OBV in Python — Coding It from Scratch (pandas)

OBV is the most beginner-friendly indicator in this whole series: a running total with an up/down rule, as the explainer covered. In pandas it’s a genuine one-liner once you spot the trick.

What you’ll need

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

The vectorized version

The rule “add volume on up closes, subtract on down closes, ignore flat ones” is exactly sign(price change) × volume, accumulated:

import numpy as np
import pandas as pd


def obv(close: pd.Series, volume: pd.Series) -> pd.Series:
    direction = np.sign(close.diff()).fillna(0)   # +1 up, -1 down, 0 flat/first bar
    return (direction * volume).cumsum().rename("obv")

np.sign(close.diff()) gives +1/−1/0 per bar, multiplying by volume applies the sign, and cumsum() accumulates it into the running total from the chart above. No loop needed.

Reading it in code: divergence

Since the OBV level is meaningless, you compare its slope to price’s. A crude but illustrative divergence check:

obv_line = obv(df["close"], df["volume"])
win = 20
price_up = df["close"] > df["close"].shift(win)
obv_down = obv_line < obv_line.shift(win)
df["bearish_divergence"] = price_up & obv_down   # price up, volume flow down

That flags spots where price rose over the window but OBV fell — a classic distribution warning to test in AlgoGen.

Gotchas

• The first bar. close.diff() is NaN on bar 0; fillna(0) keeps it out of the sum so OBV starts flat.
• Exactly-flat closes. np.sign(0) is 0, so unchanged closes correctly leave OBV untouched — matching Granville’s rule.
• Absolute value is noise. Only compare OBV’s direction to price. Never threshold on the raw OBV number; it depends on your arbitrary start point.
• Starting value. We start at 0. Some platforms seed with the first bar’s volume; it shifts the whole line by a constant and changes nothing that matters.

An OBV signal line

Since the raw level is arbitrary, a common practical move is to smooth OBV and trade the crossover — turning a directionless total into discrete signals:

obv_line = obv(df["close"], df["volume"])
obv_ma = obv_line.rolling(20).mean()
df["obv_bull"] = (obv_line > obv_ma) & (obv_line.shift(1) <= obv_ma.shift(1))

obv_bull fires when OBV crosses above its own 20-period average — a cleaner, more testable event than eyeballing the slope, and a good companion to the divergence check above.

The lazy one-liner

import pandas_ta as ta
df["obv"] = ta.obv(df["close"], df["volume"])

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.