if a strategy is profitable for a few days and then loses money then you have to decide how long you want to keep running it to be statistically confident that it does not work and does not mirror the behaviour you saw in backtesting.

if it turns out you need to keep running it for a month and each day it loses 3k and you don’t have 90k to risk then you have to pull the plug.

keep in mind this is just running the strategy. colocation and data also cost money for example.