What's the actual difference between local and server backtests?

Local runs on your own machine using data you have locally — unlimited, fast to iterate, ideal while you're shaping an idea. Server runs on our infrastructure over a longer history than you'd want to keep on a laptop, which is what you want for a final check across many months and symbols. Practical workflow: iterate locally until the idea stops changing shape, then run it on the server once for the wider period. Server runs are queued and count against your plan's daily allowance, so they're worth spending on a strategy you're fairly confident in rather than on exploration.

"Iterate locally until the idea stops changing shape" is a good rule. I was burning server runs on half-formed ideas.

If the same strategy over the same date range gives different results local vs server, which one do I trust — or is that itself a sign my data is patchy?

for crypto it matters less for me since my local data is a mess of gaps from when my machine was asleep lol. i shape stuff locally then always redo it on server before i actually run it live

- local = drafting - server = final check - never skip the second one just because the first looked good