EM() as the default solver?

[oameye]

Why do we use EM() as the default solver? From the example of StochasticDiffEq.jl docs this method seems to be prone to errors. The same can be said for Euler() as the deterministic solver.

[reykboerner]

Indeed Euler-Maruyama can be inadequate especially for large step sizes, but it is still so commonly used that one could argue it is still a reasonable default solver. Users would potentially be less familiar with the other SDE solvers. Which one would you use?

For ODEs I argee that one could replace Euler by something like RK4() or DP5().

[oameye]

DP5() is dangerous as it isn't useful for damped systems. I would use the default solvers that SciML uses which would be Tsit5() and SRIW1() I think.

[reykboerner]

Hmm, Tsit5() caused some problems when used with DynamicalSystems.jl (see this issue). But I agree that it's best to go with established defaults if we don't stick with the most basic defaults.