Implement Incremental BBRS Strategy for Real-time Data Processing
- Introduced `BBRSIncrementalState` for real-time processing of the Bollinger Bands + RSI strategy, allowing minute-level data input and internal timeframe aggregation. - Added `TimeframeAggregator` class to handle real-time data aggregation to higher timeframes (15min, 1h, etc.). - Updated `README_BBRS.md` to document the new incremental strategy, including key features and usage examples. - Created comprehensive tests to validate the incremental strategy against the original implementation, ensuring signal accuracy and performance consistency. - Enhanced error handling and logging for better monitoring during real-time processing. - Updated `TODO.md` to reflect the completion of the incremental BBRS strategy implementation.
This commit is contained in:
Vasily.onl
@@ -12,7 +12,7 @@ from .moving_average import ExponentialMovingAverageState
|
||||
|
||||
class RSIState(SimpleIndicatorState):
|
||||
"""
|
||||
Incremental RSI calculation state.
|
||||
Incremental RSI calculation state using Wilder's smoothing.
|
||||
|
||||
RSI measures the speed and magnitude of price changes to evaluate overbought
|
||||
or oversold conditions. It oscillates between 0 and 100.
|
||||
@@ -20,13 +20,14 @@ class RSIState(SimpleIndicatorState):
|
||||
RSI = 100 - (100 / (1 + RS))
|
||||
where RS = Average Gain / Average Loss over the specified period
|
||||
|
||||
This implementation uses exponential moving averages for gain and loss smoothing,
|
||||
which is more responsive and memory-efficient than simple moving averages.
|
||||
This implementation uses Wilder's smoothing (alpha = 1/period) to match
|
||||
the original pandas implementation exactly.
|
||||
|
||||
Attributes:
|
||||
period (int): The RSI period (typically 14)
|
||||
gain_ema (ExponentialMovingAverageState): EMA state for gains
|
||||
loss_ema (ExponentialMovingAverageState): EMA state for losses
|
||||
alpha (float): Wilder's smoothing factor (1/period)
|
||||
avg_gain (float): Current average gain
|
||||
avg_loss (float): Current average loss
|
||||
previous_close (float): Previous period's close price
|
||||
|
||||
Example:
|
||||
@@ -52,30 +53,32 @@ class RSIState(SimpleIndicatorState):
|
||||
ValueError: If period is not a positive integer
|
||||
"""
|
||||
super().__init__(period)
|
||||
self.gain_ema = ExponentialMovingAverageState(period)
|
||||
self.loss_ema = ExponentialMovingAverageState(period)
|
||||
self.alpha = 1.0 / period # Wilder's smoothing factor
|
||||
self.avg_gain = None
|
||||
self.avg_loss = None
|
||||
self.previous_close = None
|
||||
self.is_initialized = True
|
||||
|
||||
def update(self, new_close: Union[float, int]) -> float:
|
||||
"""
|
||||
Update RSI with new close price.
|
||||
Update RSI with new close price using Wilder's smoothing.
|
||||
|
||||
Args:
|
||||
new_close: New closing price
|
||||
|
||||
Returns:
|
||||
Current RSI value (0-100)
|
||||
Current RSI value (0-100), or NaN if not warmed up
|
||||
|
||||
Raises:
|
||||
ValueError: If new_close is not finite
|
||||
TypeError: If new_close is not numeric
|
||||
"""
|
||||
# Validate input
|
||||
if not isinstance(new_close, (int, float)):
|
||||
# Validate input - accept numpy types as well
|
||||
import numpy as np
|
||||
if not isinstance(new_close, (int, float, np.integer, np.floating)):
|
||||
raise TypeError(f"new_close must be numeric, got {type(new_close)}")
|
||||
|
||||
self.validate_input(new_close)
|
||||
self.validate_input(float(new_close))
|
||||
|
||||
new_close = float(new_close)
|
||||
|
||||
@@ -83,8 +86,8 @@ class RSIState(SimpleIndicatorState):
|
||||
# First value - no gain/loss to calculate
|
||||
self.previous_close = new_close
|
||||
self.values_received += 1
|
||||
# Return neutral RSI for first value
|
||||
self._current_value = 50.0
|
||||
# Return NaN until warmed up (matches original behavior)
|
||||
self._current_value = float('nan')
|
||||
return self._current_value
|
||||
|
||||
# Calculate price change
|
||||
@@ -94,17 +97,30 @@ class RSIState(SimpleIndicatorState):
|
||||
gain = max(price_change, 0.0)
|
||||
loss = max(-price_change, 0.0)
|
||||
|
||||
# Update EMAs for gains and losses
|
||||
avg_gain = self.gain_ema.update(gain)
|
||||
avg_loss = self.loss_ema.update(loss)
|
||||
|
||||
# Calculate RSI
|
||||
if avg_loss == 0.0:
|
||||
# Avoid division by zero - all gains, no losses
|
||||
rsi_value = 100.0
|
||||
if self.avg_gain is None:
|
||||
# Initialize with first gain/loss
|
||||
self.avg_gain = gain
|
||||
self.avg_loss = loss
|
||||
else:
|
||||
rs = avg_gain / avg_loss
|
||||
rsi_value = 100.0 - (100.0 / (1.0 + rs))
|
||||
# Wilder's smoothing: avg = alpha * new_value + (1 - alpha) * previous_avg
|
||||
self.avg_gain = self.alpha * gain + (1 - self.alpha) * self.avg_gain
|
||||
self.avg_loss = self.alpha * loss + (1 - self.alpha) * self.avg_loss
|
||||
|
||||
# Calculate RSI only if warmed up
|
||||
# RSI should start when we have 'period' price changes (not including the first value)
|
||||
if self.values_received > self.period:
|
||||
if self.avg_loss == 0.0:
|
||||
# Avoid division by zero - all gains, no losses
|
||||
if self.avg_gain > 0:
|
||||
rsi_value = 100.0
|
||||
else:
|
||||
rsi_value = 50.0 # Neutral when both are zero
|
||||
else:
|
||||
rs = self.avg_gain / self.avg_loss
|
||||
rsi_value = 100.0 - (100.0 / (1.0 + rs))
|
||||
else:
|
||||
# Not warmed up yet - return NaN
|
||||
rsi_value = float('nan')
|
||||
|
||||
# Store state
|
||||
self.previous_close = new_close
|
||||
@@ -118,14 +134,15 @@ class RSIState(SimpleIndicatorState):
|
||||
Check if RSI has enough data for reliable values.
|
||||
|
||||
Returns:
|
||||
True if both gain and loss EMAs are warmed up
|
||||
True if we have enough price changes for RSI calculation
|
||||
"""
|
||||
return self.gain_ema.is_warmed_up() and self.loss_ema.is_warmed_up()
|
||||
return self.values_received > self.period
|
||||
|
||||
def reset(self) -> None:
|
||||
"""Reset RSI state to initial conditions."""
|
||||
self.gain_ema.reset()
|
||||
self.loss_ema.reset()
|
||||
self.alpha = 1.0 / self.period
|
||||
self.avg_gain = None
|
||||
self.avg_loss = None
|
||||
self.previous_close = None
|
||||
self.values_received = 0
|
||||
self._current_value = None
|
||||
@@ -137,22 +154,18 @@ class RSIState(SimpleIndicatorState):
|
||||
Returns:
|
||||
Current RSI value (0-100), or None if not enough data
|
||||
"""
|
||||
if self.values_received == 0:
|
||||
if not self.is_warmed_up():
|
||||
return None
|
||||
elif self.values_received == 1:
|
||||
return 50.0 # Neutral RSI for first value
|
||||
elif not self.is_warmed_up():
|
||||
return self._current_value # Return current calculation even if not fully warmed up
|
||||
else:
|
||||
return self._current_value
|
||||
return self._current_value
|
||||
|
||||
def get_state_summary(self) -> dict:
|
||||
"""Get detailed state summary for debugging."""
|
||||
base_summary = super().get_state_summary()
|
||||
base_summary.update({
|
||||
'alpha': self.alpha,
|
||||
'previous_close': self.previous_close,
|
||||
'gain_ema': self.gain_ema.get_state_summary(),
|
||||
'loss_ema': self.loss_ema.get_state_summary(),
|
||||
'avg_gain': self.avg_gain,
|
||||
'avg_loss': self.avg_loss,
|
||||
'current_rsi': self.get_current_value()
|
||||
})
|
||||
return base_summary
|
||||
|
||||
Reference in New Issue
Block a user