable returns

Chapter 9: Practical Usage Suggestions

9.1 Parameter Optimization Process

Step 1: Define optimization objectives

Clarify whether pursuing high win rate, high profit/loss ratio, or Sharpe ratio
Choose appropriate loss function

Step 2: Limit parameter space

Reduce condition quantity
Limit time period range
Narrow real number parameter range

Step 3: Phased optimization

First optimize buy conditions
Then optimize sell conditions
Finally joint optimization

Step 4: Validate results

Walk-Forward validation
Out-of-sample testing
Parameter stability analysis

9.2 Result Interpretation

After optimization completion, need to interpret actual meaning of parameters:

Example Interpretation:

formula0: A!=R, indicator0: TEMA, timeframe0: 9, real0: 0.5

Meaning: 9-period TEMA not equal to 0.5 (price away from specific value)

Reasonability Check:

Are parameters at boundary? (e.g., real exactly -2.0 or 2.0)
Are time periods reasonable? (e.g., 189 periods might be too long)
Are formulas meaningful? (e.g., A==R might be too strict)

9.3 Risk Control Supplements

Recommend adding when using:

Stop Loss Protection: Add MaxDrawdown, StoplossGuard
Fund Management: Limit single position size and total holdings
Market Filter: Add trend or volatility filters
Volume Confirmation: Add volume-related conditions

Chapter 10: Comparison with Other Strategies

10.1 Comparison with Fixed Indicator Strategies

Fixed Indicator Strategies (like NostalgiaForInfinity series):

Clear logic, easy to understand
Relatively fixed parameters
Suitable for specific market environments

Persia Strategy:

Variable logic, difficult to interpret
Completely open parameters
Can adapt to multiple markets

10.2 Comparison with Machine Learning Strategies

Machine Learning Strategies:

Use neural networks and other algorithms
Need massive data training
Truly "black box"

Persia Strategy:

Use traditional technical indicators
Based on rule-based condition combinations
Partially interpretable

10.3 Applicability Comparison

Dimension	Persia	Fixed Strategies	ML Strategies
Flexibility	Extremely high	Low	High
Interpretability	Medium	High	Low
Overfitting Risk	High	Medium	High
Computational Needs	High	Low	Extremely high
Learning Value	High	Medium	Low

Chapter 11: Summary

Persia is a unique strategy framework whose core value lies in providing a completely open parameter space, letting machines automatically discover optimal trading strategy combinations.

Core Features:

Four moving average types
Eight time periods
Ten mathematical relationship formulas
Fully parameterized buy and sell conditions

Greatest Advantage: Extremely high flexibility, can simulate and discover various moving average strategies

Greatest Risk: Extremely high overfitting risk, optimization results may only be effective in historical data

Usage Recommendations:

Use only for research and learning
Strictly limit parameter space
Conduct sufficient validation
Add risk management mechanisms
Don't directly use optimization results for live trading

Persia is more like a research tool than a finished strategy. It demonstrates an interesting direction in quantitative trading—letting machines automatically discover trading patterns, while also reminding us of overfitting dangers. For traders wanting to deeply understand technical indicator and time period effects, this is an excellent learning platform.

Chapter 1: Strategy Overview and Design Philosophy​

1.1 Strategy Background​

1.2 Core Design Philosophy​

1.3 Strategy Positioning​

Chapter 2: Technical Indicator System​

2.1 Moving Average Types​

2.2 Time Period System​

2.3 Formula System​

2.4 Real Number Parameter System​

Chapter 3: Parameter Structure Details​

3.1 Condition Quantity Configuration​

3.2 Buy Parameter Example​

3.3 Sell Parameter Example​

Chapter 4: Signal Generation Mechanism​

4.1 Indicator Calculation Logic​

4.2 Buy Signal Generation​

4.3 Current Buy Conditions Interpretation​

4.4 Sell Signal Generation​

Chapter 5: Risk Management Configuration​

5.1 ROI Configuration​

5.2 Stop Loss Configuration​

5.3 Timeframe​

Chapter 6: Hyperparameter Optimization Mechanism​

6.1 Hyperopt Basics​

6.2 Parameter Space​

6.3 Optimization Suggestions​

Chapter 7: Strategy Feature Analysis​

7.1 High Flexibility​

7.2 Auto Machine Learning Characteristics​

7.3 Overfitting Risk​

Chapter 8: Strategy Advantages and Limitations​

8.1 Strategy Advantages​

8.2 Strategy Limitations​

8.3 Applicable Scenarios​

Chapter 9: Practical Usage Suggestions​

9.1 Parameter Optimization Process​

9.2 Result Interpretation​

9.3 Risk Control Supplements​

Chapter 10: Comparison with Other Strategies​

10.1 Comparison with Fixed Indicator Strategies​

10.2 Comparison with Machine Learning Strategies​

10.3 Applicability Comparison​

Chapter 11: Summary​