I've built an simulated Electronic Exchange with its own orderbook, completely isolated from the real financial markets. I want to use my creation for a game with a few friends, but I have to provide liquidity. I've been looking around the internet for algorithimc market making strategies, buit they're all pretty complicated (way beyond what I need for this project) and some have the opposite problem of beeing too simple (like AMM x * y = k). WHat I'm looking for is a simple (but not too simple) way to algorithimcally place orders (as a third party) on my simulated exchange following classical log depth and with the assumption that there is no historical data (So no mean reversion etc). Basiclly, generate an array of orders, distributed log-wise through market depth on a given ticker symbol while only having access to:
- A couply of previous human-placed orders
- Previous bid and ask prices (If it's not the first order)
- Previous close (If it's not the first session)
- Other pending orders
- Market Maker account balance and portfolio
NOTE: The market maker in this case is a normal user which connects to the platform and issues order programatically. It's not part of the exchange.
Also, over time historical data would accumulate, but the problem is making something which can work the first day, the seocnd and the 50th, not something which may work with 90 days of data.
Project GitHub Repo: https://github.com/Alessandro-Salerno/MC-UMSR-NSE-Market-System
I know the expanation may not be amazing, but I couldn 't think of another way to lay this all out without making it too long.
Many thanks
