Hello, I was wondering, if and how common it is that quant traders collaborate on building blocks for their strategy with peers? (Inside their organization, or independent traders with each other) For example the following building blocks: latency compensation between different data feeds. Different iceberg estimators for market by order, market by price, to top of book to NBBO feeds. Specific behaviour tracking of big order placements and migrations over time in the Orderbook. How to statistically remember a big quote for venues where depth data is not bought yet or not available. I would asume a quant in a trading firm can just use a framework of cleaned data with a depth book that already includes the static possibilities of detected icebergs and so on. In a firm it is also easy to share, because everyone uses the exact same datafeed. I try to do everything on my own right now, but regarding all the possible aspects one trading decision could consider, it will take some time to test them out and find hopefully working combinations. My question would be: Do you do every little step of the data to strategy building process yourself? Or do you collaborate also with other individual traders or peers on some of your algo building blocks?
I perform variants of anomaly detection (ML technique) as part of my toolbox to generate new ideas that I subsequently test.
Thanks for your answers, motivating to know that you achieved it by your own. I guess there is no shortcut in building the personal tool box. And the thought that everything would be easier once I am able to access historical depth might be theoretical true, but theoretical if no goal is achived with lower resolution data first. Regarding my latency problem I just added some artificial latency now, that way I have time to order the events and can garantie that my backtests and real time app can see frames in the exact same order. The other question would be: Do you find out everything by trail and error or did you read useful documentation or got the knowledge by good books? (For example, I have mistaken the NBBO to send quote size updates for a longer time. And thought it would be an error and few quotes are just missing. But with the knowledge, that is not an error but designed that way by the data vendor in combination with the added artificial latency and "right" order and window size of events. My Quote sizes are alot cleaner an more correct then in the past.)