I have tried both: a home fibre “gaming” setup and a cheap cloud VPS near the broker’s matching engine city. The honest answer for discretionary M1 work: location and stability usually beat raw download speed at home, but a bad VPS is worse than a clean local connection.
What I measure:
- Ping / route consistency to the broker endpoint (not speedtest to random CDN)
- Platform freeze count during London open
- Fill quality on the same pair/session for two weeks each path
- Cost of the VPS vs any improvement in median slip
What have you measured that actually changed your setup?