What varies
Fill quality, requote behaviour, latency to the matching engine, whether the counterparty internalises your flow or passes it through, and what happens to your stop during a gap. None of these appear in a strategy description and all of them appear in the returns.
Distance is measurable
Round-trip latency from a retail connection to a broker's matching engine is commonly tens of milliseconds; from a co-located instance it can be single digits. Whether that difference matters is entirely a function of holding period — it is decisive for a scalper and irrelevant for a carry book held for weeks.
The check worth running
Run the same strategy on paper against two brokers simultaneously and compare fills, not returns. Returns are noisy over short periods; fill quality is visible almost immediately, and it is the thing you are actually choosing between.