← sources

Better analogies than book/sleeve for a multi-strategy system

source: reddit (reddit:algotrading)published 2026-09-17 13:44ingested 2026-09-18 02:42

retained text

I've been working on a massive project for the better part of a year now, and it's been iterative with many improvements made along the way, but recently I decided to take a step back and evaluate how the full book of strategies operates and how to improve the process going forward so I can backtest new ideas, variants on existing strategies, etc. and I came up with an analogy that I like. (Yes, I'm using AI tools to code, and this is how I was able to convey my idea to Claude and Codex...) 

Originally it was just book and sleeve. The book is the full suite of strategies and each strategy is a sleeve, but when you want to go back and make changes to a sleeve does it become a new sleeve? If you want to set limitations for certain sleeves that perform best in RTH hours and others that are allowed to run overnight, how do you classify those criteria? 

The new analogy I came up with was building a gaming console. The console is the architecture that can load and play a strategy, so each strategy becomes a cartridge that is played on the console. This gives us a set of rules that are needed to allow the cartridge to play and every strategy needs to have this rule set. 

From there every cartridge can also have an options menu and you can have different presets or playable characters that you can use to play a level on that cartridge. Those settings can be saved and each character and options menu choice that you choose are what is used for the walk forward and replay to confirm the results, and the score that you get. 

Each daily session is a level that needs to be cleared and the state of the market along with any of the defined dimensions used by the cartridge are saved in a 'save state' that can be quickly loaded to resume play when you move on to the next level. This avoids pre-warming and can know exactly where you left off. The importance of this shows up in the replays and backtesting. It takes a lot more time to replay a single strategy over a long period of time (I'm using MBP1 data so it replays each tick) and then do the same thing over and over again for each strategy and variant when testing different parameters. The faster method is to use these save states and load a single session (level) and then play each cartridge and character for that level, save the game, and then load the next level and repeat. 

Does anyone else do something similar? How has it worked out for you if so? 

document fetched 2026-09-18 00:45 · canonical https://reddit.com/r/algotrading/comments/1wiuf03/better_analogies_than_booksleeve_for_a