CMF in Python — Coding It from Scratch (pandas)

Chaikin Money Flow is three clean steps in pandas — a multiplier, a volume weight, and a normalized rolling sum, as the explainer covered. Here it is, with the flat-bar guard that most implementations forget.

What you’ll need

• Python 3.9+, pandas/numpy
• high, low, close, volume series.

From scratch

import numpy as np
import pandas as pd


def cmf(high, low, close, volume, period: int = 20):
    rng = (high - low).replace(0, np.nan)                 # guard flat bars
    mfm = ((close - low) - (high - close)) / rng          # money flow multiplier
    mfm = mfm.fillna(0.0)                                  # flat bar -> neutral

    mfv = mfm * volume                                     # money flow volume
    return (mfv.rolling(period).sum()
            / volume.rolling(period).sum()).rename("cmf")

mfm is the close-location value (+1 at the high, −1 at the low), mfv weights it by volume, and dividing the rolling sums normalizes CMF into its bounded, zero- centered range — the oscillator from the output chart.

A confirmation filter

CMF earns its keep confirming price:

c = cmf(df["high"], df["low"], df["close"], df["volume"])
df["accum"] = c > 0.05      # meaningful buying pressure
df["distrib"] = c < -0.05   # meaningful selling pressure
# e.g. only take long entries while accumulation is present:
df["long_ok"] = df["accum"]

Use long_ok to gate another entry signal and test whether it helps in AlgoGen.

Gotchas

• Flat-bar guard. When high == low, the multiplier divides by zero. replace(0, np.nan) then fillna(0.0) sets those bars neutral — the standard fix.
• Sum, then divide. CMF divides the sum of money flow volume by the sum of volume over the window — not the average of per-bar ratios. Order matters.
• Gap blindness. The multiplier only sees inside each bar’s range, so overnight gaps don’t register. Know this limitation; don’t expect CMF to catch gap moves.

The lazy one-liner

import pandas_ta as ta
df["cmf20"] = ta.cmf(df["high"], df["low"], df["close"], df["volume"], length=20)

Same indicator elsewhere: MQL5, Pine Script, EasyLanguage, NinjaScript.

---

This post is educational, not financial advice. Indicators describe the past; they don’t predict the future. Backtest anything before you risk real money on it.