Most systematic strategies are born in a notebook. It’s the right tool for the job: load some data, sketch an idea, plot a curve, iterate fast. The notebook is wonderful for exploration. It’s also where a lot of strategies get stuck, because the qualities that make it great for research make it terrible for operating. Here’s where the path to live actually breaks.
Break 1: the run isn’t a thing you keep
A notebook cell produces output, then you change it and the old output is gone. There’s no durable record of what you ran, with which parameters, on which data. The first time you want to compare two runs or reproduce one from last month, you discover you can’t. Research is fine with this. Operating is not.
Break 2: there’s no real-time anything
Notebooks run a block and stop. A live strategy is a long-lived process reacting to a market in real time. Nothing about a notebook is built to stay alive for a session, stream live data, and react as events arrive. You can’t incrementally “turn a notebook live”, it’s a different shape of program.
Break 3: the broker doesn’t fit in a cell
Backtesting in a notebook needs no broker. Going live needs order routing, fill tracking, position reconciliation, and an account that’s the source of truth. None of that lives naturally in a notebook, so you end up bolting on scripts and services around it until the notebook is just one increasingly awkward piece of a system it was never meant to anchor.
Break 4: nobody’s watching
A live strategy needs monitoring, logging, and a way to stop it. A notebook offers none of these as first-class things. So you either fly blind or build the monitoring yourself, at which point, again, you’ve left the notebook behind.
The pattern
Notice that every break ends the same way: and then you build the thing the notebook isn’t. The notebook didn’t get you most of the way to live; it got you a great research artifact and a clear view of how much is missing.
The useful move is to keep the notebook for what it’s best at, exploring ideas, and stop trying to make it the home of a running system. The strategy is yours; the run record, the real-time execution, the broker handling, and the monitoring are a different job, and pretending the notebook can stretch to cover them is exactly where the workflow breaks.