Many trading strategies combine signals from multiple time frames — for example, a weekly trend filter with daily entry signals. backtesting.py supports this via resample_apply() from backtesting.lib, which handles resampling, indicator application, and back-alignment in a single call.
The core concept
You always pass the lowest time frame data to Backtest. To use a higher time frame, you resample up, apply an indicator, then forward-fill the values back down to the original frequency.
The critical detail is label='right': each resampled bar is labeled with the end of its period, not the start. This ensures the indicator value for a given week is only available after that week closes — preventing look-ahead bias.
resample_apply()
from backtesting.lib import resample_apply
resample_apply(rule, func, series, *args, agg=None, **kwargs)
| Parameter | Description |
|---|
rule | Pandas offset string: 'W-FRI', 'D', '4H', 'ME', etc. |
func | Indicator function to apply on the resampled series |
series | A Strategy.data.* array or pd.Series with a DatetimeIndex |
*args | Extra positional arguments forwarded to func |
agg | Aggregation function for resampling. Defaults to OHLCV_AGG for DataFrames, or the appropriate entry from OHLCV_AGG for named Series, otherwise 'last' |
**kwargs | Extra keyword arguments forwarded to func (except plot, overlay, color, scatter which go to self.I()) |
Strategy.init(), the result is automatically wrapped with self.I(), so it is plotted and managed like any other indicator.
Complete example: weekly RSI with daily MAs
This strategy uses RSI computed on daily and weekly close prices, plus four daily moving averages, to decide when to enter and exit long trades:
import pandas as pd
from backtesting import Strategy, Backtest
from backtesting.lib import resample_apply
from backtesting.test import GOOG
def SMA(array, n):
"""Simple moving average."""
return pd.Series(array).rolling(n).mean()
def RSI(array, n):
"""Relative strength index (approximate)."""
gain = pd.Series(array).diff()
loss = gain.copy()
gain[gain < 0] = 0
loss[loss > 0] = 0
rs = gain.ewm(n).mean() / loss.abs().ewm(n).mean()
return 100 - 100 / (1 + rs)
class System(Strategy):
d_rsi = 30 # Daily RSI lookback
w_rsi = 30 # Weekly RSI lookback
level = 70 # RSI entry threshold
def init(self):
# Daily moving averages — standard self.I() usage
self.ma10 = self.I(SMA, self.data.Close, 10)
self.ma20 = self.I(SMA, self.data.Close, 20)
self.ma50 = self.I(SMA, self.data.Close, 50)
self.ma100 = self.I(SMA, self.data.Close, 100)
# Daily RSI computed on the original (daily) time frame
self.daily_rsi = self.I(RSI, self.data.Close, self.d_rsi)
# Weekly RSI: resample daily Close to weekly (Friday end),
# apply RSI, then forward-fill back to daily index.
# resample_apply() handles all of this automatically.
self.weekly_rsi = resample_apply(
'W-FRI', RSI, self.data.Close, self.w_rsi)
def next(self):
price = self.data.Close[-1]
if (not self.position
and self.daily_rsi[-1] > self.level
and self.weekly_rsi[-1] > self.level
and self.weekly_rsi[-1] > self.daily_rsi[-1]
and self.ma10[-1] > self.ma20[-1] > self.ma50[-1] > self.ma100[-1]
and price > self.ma10[-1]):
# Enter long with an 8% stop-loss
self.buy(sl=.92 * price)
elif price < .98 * self.ma10[-1]:
self.position.close()
backtest = Backtest(GOOG, System, commission=.002)
stats = backtest.run()
print(stats)
How look-ahead bias is prevented
Internally, resample_apply() resamples with label='right':
resampled = series.resample(rule, label='right').agg(agg).dropna()
'W-FRI'), not the Monday open. The resampled indicator is then forward-filled back to the original daily index:
result = result.reindex(
index=series.index.union(resampled.index),
method='ffill'
).reindex(series.index)
self.weekly_rsi[-1] reflects the RSI value of the last completed week, never the current in-progress week.
Manual equivalent
For complete transparency, here is what resample_apply('D', SMA, self.data.Close, 10) does internally when working with hourly data:
class System(Strategy):
def init(self):
# 1. Get closing prices as a pandas Series with DatetimeIndex
close = self.data.Close.s
# 2. Resample to daily; label='right' prevents look-ahead bias
daily = close.resample('D', label='right').agg('last')
# 3. Apply indicator on the resampled series,
# then reindex back to the original hourly frequency
def daily_sma(series, n):
from backtesting.test import SMA
return SMA(series, n).reindex(close.index).ffill()
# 4. Wrap in self.I() so it is plotted and properly managed
self.sma = self.I(daily_sma, daily, 10, plot=False)
resample_apply() reduces all of the above to a single line.
OHLCV aggregation
When resampling a full OHLCV DataFrame rather than a single Series, use the OHLCV_AGG dictionary for correct aggregation:
from backtesting.lib import OHLCV_AGG
# Manually resample a DataFrame
weekly_df = data.resample('W-FRI', label='right').agg(OHLCV_AGG).dropna()
OHLCV_AGG is defined as:
OHLCV_AGG = {
'Open': 'first',
'High': 'max',
'Low': 'min',
'Close': 'last',
'Volume': 'sum',
}
resample_apply() receives a DataFrame or a Series named after one of these columns.
Common Pandas offset strings
| Rule | Meaning |
|---|
'D' | Calendar day |
'W-FRI' | Week ending Friday |
'ME' | Month end |
'QE' | Quarter end |
'4H' | 4-hour bars |
'2W-MON' | Bi-weekly, ending Monday |
Always pass data in the finest time frame your strategy uses. If you need both hourly and daily signals, pass hourly OHLCV data and use resample_apply('D', ...) to compute daily indicators.
resample_apply() requires that series has a DatetimeIndex. If your data uses a plain integer index, the function will raise an assertion error.