Microprice (size-weighted mid)
The microprice weights each side's price by the size on the opposite side. When the bid queue is much bigger than the ask queue, the price is more likely to tick up, so the fair value leans towards the ask.
01The formula
- the microprice: a fair-value estimate inside the spread
- best bid and best ask prices
- shares resting at the best bid and at the best ask
- the mid price, (b + a)/2
02See it move
03Where it is used
- Market makerQuotes around the microprice instead of the mid, so it does not sell cheaply just before the thin ask is taken out.
- Execution traderWaits a moment to buy when the microprice sits near the bid, and crosses quickly when it sits near the ask.
- Quant researcherUses microprice minus mid as one of the simplest short-term price prediction features.
04Why traders use it
The next price change is not a coin flip: a thin queue is likely to be eaten first. The microprice turns that into a fair price you can quote around, which matters when your whole edge is a fraction of a tick.
05Worked example
With 3,000 shares on the bid and 1,000 on the ask, the microprice is $100.0075, against a mid of $100.0050. It sits $0.0025 above the mid because 75.0% of the visible size is on the bid.
06How it is built
- The thinner side is more likely to be cleared first, moving the price towards it.
- So give the ask price a weight proportional to the bid size, and the bid price a weight proportional to the ask size.
- That gives .
- Rearranging shows it is the mid plus the half-spread times the imbalance , so it always stays between and .
07The common mistake
Swapping the weights. The bid price is weighted by the ask size, not the bid size; get it backwards and your fair value moves the wrong way, towards the side that is about to hold.
08In Python and LaTeX
b, a, Vb, Va = 100.00, 100.01, 3000, 1000 micro = (b * Va + a * Vb) / (Vb + Va) print(round(micro, 4)) # 100.0075
p_{\text{micro}} = \frac{b\,V_a + a\,V_b}{V_b + V_a} = m + \frac{a - b}{2}\cdot\frac{V_b - V_a}{V_b + V_a}09Where it fits
Rather not code the maths yourself? Describe your trading idea and let AI build the bot.
Build a bot with AgenKitThe worked example and the model run the same code, so every number matches. Models simplify markets on purpose. This is education, not financial advice.