A backtest in twenty lines

prices = [100.0, 102.0, 101.0, 103.0, 106.0]

rets = [prices[i] / prices[i - 1] - 1 for i in range(1, len(prices))]
signals = [1 if r > 0 else 0 for r in rets]          # known at each close
strat = [signals[i - 1] * rets[i] for i in range(1, len(rets))]

equity = 1.0
for r in strat:
    equity *= 1 + r

print("returns:", [f"{r:+.4%}" for r in rets])
print("signals:", signals)
print("strategy earns:", [f"{r:+.4%}" for r in strat])
print(f"equity: {equity:.6f}  ({equity - 1:+.4%})")
print(f"buy & hold: {prices[-1] / prices[0] - 1:+.4%}")

import numpy as np
import qadata

prices = qadata.trending_prices()          # 756 seeded daily closes

def make_positions(prices):
    ma = prices.rolling(20).mean()
    return (prices > ma).astype(float)     # NaN comparisons are False -> flat

rets = prices.pct_change().fillna(0.0)
pos = make_positions(prices)
strat_rets = pos.shift(1).fillna(0.0) * rets
equity = (1.0 + strat_rets).cumprod()
sharpe = strat_rets.mean() / strat_rets.std() * np.sqrt(252)

bh = (1.0 + rets).cumprod()
bh_sharpe = rets.mean() / rets.std() * np.sqrt(252)

print(f"days long : {pos.sum():.0f} of {len(pos)}")
print(f"strategy  : {equity.iloc[-1] - 1:+.2%}  Sharpe {sharpe:.2f}")
print(f"buy & hold: {bh.iloc[-1] - 1:+.2%}  Sharpe {bh_sharpe:.2f}")

import numpy as np
import qadata

prices = qadata.trending_prices()
rets = prices.pct_change().fillna(0.0)

def make_positions(prices):
    return (prices > prices.rolling(20).mean()).astype(float)

pos = make_positions(prices)
honest = (1.0 + pos.shift(1).fillna(0.0) * rets).cumprod()
cheat = (1.0 + pos * rets).cumprod()      # same bar: position knows its own close

for name, eq in [("honest", honest), ("no shift", cheat)]:
    r = eq.pct_change().fillna(0.0)
    print(f"{name:9}: {eq.iloc[-1] - 1:+8.2%}  Sharpe {r.mean() / r.std() * np.sqrt(252):.2f}")

qadata.assert_causal(make_positions, prices)
print("assert_causal: the rule itself is causal")

try:  # a rule built on the FULL-SAMPLE mean peeks at the future
    qadata.assert_causal(lambda p: (p > p.mean()).astype(float), prices)
except AssertionError as e:
    print(f"caught: {e}")