1.
요즘 AI가 대세인지라 시장미시구조와 관련한 글은 구닥다리로 취급당하는 듯 합니다. 국내 글은 차지하고 해외 글을 보더라도 자주 보이지 않습니다. 물론 LOB를 머신러닝과 결합한 논문은 있지만 주된 관심을 아닌 듯 합니다. 논문이 아닌 블로그나 기사로 더욱더 귀합니다.
우현히 북마크한 글을 읽었습니다.
위 글중 일부분입니다.
Understanding LOB and Its Imbalances
A LOB is essentially a list of buy and sell orders for a specific asset, organized by price level. (중략) Now, let’s talk about imbalances. In a perfectly balanced market, the number of buy orders would match the number of sell orders. But markets are rarely perfectly balanced. At any given moment, there might be more buy orders than sell orders, or vice versa. This is what we call an ‘imbalance’ in the LOB.
Why are these imbalances important? Well, they can provide insights into market sentiment. For example, if there are significantly more buy orders than sell orders at a particular price level, it could indicate bullish sentiment – a belief among traders that the asset’s price will rise. Conversely, if there are more sell orders, it could signal bearish sentiment – a belief that the price will fall.
But can these imbalances be used as a trading signal? Can they help traders predict price movements and make profitable decisions? That’s the million-dollar question – and it’s the question that the research papers we’re about to discuss have tried to answer.
LOB Imbalance. 논문을 보면 자주 등장하는 개념입니다. 여기까지는 그런가 했습니다. 위 글에서 소개한 VisualHFT의 기능을 보는데 아래 내용이 있습니다.
Advanced microstructure studies: Built‑in study plug‑ins compute microstructure analytics like VPIN (Volume‑synchronised Probability of Informed Trading), LOB Imbalance, Market Resilience and OTT Ratio. Each study listens to order book/trade events and publishes computed metrics via the trigger engine.
앞서 글을 보면 이런 표현이 등장합니다.
“LOB Imbalance”가 매매시그날로 사용할 수 있는가?
2.
위 글에서는 LOB Imbalance를 매매시그날로 이용하기 위한 이론적 배경으로 논문을 소개합니다.
Trade arrival dynamics and quote imbalance in a limit order book
글쓴이의 설명입니다. 매매시그날로 사용할 수 있지만 전제와 한계가 있다고 이야기합니다.
The authors focus on the probability of price movements and trade arrivals as a function of the quote imbalance at the top of the LOB. In other words, they’re interested in how the imbalance between buy and sell orders at the best prices can affect the likelihood of trades happening and prices changing.
To do this, they propose a stochastic model. A stochastic model is a fancy term for a mathematical model that includes some element of randomness. It’s a way of capturing the inherent uncertainty and variability in financial markets.(중략)
So, what do they find? The authors discuss the quality of fit and practical applications of the results. They find that the model can indeed capture some of the dynamics of the LOB and the trading process. The results suggest that the quote imbalance at the top of the LOB can indeed provide valuable information about the likelihood of price movements and trade arrivals.
This study provides a solid foundation for our understanding of LOB imbalances. It shows that these imbalances can have a significant impact on market dynamics and can potentially be used as a trading signal.
두번째 논문은 The Price Impact of Order Book Events입니다.
혹 기억하실 분도 계시지만 rama cont가 나옵니다. 미시시장구조를 이용할 때 자주 등장하는 분입니다. 글쓴이는 LOB Imbalance(Order flow Imbalance)를 통한 가격예측가능한다고 이를 linear price impact model이라고 부른다고 정리합니다.
The authors analyze the price impact of three types of order book events: limit orders, market orders, and cancellations. They use the NYSE TAQ data for 50 U.S. stocks, providing a broad and diverse dataset for their analysis.
One of the key concepts in this paper is the order flow imbalance. This is defined as the imbalance between supply and demand at the best bid and ask prices. In other words, it’s a measure of how much more buying or selling pressure there is at the top of the LOB.The authors find that, over short time intervals, price changes are mainly driven by the order flow imbalance. This suggests that LOB imbalances can indeed be a powerful predictor of price movements.
But the insights don’t stop there. The study reveals a linear relationship between order flow imbalance and price changes. The slope of this relationship is inversely proportional to the market depth, meaning that the impact of imbalances is greater when the market is thin and less when the market is deep.
These results are robust to seasonality effects and are stable across different time scales and stocks. This suggests that the relationship between LOB imbalances and price changes is not just a quirk of a particular market or time period, but a fundamental feature of financial markets.
Interestingly, the authors argue that this linear price impact model, together with a scaling argument, implies the empirically observed “square-root” relation between price changes and trading volume. However, they also note that the relationship between price changes and trade volume is noisier and less robust than the one based on order flow imbalance.
Order Flow Imbalance를 주제로 한 다른 글쓴이의 글도 좋습니다.
Order Flow Imbalance – A High Frequency Trading Signal
세번째 논문은 Limit Order Book as a Market for Liquidity입니다. LOB Imbalance 혹은 Order flow Imbalance에서 나아가 Liquidity Imbalance라는 개념을 소개합니다.
The authors propose a new measure of liquidity that is based on the LOB. They argue that traditional measures of liquidity, such as the bid-ask spread or trading volume, do not fully capture the state of the market. Instead, they suggest looking at the entire LOB to get a more comprehensive view of liquidity.
Their measure of liquidity, which they call the “liquidity imbalance,” is based on the imbalance between the supply and demand of liquidity in the LOB. This is calculated as the difference between the volume of limit orders on the buy side and the sell side of the book.
The authors then investigate the predictive power of this liquidity imbalance. They find that it can predict future price changes, suggesting that it can be used as a trading signal. However, the predictive power is asymmetric: it is stronger for negative price changes than for positive ones.
This study provides a fresh perspective on LOB imbalances. It shows that these imbalances can provide valuable information about the liquidity of the market, which can in turn be used to predict price movements. This adds another layer to our understanding of the potential uses of LOB imbalances in trading.
네번째 논문은 Queue Imbalance as a One-Tick-Ahead Price Predictor in a Limit Order Book입니다. 논문은 Queue Imbalance라는 개념을 소개합니다. 방향과 가격을 예측하기 위한 도구입니다.
This paper takes a deep dive into the predictive power of LOB imbalance, specifically focusing on its ability to predict the direction of the next mid-price movement.
The authors start by introducing the concept of queue imbalance. This is defined as the difference between the number of orders in the best bid queue and the best ask queue, divided by the total number of orders in both queues. In other words, it’s a measure of how much more demand there is than supply (or vice versa) at the best bid and ask prices.
이상의 논문들은 LOB Imbalance에 기초하여 Order Flow Imbalance, Liquidity Imbalance, Queue Imbalance로 확장하면서 LOB Imbalance를 시장의 앞날을 예측하는 매매시그날로 사용할 수 있음을 보여주고 있습니다.
3.
Deutsche Borse 데이타를 이용하여 LOB를 분석한 Python코드는 LOB-feature-analysis에서 참고하실 수 있습니다. 앞서 VisualHFT를 보면 미시시장구조를 분석하기 위한 지표로 LOB IMbalance외에 몇가지가 더 있습니다.
VPIN
LOB Imbalance
OTT Ratio
Market Resilence
LOB Imbalance를 한국 자본시장에 적용할 수 있을까요? 한국은 호가정보와 체결정보를 가공해서 보내줍니다. 코스콤 시세분배시스템이 실시간데이타를 가공하여 LOB를 구축하는 업무를 하고 있기때문에 감춰진 호가정보가 있을 수 있지만 LOB정보를 보내줍니다. 이를 기초로 유사한 작업을 하고자 하면 가능합니다. 100G로 시세환경이 바뀌면서 데이타의 업데이트 주기가 이벤트기준으로 바뀐 경우가 많기때문입니다.
마지막으로 기계학습과 LOB를 결합하여 분석하려는 시도가 많습니다. LOB을 이용한 기계학습 주가예측 모형들에서 몇 논문을 소개했습니다. 아래도 같은 논문입니다.
To Make, or to Take, That Is the Question: Impact of LOB Mechanics on Natural Trading Strategies
Deep Limit Order Book Forecasting
Representation Learning of Limit Order Book: A Comprehensive Study and Benchmarking
LOB-Bench: Benchmarking Generative AI for Finance — an Application to Limit Order Book Data
