Fun read. I've felt that MM in less liquid markets are either domain specialists I'd hesitate to trade against, or bots I would be happy to challenge. Did you believe that it was the latter (or perhaps the manual flavor of uninformed MM)? What happened to the order book when you lifted the offers?
Without the liquidity rewards I would have given the size displayed much more credit. There wasn't much reaction in the orderbook right after my trades and all the sizes across markets were similar. There are def domain specialists out there but the incentives lower P(specialist).
I hinted at this in another article but I obviously can't mark my trades to T+30min - i just had market impact of course the trades look good short-term.
If it was a domain specialist I likely lost ~$0.005+fees on average. Definite possibility but I thought odds in my favor.
That is a great read! Thnx for sharing the details. Facilitating to see the actual mechanics to get real profit. Indeed choosing the right table is critical. I’ve done some llm agent for crypto trading for https://trade-bench.live/. Idea is to give an agent a portfolio and ask it to rebalance it according to latest price move and news. Goal of project is to evaluate bunch of models to see who is better at market prediction. Your case is much more focused and nuanced. Love how you merged your judgment and llm capabilities
Fun read. I've felt that MM in less liquid markets are either domain specialists I'd hesitate to trade against, or bots I would be happy to challenge. Did you believe that it was the latter (or perhaps the manual flavor of uninformed MM)? What happened to the order book when you lifted the offers?
Without the liquidity rewards I would have given the size displayed much more credit. There wasn't much reaction in the orderbook right after my trades and all the sizes across markets were similar. There are def domain specialists out there but the incentives lower P(specialist).
I hinted at this in another article but I obviously can't mark my trades to T+30min - i just had market impact of course the trades look good short-term.
If it was a domain specialist I likely lost ~$0.005+fees on average. Definite possibility but I thought odds in my favor.
There's a lot to think about here, really. Thank you for this analysis, thank you for this article.
That is a great read! Thnx for sharing the details. Facilitating to see the actual mechanics to get real profit. Indeed choosing the right table is critical. I’ve done some llm agent for crypto trading for https://trade-bench.live/. Idea is to give an agent a portfolio and ask it to rebalance it according to latest price move and news. Goal of project is to evaluate bunch of models to see who is better at market prediction. Your case is much more focused and nuanced. Love how you merged your judgment and llm capabilities
Excellent overview...ty
Interesting read