On-Balance Volume (OBV): Formula, Signals, and What Tests Reveal

On-Balance Volume, usually shortened to OBV, adds volume after an up close and subtracts it after a down close. The resulting line summarizes volume signed by closing direction.

Our 129,887 configuration evaluations compare that volume weighting with an equal vote for each candle’s direction. At a 20-bar lookback, the two measures had strictly opposite signs on about 5–11% of checked candles across the historical samples. Volume confirmation can materially disagree with a simple tally of up and down closes.

How is OBV calculated?

The update has three cases:

Close above the previous close: add this bar's volume
Close below the previous close: subtract this bar's volume
Close unchanged:                leave OBV unchanged

TradingView’s OBV documentation describes that cumulative construction and its conventional uses for trend confirmation and divergence.

Suppose three candles have volumes of 100, 300, and 200. Their closes move up, down, and up. OBV’s net change is 100 − 300 + 200 = 0, although two of the three closes rose.

A move from 100 to 100.01 and a move from 100 to 110 both add the entire candle’s volume when they are up closes. OBV preserves the sign of the price change and discards its magnitude. It also does not identify the aggressor of each transaction: assigning all an up candle’s volume a positive sign is a calculation convention.

What does a rising OBV mean?

A rise means more volume has accumulated on up closes than on down closes over that portion of the line. Traders compare that direction with price, look for range breaks, or add a smoothing average.

It is a description of the signed-volume history. Price can still decline afterward. One large-volume candle can dominate the line, and a different venue or volume feed can produce a different path.

Calling OBV a leading indicator expresses the hope that a change in volume behavior comes before a price turn. The calculation itself supplies no guarantee that it will lead the next move.

Divergence means price and OBV fail to confirm one another’s highs, lows, or direction. A precise divergence strategy needs defined pivots, confirmation delays, and an entry rule. Our tested rules use rolling signed flow; they do not test discretionary chart divergence.

Why OBV’s zero level is arbitrary

If a chart starts OBV at zero and another starts it at one million, their curves differ by a constant. The signed-volume changes are identical.

That shift changes whether the raw OBV line lies above or below zero. It does not change these comparisons:

OBV now − OBV N bars ago

OBV now − its moving average

The constant cancels. A raw zero crossing therefore needs a deliberately fixed starting convention before it has a stable interpretation. An OBV-to-average comparison or rolling change asks a question that survives that shift.

This is an identity in the calculation, not a discovery established by counting more backtests.

How much difference does volume weighting make?

To compare volume with an equal vote, we put both measures on a shared scale:

Volume-weighted flow = rolling OBV change ÷ total volume
Equal-vote flow       = (up closes − down closes) ÷ number of bars

Both lie between −1 and +1 when volume is positive. The first gives a larger vote to a candle with more volume. The second gives each closing direction the same weight.

Select Volume and votes disagree. The same twenty closing directions produce opposing signs because their volumes differ.

Across all eight historical samples, strict opposite signs covered about 5–11% of post-startup candles at that lookback. Ties in the equal-vote measure are counted separately; the percentage does not include them. This distinction matters because equal votes can sum to zero while signed volume points either way.

The observations share overlapping windows. They describe how often the calculations disagreed in these samples, not independent predictions.

Why some threshold changes make no difference

An equal vote produces a discrete score. With two candles and no unchanged closes, the possible averages are −1, 0, and +1.

A positive entry threshold of 0.10 or 0.80 then selects the same +1 state. Searching more numbers between them does not create a different signal. Volume weighting can distinguish mixed-direction pairs because unequal volumes need not cancel. With unchanged closes, zero votes introduce extra possible scores: at two bars, one up close and one unchanged close average +0.5.

The backtests compare strength entries above a positive flow threshold with fade entries below the negative threshold. Both exit when the flow returns across zero.

At twenty bars and a 0.20 imbalance threshold, one matched sample recorded 442 executions with volume weighting and 242 with equal votes. The price history, threshold, and strength rule were the same. Changing the weights changed more than the appearance of the line. At that same lookback and threshold, the volume-weighted rule also recorded more executions in all eight historical replications.

This separate comparison uses small positions and fixes the threshold and strength rule while varying lookback. Recorded execution counts include position state and the platform’s fill model. They are not counts of every qualifying candle.

A larger imbalance requirement can reduce activity, but that alone does not improve a strategy. Lower activity changes exposure and fee costs as well as which moves the strategy catches. The evidence gives returned performance summaries alongside their observed-row counts.

OBV settings and comparisons

Raw OBV has no rolling period: it accumulates. A period enters when a trader adds smoothing, measures a rolling change, or defines a breakout window. Specify which of those is being changed.

OBV and VWAP also answer different questions. OBV signs volume by the direction of the close. VWAP averages prices using volume weights. Neither is an interchangeable measure of actual order flow.

For a reproducible test, keep the volume feed fixed and define the lookback, normalization, threshold, and position rule. Comparing the two weightings under the same rules makes the effect of volume easier to assess.

Reproduce the comparison

The script defaults to twenty-bar volume-weighted flow and a zero-threshold strength rule. Equal-vote mode is a matched control for closing direction.

//@version=6
strategy("OBV Lab: signed volume and equal votes", overlay=false, initial_capital=10000, default_qty_type=strategy.percent_of_equity, default_qty_value=99, process_orders_on_close=false)

flow_length = input.int(20, "Flow lookback", minval=2, maxval=198)
threshold = input.float(0.0, "Imbalance threshold", minval=0.0, maxval=0.8)
weight_mode = input.int(0, "Weight: 0 volume, 1 equal votes", minval=0, maxval=1)
rule_mode = input.int(0, "Rule: 0 strength, 1 fade", minval=0, maxval=1)

vote = close > close[1] ? 1.0 : close < close[1] ? -1.0 : 0.0
obv_value = ta.obv
volume_total = flow_length * ta.sma(volume, flow_length)
volume_flow = volume_total > 0 ? (obv_value - obv_value[flow_length]) / volume_total : na
equal_flow = ta.sma(vote, flow_length)
flow = weight_mode == 0 ? volume_flow : equal_flow
ready = bar_index >= 500 and volume_total > 0
enter_long = ready and (rule_mode == 0 ? flow > threshold : flow < -threshold)
exit_long = ready and (rule_mode == 0 ? flow < 0 : flow >= 0)

plot(flow, "Normalized signed flow", color=color.blue)
plot(threshold, "Upper threshold", color=color.orange)
plot(-threshold, "Lower threshold", color=color.green)
plot(0, "Balance", color=color.gray)
if exit_long and strategy.position_size > 0
    strategy.close("Long")
if enter_long and not exit_long and strategy.position_size == 0
    strategy.entry("Long", strategy.long)

See VWAP for another use of volume weights, and RSI for a different way to summarize the direction and size of recent price changes.