Steson Tan

Call Website Contact the business

Steson Tan Trader | Risk Management Coach
The universe doesn't care about your attachment – neither does the market. Only objectivity reveals order within chaos.

Dark Pools: The Hidden Architecture of Modern Equity MarketsHow Institutional Investors Execute Large Orders, How Hidden...
10/11/2026

Dark Pools: The Hidden Architecture of Modern Equity Markets

How Institutional Investors Execute Large Orders, How Hidden Liquidity Affects Price Discovery, and What Traders Can Actually Learn From Dark Pool Activity
A significant part of modern equity trading takes place outside the publicly displayed order book.

While retail traders often analyse candlesticks, exchange volume, bid–ask spreads, and Level 2 liquidity, institutional investors have access to additional ex*****on venues that allow them to trade without publicly displaying their intentions before ex*****on.

These venues are commonly known as Dark Pools.

Dark pools play an important role in modern financial market structure. They allow institutional investors, asset managers, hedge funds, and other participants to search for counterparties while limiting the information leakage associated with displaying large orders on public exchanges.

But this creates an important challenge.

If some trading interest is not publicly visible before ex*****on, how complete is the picture of supply and demand presented by the conventional order book?

For quantitative traders, understanding dark pools is not about identifying secret institutional trades or attempting to predict every movement from large transaction prints.

It is about understanding how liquidity is distributed across different trading venues, how institutional ex*****on works, and how hidden trading activity interacts with public price discovery.

The visible order book is only one part of the market's liquidity structure.

Let's examine the mechanics, academic research, institutional behaviour, and practical implications.

1. What Are Dark Pools?
A dark pool is a non-displayed trading venue where participants can submit orders without publicly revealing their trading interest before ex*****on.

In the United States, the term generally refers to non-displayed Alternative Trading Systems (ATSs).

These systems match buyers and sellers outside the traditional displayed order books of national securities exchanges.

An ATS is an SEC-regulated electronic trading system that matches securities orders. It is not itself a national securities exchange.

The defining characteristic of dark trading is the absence of publicly displayed pre-trade quotations.

Unlike a conventional exchange order book, other market participants generally cannot see the price and quantity of resting dark-pool orders.

However, this does not mean that dark-pool transactions are completely invisible.

Executed transactions in listed stocks are subject to applicable trade-reporting requirements. FINRA explains that ATS transactions are submitted to a Trade Reporting Facility (TRF) and published on the consolidated tape alongside exchange transactions.

This creates an important distinction:
- Before ex*****on: Trading interest is generally not publicly displayed.
- After ex*****on: The transaction is reported under applicable reporting rules.
Dark pools therefore provide pre-trade confidentiality, not absolute post-trade anonymity.

References: SEC Investor.gov — Alternative Trading Systems | FINRA — Where Do Stocks Trade?
2. Why Do Institutional Investors Use Dark Pools?
Consider an asset manager that wants to purchase 2 million shares of a publicly traded company.

Executing the entire order immediately on a public exchange may reveal significant buying demand.

Other market participants could observe the large order, adjust their quotations, or trade ahead of the expected demand.

As the institution continues buying, the available selling liquidity may become more expensive.

The institution could therefore experience adverse price movement before completing its transaction.

This problem is known as market impact.

Dark pools were developed partly to address this challenge.

By placing orders in non-displayed venues, institutional investors can search for available counterparties without broadcasting their full trading intentions to the public market.

Potential advantages include:

Reduced information leakage

Other public-market participants cannot directly inspect the complete resting order before ex*****on.

Potentially lower market impact

Large transactions may be matched without consuming multiple visible price levels on an exchange.

Price improvement

Some dark venues facilitate ex*****ons at prices better than the prevailing displayed bid or ask, often using reference prices derived from public markets.

Institutional ex*****on flexibility

Different ex*****on mechanisms allow institutions to seek liquidity while managing ex*****on costs and timing.

However, dark pools do not guarantee better ex*****on.

An order may remain unfilled, execute only partially, or receive an outcome that is less favourable than an alternative ex*****on venue.

The actual result depends on available counterparties, venue design, order priority, market conditions, and ex*****on quality.
3. How Large Is the Dark Trading Market?
One of the most important developments in modern equity market structure is the expansion of off-exchange trading.

However, it is essential to distinguish dark-pool activity from total off-exchange activity.

According to Cboe's January 2026 review of US equity markets:
- Off-exchange trading represented 50.6% of total consolidated US equity volume in 2025, measured through Trade Reporting Facility (TRF) market share.
- Of that TRF volume, 18.7% occurred on Alternative Trading System platforms.
- The remaining 81.3% occurred through principal dealers.
These are different categories.

Principal-dealer activity includes off-exchange ex*****ons that are not necessarily dark-pool ATS transactions.

Therefore, it would be inaccurate to describe the entire 50.6% off-exchange share as dark-pool trading.

The figures highlight just how fragmented modern equity ex*****on has become.

Trading is distributed across traditional exchanges, ATSs, electronic market makers, wholesalers, and other ex*****on mechanisms.

For quantitative traders, the implication is significant:

Exchange volume alone does not represent the entirety of trading activity in US equities.

Source: Cboe — 2025 US Equities Year in Review.
4. How Dark Pools Work: The Institutional Ex*****on Process
To understand dark-pool activity, it helps to examine a simplified institutional trading workflow.

Imagine an asset manager that wants to sell 500,000 shares without immediately revealing the entire order.

The institution may submit its order to a broker or ex*****on algorithm.

The ex*****on system can evaluate several potential venues, including public exchanges and non-displayed ATSs.

If the dark venue finds a compatible counterparty, the transaction may be executed according to the venue's matching rules.

Depending on the venue, ex*****ons may use prices linked to the National Best Bid and Offer (NBBO), midpoint prices, or other permitted pricing mechanisms.

After ex*****on, the transaction is reported through the applicable reporting infrastructure.

The simplified process is:
1. The institution submits a large order.
2. The broker or ex*****on algorithm evaluates available venues.
3. A non-displayed venue searches for compatible trading interest.
4. A transaction is executed if a suitable counterparty is found.
5. The executed trade is reported under applicable trade-reporting requirements.
The important distinction is that the public market may observe the completed transaction without having observed the original resting order.

The trade becomes visible after ex*****on, but the full pre-trade intention remains hidden.
5. Dark Pools and Price Discovery: Do Hidden Markets Distort Prices?
This is one of the most important debates in financial market microstructure.

Price discovery is the process through which market participants incorporate information into asset prices.

Public exchanges play a central role because they display quotations and execute transactions that help establish widely observed reference prices.

Dark pools generally operate differently.

Many use publicly derived reference prices rather than independently establishing their own publicly displayed quotations.

This creates an interesting relationship between lit and dark markets.
The Lit Market: Public Price Formation
Public exchanges display bid and ask quotations.

Participants can observe available liquidity, submit orders, and interact with other market participants.

These interactions contribute to publicly observable price discovery.
The Dark Market: Non-Displayed Ex*****on
Dark pools allow participants to submit orders without publicly displaying their trading interest.

They can facilitate transactions at prices linked to the publicly available market.

Consequently, dark venues may depend on the pricing information generated by lit markets.
What Does Academic Research Say?
The research does not provide a simple answer that all dark trading improves or damages price discovery.

Different studies identify different effects.

Research 1: Zhu (2014) — Do Dark Pools Harm Price Discovery?

Haoxiang Zhu's study, published in The Review of Financial Studies, examined the interaction between dark pools and public exchanges.

The research showed that under certain conditions, dark pools can improve price discovery by concentrating informed trading activity on public exchanges.

At the same time, this can reduce liquidity on the exchange.

This is an important insight because it challenges the assumption that all dark trading necessarily damages price discovery.

Research 2: Comerton-Forde and Putniņš (2015) — Dark Trading and Price Discovery

Their empirical research in the Journal of Financial Economics found that the impact of dark trading depends on its scale and structure.

They reported that lower levels of non-block dark trading could be benign or even beneficial for informational efficiency.

However, higher levels were associated with deteriorating price discovery and greater adverse-selection risk on public exchanges.

Importantly, their study found no evidence that block trades in dark venues impeded price discovery in the same way.

Research 3: Ye (2024) — Understanding the Impacts of Dark Pools on Price Discovery

Linlin Ye's research in the Journal of Financial Markets provides another perspective.

The study examines how the information environment affects the relationship between dark pools and price discovery.

Its findings suggest that the impact depends on the precision and distribution of information among market participants.

Under some conditions, dark pools can improve informational efficiency. Under others, they can weaken it.
The Research Takeaway
Dark pools do not automatically destroy price discovery.

Their effect depends on:
- The proportion of trading activity executed in dark venues.
- Whether trading interest is informed or uninformed.
- The structure of the matching mechanism.
- The liquidity available on public exchanges.
- How information moves between trading venues.
For professional traders, the most useful interpretation is that dark pools contribute to market fragmentation, while their actual impact on price discovery is conditional.

Research references:
- Zhu (2014) — Do Dark Pools Harm Price Discovery?
- Comerton-Forde & Putniņš (2015) — Dark Trading and Price Discovery
- Ye (2024) — Understanding the Impacts of Dark Pools on Price Discovery
6. Understanding Dark Pool Prints: What Can Traders Actually Observe?
One of the most popular forms of dark-pool analysis involves monitoring large off-exchange transactions.

These are often called dark-pool prints.

A print represents an executed transaction that has been reported to the market.

For example, imagine the consolidated tape displays several transactions:
Time Ex*****on price Quantity
10:15:12 $100.05 25,000
10:18:34 $100.04 40,000
10:22:51 $100.05 30,000
10:31:09 $100.03 50,000

These transactions demonstrate that substantial trading activity occurred around $100.03–$100.05 during the observed period.

This may identify an area of historical ex*****on interest.

However, there is a major limitation.

A publicly reported trade print does not automatically reveal whether the institution was buying or selling.

Every transaction has both a buyer and a seller.

Without reliable information about the initiating side, the participant identities, and the full position context, it is not possible to conclude that the transaction represents institutional accumulation or distribution.

A large trade may represent:
- A buyer establishing a new position.
- A seller reducing an existing position.
- A portfolio rebalancing transaction.
- A hedge against another instrument.
- A closing or transfer of existing exposure.
The ex*****on itself is observable.

The underlying investment intention is generally not.
7. Can Dark Pool Prints Identify Support and Resistance?
Many traders monitor repeated dark-pool transactions around particular price levels.

The idea is that areas with significant historical ex*****on activity may be relevant when price returns.

This can be a useful research hypothesis, but it needs careful interpretation.

Suppose multiple large transactions occur around $100.

A trader might identify $100 as an area of historical ex*****on activity.

If price later approaches the same level, it may be worth observing:
- Whether trading activity increases.
- Whether price accepts or rejects the area.
- Whether public exchange liquidity changes.
- Whether subsequent price movements show continuation or reversal.
However, the historical trades do not prove that the participants involved still hold their positions.

They may already have exited, hedged, or changed their exposure.

Furthermore, price levels associated with large dark-pool ex*****ons may not represent meaningful levels of future buying or selling demand.

A professional approach is therefore to use large prints as historical ex*****on references, not automatic support and resistance levels.
8. The Dark Index (DIX): Measuring Dark-Market Activity
Some market analytics services publish proprietary indicators designed to estimate directional information from dark-market trading data.

One example is the Dark Index (DIX), popularised by SqueezeMetrics.

DIX is based on aggregated off-exchange short-volume data and uses a particular methodology to construct a market-wide proxy.

The idea is to examine patterns in off-exchange trading activity that may contain information about institutional positioning or future market behaviour.

However, DIX should not be interpreted as a direct measurement of institutional buying.

Why?

Because short-volume data does not necessarily mean that an investor is expressing a bearish view.

Short sales can occur for several reasons, including:
- Market-making activity.
- Hedging.
- Arbitrage.
- Risk management.
- Other ex*****on-related purposes.
The metric is therefore an analytical proxy rather than a direct record of net institutional buying or selling.

It should also be recognised that proprietary indicators can differ in methodology and data treatment.

For traders, the best approach is to examine the methodology, test the historical relationship, and avoid treating a single indicator threshold as a guaranteed market signal.
9. The Limitations of Dark Pool Analysis
Dark-pool analysis can provide useful additional context, but several limitations must be considered.
A. Incomplete Visibility
The public consolidated tape shows executed transactions, not every resting order or unexecuted trading intention.

The visible data is therefore an incomplete representation of the total liquidity structure.
B. Directional Ambiguity
A trade print includes both buying and selling.

Without a reliable method of identifying the initiating side, its directional interpretation remains uncertain.
C. Delayed and Aggregated Information
Some off-exchange data is available only through periodic aggregate reports.

FINRA publishes ATS and non-ATS trading information, including weekly data, but these datasets are not equivalent to a real-time view of every venue's order book.
D. Hedging and Arbitrage
Large ex*****ons can reflect hedging, portfolio rebalancing, and arbitrage rather than outright directional conviction.
E. Market Conditions Change
The usefulness of dark-pool observations depends on volatility, liquidity, ex*****on mechanisms, and the wider market environment.

A pattern observed in one market regime may not remain informative in another.

These limitations reinforce one key principle:

Dark-pool activity should be treated as contextual market information, not an independent trading signal.
10. A Practical Framework for Analysing Dark Pool Activity
For quantitative and systematic traders, dark-pool analysis becomes more useful when combined with broader market information.

A structured process can follow six steps.

Step 1 — Identify the Trading Source

Establish whether a transaction is classified as ATS activity or another form of off-exchange ex*****on.

Do not assume that every trade reported through a TRF originated from a dark pool.

Step 2 — Examine the Transaction Distribution

Analyse transaction size, ex*****on price, time, frequency, and concentration.

Look for repeated activity rather than interpreting one isolated large trade.

Step 3 — Compare Ex*****ons with Public Market Prices

Compare the print with the prevailing NBBO, midpoint, VWAP, and public-market trading conditions at the time.

This provides greater context about the ex*****on.

Step 4 — Evaluate Subsequent Price Behaviour

Observe how price behaves after the reported transactions.

Does it continue higher, move lower, or remain around the same region?

Do not assume that the transaction itself establishes a future support or resistance level.

Step 5 — Integrate Other Market Data

Combine dark-pool observations with:
- Public exchange volume.
- Order Book Imbalance.
- Order Flow Imbalance.
- Market depth.
- VWAP.
- Volatility.
- Broader market structure.
This helps place off-exchange ex*****ons within the wider trading environment.

Step 6 — Validate the Research

If using dark-pool information in a systematic strategy, define precise rules and test them using historical data.

Account for reporting delays, transaction costs, data quality, market regimes, and the possibility that apparent directional signals are simply noise.

The goal is to determine whether dark-pool information adds predictive or ex*****on value beyond conventional market data.

Conclusion: The Visible Market Is Not the Entire Market
Dark pools are an established feature of modern financial markets.

They address a real institutional problem: executing large orders while limiting information leakage and potential market impact.

But they also contribute to market fragmentation and create additional challenges for traders trying to interpret the full distribution of supply and demand.

Public exchanges continue to play an essential role in price discovery, while non-displayed venues provide alternative ways for participants to find liquidity.

For traders, the opportunity is not to assume that every large dark-pool print reveals institutional intentions.

It is to understand what the data actually represents, recognise what remains hidden, and combine ex*****on information with public market structure.

Dark pools reveal where certain transactions have occurred. They do not reveal the complete intentions or positions of the participants involved.

The professional edge comes from understanding both the visible and non-visible components of market structure.

Research references and further reading
1. SEC Investor.gov — Alternative Trading Systems (ATSs)
Explains the regulatory definition of ATSs and the non-displayed nature of dark pools.

2. FINRA — Where Do Stocks Trade?
Explains the distinction between ATS trading, principal dealers, and the reporting of off-exchange transactions.

3. Cboe — 2025 U.S. Equities Year in Review
Reviews off-exchange trading share and the distinction between ATS and principal-dealer activity in 2025.

4. Zhu, H. (2014) — Do Dark Pools Harm Price Discovery?
The Review of Financial Studies.

5. Comerton-Forde, C. & Putniņš, T. J. (2015) — Dark Trading and Price Discovery
Journal of Financial Economics.

6. Ye, L. (2024) — Understanding the Impacts of Dark Pools on Price Discovery
Journal of Financial Markets.

7. FINRA — OTC Transparency Data
Provides aggregate ATS and non-ATS trading information.

LULU | Quantitative Research & Trading

Educational content only. Dark-pool transaction data is incomplete and should not be interpreted as direct evidence of institutional directional positioning or future price movements.

The central research conclusion is that dark pools serve a legitimate ex*****on function, but their impact on price discovery depends on market structure and the amount and type of trading they attract. The 2025 Cboe statistics are particularly important because they separate ATS activity from principal-dealer off-exchange activity rather than incorrectly labelling all off-exchange trading as dark-pool volume

Beyond the Bid–Ask Spread: How Order Book Imbalance and Microprice Reveal Short-Term Price PressureA Quantitative Framew...
07/11/2026

Beyond the Bid–Ask Spread: How Order Book Imbalance and Microprice Reveal Short-Term Price Pressure

A Quantitative Framework for Understanding Limit Order Books, Liquidity, and Short-Horizon Price Discovery
Most traders focus on what price has already done.

They study candlestick patterns, moving averages, support and resistance, and momentum indicators to understand market direction.

But there is another layer of information that receives considerably less attention from ordinary traders: the limit order book.

Before a transaction takes place, buyers and sellers submit orders at different price levels. These orders form a constantly changing structure of displayed liquidity, revealing where participants are willing to trade and how much quantity they are prepared to offer.

For quantitative traders, this information provides an opportunity to study short-term price formation before it becomes visible in conventional price charts.

Two important concepts help translate this information into measurable signals:

Order Book Imbalance (OBI): Measures the relative distribution of displayed liquidity between the bid and ask sides.

Microprice: Adjusts the midpoint using order book information to estimate short-term price pressure.

Together, these concepts form an important part of modern market microstructure research, particularly in high-frequency trading, electronic market making, and algorithmic ex*****on.

However, there is an important distinction between detecting order book pressure and predicting a profitable trade.

A large bid does not guarantee that price will rise. A large ask does not guarantee that price will fall.

The real objective is to understand how displayed liquidity, order flow, and market depth interact to influence short-term price movements.

Let's examine the mathematics, empirical evidence, and practical applications.

1. Understanding the Limit Order Book: Where Price Is Formed
A limit order book is an electronic record of outstanding buy and sell limit orders.

It is organised into two sides:

Bid side: Orders from participants willing to buy at specified prices.

Ask side: Orders from participants willing to sell at specified prices.

The highest available buying price is called the Best Bid, while the lowest available selling price is called the Best Ask.

For example:

Bid price

Bid quantity

Ask price

Ask quantity

$100.00

200

$100.02

800

$99.99

1,000

$100.03

600

$99.98

500

$100.04

400

The best bid is $100.00, and the best ask is $100.02.

The difference between these prices is the bid–ask spread:

\[ Spread=Ask-Bid \]

\[ Spread=100.02-100.00=\$0.02 \]

The conventional midpoint is:

\[ Midprice=\frac{Bid+Ask}{2} \]

\[ Midprice=\frac{100.00+100.02}{2} =\$100.01 \]

The midpoint is a useful reference for measuring transaction costs and market movements.

However, it does not account for the quantity of liquidity resting on either side.

Consider the example above.

The bid side contains 200 units at the best bid, while the ask side contains 800 units at the best ask.

Although the midpoint is $100.01, the displayed liquidity is not evenly distributed.

This is where order book imbalance becomes useful.

2. Order Book Imbalance (OBI): Measuring the Pressure of Displayed Liquidity
Order Book Imbalance measures the relative difference between buying and selling quantities in a limit order book.

The standard top-of-book queue imbalance formula is:

\[ \boxed{ OBI=\frac{Q_{bid}-Q_{ask}} {Q_{bid}+Q_{ask}} } \]

Where:

\(Q_{bid}\) = Quantity resting at the best bid.

\(Q_{ask}\) = Quantity resting at the best ask.

The result ranges between −1 and +1.

Understanding the OBI values
OBI reading

Interpretation

+1.00

Only bid quantity is present

+0.50

Bid quantity dominates

0.00

Equal displayed quantities

−0.50

Ask quantity dominates

−1.00

Only ask quantity is present

A positive reading indicates greater displayed quantity on the bid side.

A negative reading indicates greater displayed quantity on the ask side.

A reading close to zero indicates relatively balanced displayed quantities.

Numerical example
Using the previous order book:

Bid quantity: 200

Ask quantity: 800

Therefore:

\[ OBI=\frac{200-800}{200+800} \]

\[ OBI=\frac{-600}{1000} \]

\[ \boxed{OBI=-0.60} \]

This indicates that displayed ask quantity is substantially larger than bid quantity at the best quotes.

The book is therefore ask-side dominant.

This may indicate downward short-term queue pressure because the bid queue is relatively smaller and may be depleted more easily by incoming sell orders.

However, this is a probabilistic interpretation, not a directional guarantee.

Displayed orders can be cancelled, replenished, or executed. A static OBI reading does not reveal the complete intentions of market participants or the future sequence of order book events.

OBI measures the distribution of displayed liquidity, not actual buying and selling aggression.

3. Multi-Level Order Book Imbalance
The basic OBI formula considers only the best bid and ask.

However, professional quantitative traders may extend the calculation across multiple levels of the order book.

This provides a broader view of displayed liquidity beyond the top quotes.

A general weighted multi-level imbalance measure is:

\[ OBI_N= \frac{ \sum_{i=1}^{N}w_iQ_{bid,i} - \sum_{i=1}^{N}w_iQ_{ask,i} }{ \sum_{i=1}^{N}w_iQ_{bid,i} + \sum_{i=1}^{N}w_iQ_{ask,i} } \]

Where:

\(N\) = Number of order book levels.

\(Q_{bid,i}\) = Bid quantity at level \(i\).

\(Q_{ask,i}\) = Ask quantity at level \(i\).

\(w_i\) = Weight assigned to each level.

A common design choice is to assign greater weight to prices closer to the best bid and ask.

For example, the nearest level may receive a weight of 1.0, while deeper levels receive progressively smaller weights.

This approach reflects the idea that liquidity close to the current market price may be more immediately relevant to ex*****on.

However, the weighting scheme must be tested rather than assumed to be optimal.

Multi-level OBI can provide more information about displayed depth, but it can also introduce noise from distant orders that may never be executed.

The appropriate number of levels depends on the market, instrument, tick size, and trading horizon.

4. The Microprice: A More Responsive Alternative to the Midpoint
The midpoint is calculated using the best bid and ask prices.

It does not incorporate the relative quantities available at those prices.

A queue-weighted midpoint adjusts the midpoint according to the displayed bid and ask sizes.

One commonly used formula is:

\[ \boxed{ MP= \frac{ Ask\times Q_{bid} + Bid\times Q_{ask} }{ Q_{bid}+Q_{ask} } } \]

Where:

\(MP\) = Queue-weighted midpoint.

\(Ask\) = Best ask price.

\(Bid\) = Best bid price.

\(Q_{bid}\) = Best bid quantity.

\(Q_{ask}\) = Best ask quantity.

The larger queue receives greater weight on the opposite quote.

This produces an adjusted midpoint that moves towards the side of the smaller queue.

A practical example
Assume:

Best Bid: $100.00

Best Ask: $100.02

Bid Quantity: 200

Ask Quantity: 800

The conventional midpoint is:

\[ Midprice=\frac{100.00+100.02}{2} =\$100.01 \]

The queue-weighted midpoint becomes:

\[ MP= \frac{100.02(200)+100.00(800)} {200+800} \]

\[ MP= \frac{20,004+80,000}{1,000} \]

\[ \boxed{MP=\$100.004} \]

Compare the two values:

Conventional midpoint: $100.010

Queue-weighted midpoint: $100.004

The adjusted midpoint is $0.006 below the conventional midpoint.

Why?

Because the ask queue contains substantially more displayed quantity than the bid queue.

The smaller bid queue may be more vulnerable to incoming aggressive selling, making a downward next-price movement more plausible under certain market conditions.

This illustrates how order book quantities can influence a short-horizon price estimate.

An important technical distinction
The queue-weighted midpoint is often called a microprice in practical trading discussions.

However, Sasha Stoikov's formal 2018 paper, The Micro-Price: A High-Frequency Estimator of Future Prices, develops a more sophisticated model-based microprice.

It treats the microprice as a conditional expectation of future mid-prices based on information in the order book, including spread and imbalance.

Stoikov's empirical results show that his estimated microprice predicts short-term prices better than the conventional midpoint and simple weighted midpoint.

Therefore, it is important to distinguish the simple queue-weighted calculation from the more formal microprice estimator.

Research: Stoikov (2018) — The Micro-Price: A High-Frequency Estimator of Future Prices.

5. Order Book Imbalance vs. Order Flow Imbalance
This distinction is particularly important for quantitative traders.

OBI and Order Flow Imbalance (OFI) are related but measure different things.

Order Book Imbalance
OBI measures the distribution of resting displayed quantities at a particular point in time.

It answers:

How much displayed liquidity is currently available on the bid compared with the ask?

It is essentially a snapshot of the order book.

Order Flow Imbalance
OFI measures changes in supply and demand at the best bid and ask through order book events.

These events include:

New limit orders.

Marketable buy orders.

Marketable sell orders.

Limit order cancellations.

Changes in best bid and ask prices.

For example, a large sell order may consume several bid levels.

At the same time, new buy limit orders may replenish the bid.

A static OBI measurement might not fully capture this sequence of activity.

OFI attempts to measure the net effect of these events over a defined interval.

Why OFI matters
Cont, Kukanov, and Stoikov (2014), in their research The Price Impact of Order Book Events, examined 50 US stocks using NYSE Trades and Quotes data.

Their findings showed that short-term price changes were strongly related to OFI.

They also identified an approximately linear relationship between price changes and OFI, with the price-impact coefficient inversely related to market depth.

In simple terms:

The same order flow imbalance can have a greater price impact when available market depth is lower.

This is an important connection between order flow and liquidity.

A market with substantial depth may absorb considerable order flow without significant price movement.

A thinner market may experience much greater price changes from the same imbalance.

Research: Cont, Kukanov & Stoikov (2014) — The Price Impact of Order Book Events.

6. What Does the Academic Research Say About Predictability?
The predictive value of order book imbalance is supported by empirical research, particularly at short horizons.

However, its effectiveness varies across instruments, liquidity conditions, and market structures.

Research 1: Queue Imbalance as a One-Tick-Ahead Price Predictor
Gould and Bonart (2016) investigated whether queue imbalance could predict the direction of the next mid-price movement.

Their study examined ten liquid Nasdaq stocks using logistic regression models.

They found a statistically significant relationship between bid/ask queue imbalance and subsequent mid-price direction.

More importantly, the results differed according to stock characteristics.

Their models achieved considerable improvements over a simple benchmark for large-tick stocks and moderate improvements for small-tick stocks.

This demonstrates that the predictive value of OBI is not uniform across all instruments.

Reference: Gould & Bonart — Queue Imbalance as a One-Tick-Ahead Price Predictor.

Research 2: The Price Impact of Order Flow Imbalance
Cont, Kukanov, and Stoikov (2014) showed that short-term price movements were closely associated with changes in order flow at the best bid and ask.

Their work identified an important relationship:

\[ Price\ Impact \propto \frac{OFI}{Market\ Depth} \]

This helps explain why order flow signals may be more influential when liquidity is limited.

The study also highlighted potential applications of OFI in analysing adverse selection and intraday volatility.

Reference: The Price Impact of Order Book Events.

Research 3: The Microprice
Stoikov (2018) investigated the use of order book information to estimate future prices.

His research showed that the model-based microprice was empirically a better predictor of short-term prices than the midprice or simple weighted midpoint.

The result demonstrates that the distribution of orders around the best bid and ask can provide information beyond the conventional midpoint.

Reference: The Micro-Price: A High-Frequency Estimator of Future Prices.

Research 4: Order Imbalances in the Chinese Stock Market
A December 2025 study by Zhang, Xie, and Wang examined intraday return predictability using high-frequency Chinese stock market data.

The authors found that order imbalances positively predicted returns over 5–30-minute horizons.

Interestingly, the predictive relationship reversed over 60–120-minute horizons in their sample.

The reversal was not observed in large-size stocks, high-turnover stocks, and specially treated stocks.

This suggests that order imbalance can have different implications depending on the trading horizon and characteristics of the underlying market.

One important qualification: this research examines broader order-imbalance measures, so its findings should not be treated as universal evidence for every top-of-book OBI implementation.

Reference: Zhang, Xie & Wang (2025) — Do Order Imbalances Predict Intraday Returns? New Evidence from the Chinese Stock Market.

What can we conclude?
The evidence supports three important observations:

Order book imbalance can contain information about the probability of short-term price movements.

Order flow imbalance can help explain the relationship between trading events, available liquidity, and price impact.

Predictive effectiveness depends on market characteristics, ex*****on conditions, and the forecasting horizon.

Statistical predictability does not automatically translate into a profitable trading strategy.

Transaction costs, latency, adverse selection, and market impact must still be considered.

7. How Institutions Operate Behind the Visible Order Book
One of the biggest limitations of Level 2 data is that the displayed order book does not reveal the complete intentions or inventory of market participants.

Large institutional orders are often executed through algorithms designed to reduce market impact.

Two important mechanisms are iceberg orders and order splitting.

A. Iceberg Orders
An iceberg order displays only part of its total quantity.

For example, an institution may want to buy 50,000 shares but display only 1,000 shares at a time.

As the displayed portion is executed, additional quantity may become available according to the order's ex*****on rules.

Other market participants may observe repeated replenishment at the same price level.

However, repeated replenishment is not conclusive evidence of an iceberg order or a particular institution's activity.

B. Order Splitting
Large orders can be divided into smaller ex*****ons over time.

Ex*****on algorithms may distribute child orders across different price levels, venues, or time intervals.

This reduces the need to expose the entire order to the market at once.

For an observer studying OBI, this creates an important limitation.

The visible order book represents only the displayed portion of available liquidity.

It does not necessarily reveal the full quantity an institutional participant intends to execute.

C. Hidden Liquidity and Off-Exchange Trading
Some trading activity takes place through mechanisms that are not fully displayed in the visible order book.

Depending on the market, this may include reserve orders, dark pools, internalised ex*****ons, and other off-exchange transactions.

Consequently, Level 2 data should be interpreted as a partial representation of market supply and demand.

The order book reveals displayed intentions and observable changes—not the complete inventory or intentions of every participant.

8. Spoofing: When Displayed Imbalance Becomes Misleading
An order book imbalance can sometimes provide misleading information.

One reason is that visible orders can be cancelled before ex*****on.

Some order cancellations are completely legitimate. Traders routinely modify quotes as market conditions change.

However, submitting orders without an intention to execute them, with the purpose of creating a false impression of supply or demand, can constitute spoofing.

For example, a participant might place unusually large sell orders to make the ask side appear heavily supplied, hoping that other participants will respond by selling.

The displayed orders may then be cancelled before ex*****on.

This can create a false impression of market pressure.

The US Commodity Futures Trading Commission specifically identifies the creation of false market depth through bids or offers as an example of conduct that may constitute spoofing, depending on the circumstances and intent.

Reference: CFTC — Spoofing: Federal Register Rule.

For quantitative traders, this reinforces the need to distinguish displayed liquidity from executed trading activity.

OBI should be monitored alongside actual ex*****ons, cancellations, replenishment, and price response.

9. Practical Applications: Turning OBI and Microprice Into Trading Information
For quantitative traders, order book metrics can serve several different purposes.

A. Short-Term Directional Analysis
OBI can be used as one input in estimating the likelihood of the next mid-price movement.

For example:

Strong positive OBI may increase the model's estimated probability of an upward mid-price move.

Strong negative OBI may increase the estimated probability of a downward mid-price move.

These relationships must be calibrated for the specific market and instrument.

B. Market Making
Market makers can use microprice estimates to adjust their quotations.

If the estimated short-term price is above the midpoint, a market maker may consider adjusting the bid and ask to reduce adverse-selection risk.

If the estimated price is below the midpoint, the reverse may apply.

The objective is to quote in a way that better reflects the information contained in the order book.

C. Algorithmic Ex*****on
Ex*****on algorithms can incorporate OBI and OFI when deciding when and where to submit child orders.

For example, an algorithm executing a large buy order may prefer to participate when order book conditions are relatively favourable, while avoiding aggressive buying when the short-term order flow suggests adverse price movement.

The objective is not necessarily to predict the entire market direction.

It is to improve ex*****on quality.

D. Liquidity and Adverse-Selection Monitoring
Large displayed imbalances can alert traders to situations where the order book is vulnerable to short-term changes.

Combining imbalance with depth, spread, and actual order flow can provide a more complete view of ex*****on conditions.

10. Building a Quantitative Order Book Strategy
To transform these concepts into a systematic trading framework, a quantitative researcher needs to establish precise rules.

A practical development process can follow six stages.

Step 1 — Collect High-Frequency Data

Capture order book snapshots and event-level information, including:

Best bid and ask prices.

Bid and ask quantities.

Multiple levels of market depth.

New limit orders.

Marketable ex*****ons.

Order cancellations.

Step 2 — Calculate the Features

Construct:

Top-of-book OBI.

Multi-level weighted OBI.

Order Flow Imbalance.

Queue-weighted midpoint.

Model-based microprice, where sufficient data are available.

Spread and market-depth measures.

Step 3 — Define the Forecasting Horizon

Specify precisely what the model is predicting.

Examples include:

Next mid-price movement.

Next one-tick price change.

Mid-price return over 1 second.

Mid-price return over 10 seconds.

Short-horizon ex*****on price improvement.

Different horizons may require different features and models.

Step 4 — Develop the Predictive Model

Potential methods include:

Logistic regression.

Gradient boosting.

Random forests.

Time-series models.

Neural networks.

Start with simple statistical models before introducing additional complexity.

The purpose is to determine whether more sophisticated methods provide genuinely incremental predictive information.

Step 5 — Validate Out of Sample

Evaluate the model using chronological training, validation, and testing periods.

Avoid random data splitting that allows future information to leak into training.

Measure:

Predictive accuracy.

Calibration.

Precision and recall where appropriate.

Performance across different liquidity conditions.

Stability across instruments and trading sessions.

Step 6 — Test Economic Value

A model can predict the next price movement statistically while still losing money after ex*****on costs.

The final assessment must include:

Bid–ask spreads.

Commissions and exchange fees.

Slippage.

Market impact.

Latency.

Queue priority.

Adverse selection.

Realistic order ex*****on assumptions.

Only after these factors are considered can predictive accuracy be evaluated as a potential trading advantage.

11. The Limitations of Order Book Imbalance
Although OBI and microprice are useful analytical tools, several limitations must be recognised.

The visible book is incomplete. Hidden orders, reserve liquidity, and off-exchange transactions mean that Level 2 data cannot reveal all available supply and demand.

Displayed orders can change quickly. Cancellations and quote updates can make an imbalance disappear before ex*****on.

Predictability is horizon-dependent. A signal useful for predicting the next mid-price movement may not remain useful over several minutes or hours.

Liquidity changes the impact. A given imbalance can have different consequences depending on available market depth.

Ex*****on costs can eliminate the edge. Spread, fees, latency, and market impact can turn a statistically significant signal into an unprofitable strategy.

Market-specific calibration is essential. Results from Nasdaq equities or Chinese stock markets should not automatically be transferred to futures, FX, commodities, or other trading venues.

The most important lesson is that an order book signal must be evaluated within the market structure in which it operates.

Conclusion: The Order Book Contains More Than Price Information
Price charts show the transactions that have already occurred.

The limit order book provides an additional perspective: the displayed liquidity currently available around the market price and how that liquidity changes over time.

Order Book Imbalance quantifies the relative distribution of displayed bid and ask quantities.

Order Flow Imbalance captures changes in supply and demand caused by orders, ex*****ons, and cancellations.

Microprice uses order book information to construct a more responsive estimate of short-term price conditions.

Together, these concepts help explain how liquidity, order flow, and market depth interact in the process of short-term price discovery.

But the real advantage does not come from simply observing a large bid or ask.

It comes from measuring the information consistently, understanding its limitations, and testing whether the resulting signal provides value after realistic ex*****on costs.

Price tells you where the market has traded. The order book tells you how displayed liquidity is positioned around the current price.

Understanding both provides a more complete view of modern market microstructure.

Research references
Gould, M. & Bonart, J. (2016). Queue Imbalance as a One-Tick-Ahead Price Predictor in a Limit Order Book.

Cont, R., Kukanov, A. & Stoikov, S. (2014). The Price Impact of Order Book Events. Journal of Financial Econometrics.

Stoikov, S. (2018). The Micro-Price: A High-Frequency Estimator of Future Prices. Quantitative Finance.

Zhang, T., Xie, C. & Wang, G.-J. (2025). Do Order Imbalances Predict Intraday Returns? New Evidence from the Chinese Stock Market.

Zhao, C. et al. (2025). High-Frequency Liquidity in the Chinese Stock Market: Measurements, Patterns, and Determinants.

CFTC. Spoofing: Federal Register Final Rule.

LULU | Quantitative Research & Trading

Educational content only. Order book data provides probabilistic information and does not guarantee future price movements or profitable ex*****on.

Address

81, Jalan 2/90
Taman Pertama

Alerts

Be the first to know and let us send you an email when Steson Tan posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Steson Tan:

Shortcuts

Share