Order Book Imbalance explained for betting

In one sentence

Order book imbalance compares the unmatched money queued on the back and lay sides of the ladder near the current price, to estimate which way the next tick is more likely to go.

How it works

The Betfair ladder has two columns of waiting money. The blue "available to back" column is money from layers waiting to be matched, and the pink "available to lay" column is money from backers waiting to be matched. The best back price always sits one tick or more below the best lay price.

When one side is much thinner than the other, it is easier for incoming bets to eat through it. If the £400 sitting at the best back price is taken, the next best back price is a tick lower, so the price shortens. A large queue on the opposite side does the reverse: it acts like a wall that takes a lot of money to move.

Imbalance boils that picture down to one number between −1 and +1. Traders usually look only at the top few levels, because money far from the price is rarely matched and is often pulled before it gets close.

The maths

I = \frac{L - B}{L + B}

M = \frac{P_{b} \, L_{1} + P_{l} \, B_{1}}{L_{1} + B_{1}}

I: the imbalance, from −1 (all waiting money on the blue side) to +1 (all on the pink side).
B: total money in the blue "available to back" column over the top few levels.
L: total money in the pink "available to lay" column over the same levels.
M: the "microprice", a queue-weighted price between the two best prices.
Pb and Pl: the best back price and best lay price.
B1 and L1: the money at just those two best prices.

In plain English: the side with less money waiting is the one more likely to be cleared, and the microprice leans towards it.

Worked betting example

The home team's Match Odds ladder a few minutes before kick-off:

Blue side: 2.48 £400, 2.46 £300, 2.44 £250 (total £950).
Pink side: 2.50 £1,200, 2.52 £800, 2.54 £600 (total £2,600).

Step 1, raw imbalance: (2,600 − 950) ÷ (2,600 + 950) = 1,650 ÷ 3,550 ≈ +0.46.

Step 2, weight nearer levels more heavily (1, 0.5, 0.25). Blue becomes 400 + 150 + 62.5 = £612.50. Pink becomes 1,200 + 400 + 150 = £1,750. Imbalance = 1,137.50 ÷ 2,362.50 ≈ +0.48.

Step 3, microprice from the best level only: (2.48 × 1,200 + 2.50 × 400) ÷ 1,600 = 3,976 ÷ 1,600 = 2.485.

The microprice of 2.485 sits closer to 2.48 than 2.50, hinting that the thin £400 at 2.48 is the likelier queue to go, which would shorten the price. It is a tilt in the odds of the next tick, not a forecast of the result.

Where it's good

Scalping a tick or two in big-league football Match Odds and Over/Under goals markets before kick-off.
Timing entries: waiting for a favourable tilt before placing a back or lay.
Deciding whether to join a queue or take the price now.
As one input feature in a short-horizon machine learning model of price moves.

Limitations and pitfalls

Spoofing: large amounts are placed to look like walls and cancelled before they are matched, so the signal can be manufactured.
The edge, where it exists, is usually a fraction of a tick, so it only adds up over many trades.
It says nothing about who is right on the outcome; informed money can move the price straight through a big wall.
Betfair's data feed has a delay unless you use the streaming API, and a stale snapshot is worse than none.
Scale matters: with only £75 queued near the price, one £50 bet can swing the reading from −0.5 to +0.5, so set a minimum depth before trusting it.
The chosen depth and weights are tuning choices, so backtests can overfit easily.

How to build it

Pull ladder snapshots from the Betfair Exchange Stream API (betfairlightweight in Python handles the connection and caching).
Store top-of-book levels with millisecond timestamps in pandas or a time-series database, then test the imbalance against the next tick move.
Tip: measure predictive power per price band, as the uneven tick sizes make a tick at 1.50 and a tick at 15 very different events.

Weight of money and VWAP looks at the whole ladder and traded volume rather than just the top levels.
Queue position shows what the imbalance means for your own unmatched bet.
Market impact covers what happens when a large bet walks through the thin side.
Tick size maths explains why a one-tick move is worth different amounts across the ladder.