OrderbookEngine

A fully functioning LOB + matching engine implementation written in Java. 3 basic matching algorithms are implemented (price-time, pro-rata, pro-rata with top) in the MatchingEngine package.

Orderbook implementation:

• In each limit, maintain a doubly linked list to store individual orders. Provides constant time addition and deletion of orders
• Each individual limit level is stored in a balanced binary tree (TreeSet). The orderbook maintains two of these trees, one for buy limits and one for sell limits.
• OrderId: Order HashMap
• O(1) access for bestBid/Ask, bestBidSize/AskSize, totalBidSize/AskSize

Matching implementation:

• IOrderMatcher interface - each matcher must implement matchMarketOrder and matchAggressiveLimitOrder.
• matchMarketOrder can deplete multiple limit levels
• matchAggressiveLimitOrder can only deplete the far touch limit before creating a new passive order at the far touch for the remaining quantity.
• Price-Time: Standard price-time priority matching.
• Pro-Rata: https://atas.net/volume-analysis/basics-of-volume-analysis/cme-order-matching-algorithms-part-1/
• Threshold Pro-Rata: https://atas.net/volume-analysis/basics-of-volume-analysis/cme-order-matching-algorithms-part-3/

Simulator implementation:

• At any iteration, there are 8 events (Event.java) that the simulator can pick from with defined weights (in eventProbabilitiesDefault.properties)
• These 8 events are: PASSIVE_BUY/SELL, AGGRESSIVE_BUY/SELL, MOD_BUY/SELL, CANCEL_BUY/SELL.
• If an AGGRESSIVE_BUY/SELL event is selected, there is a 50% chance it will be a market order or an aggressive limit order.
• If any of the other 6 events are selected, we need to determine the limit price level for that event. E.g. If PASSIVE_BUY is chosen, we need to choose a bid price for the new passive order. To do this, an array of decaying probabilities is generated, where the ith element represents the probability we select a price level that is (i+1) ticks away from the opposite best quote (see OrderbookSimulator.generateDecayingProbabilities).
• The simulation is initialized with INIT_ITERATIONS (simulationConfig.properties) interations, where only PASSIVE_BUY/SELL events can be chosen to build up the orderbook. No matching/trades can occur during this phase. Afterwards, the simulation is run for ITERATIONS iterations.

To run:

1. Install maven
2. Run mvn clean install -DskipTests assembly:single to create an all in one jar.
3. Run java -jar target/OrderbookEngine-1.0-SNAPSHOT-jar-with-dependencies.jar simulationConfig.properties 123. The first argument is the config file which can be found in src/main/resources, and the second argument is a random seed value for the Random class. The defaults are simulationConfig.properties and 123 if no arguments are specified.
4. A log file will be generated for each run, located in logs/.

TODO:

Implement a light orderbook + BBO visualization program (python) based on the simulation log output

Improve the simulation - Poisson distribution to model time between events instead of using iterations (Watts/Delft paper in pdfs/)

Implement concurrent features