make bad and old examples more comparable

[vsemozhetbyt]

See the comment of @ryanmcdermott.

With this PR:

1. Bad and old examples become more comparable (let bad examples be bad only in regard to one appropriate case).
2. Both examples do not contain type coercion (i.e. they use strict comparing).
3. The new algorithm has less performance difference in small and large arrays comparing (some benchmarks have been done).

[ryanmcdermott]

Even smaller, and probably faster time complexity! Thank you!