On Mikrotik, BGP policy is expressed with route filters under /routing/filter/rule [1]. The syntax works sort of like a mix of the firewall rule chains and a pattern matching language. I’m not exactly sure of the context in which you’re running BGP, but it’s usually good practice to have inbound and outbound policies with any BGP neighbors that you don’t control. The idea is to ensure that you’re only announcing and accepting what you intend.
The client side router can implement two filter chains, one for each peering. One chain accepts anything and sets localpref to, say 200 (up from the default of 100). The other accepts the same but sets localpref to, say, 50. You can get much more granular than this, but this should influence the BGP route selection algorithm [2] to ensure that only the routes you want are installed. ECMP is only happening because the two routes have the same weight, localpref, AS path, origin, and MED plus were both learned from eBGP. Adjusting any of those metrics should cause one route to be preferred over the other(s).
It might be tempting to adjust weight but that’s local to the device, whereas localpref can be set once and will be automatically be preserved within your AS.
[1] https://help.mikrotik.com/docs/display/ROS/Route+Selection+and+Filters
[2] https://help.mikrotik.com/docs/display/ROS/BGP#BGP-Best-PathSelection