How predict the knight moves ?

well maybe it's just a stupid question or a wrong point of view because i'm basically a really bad player but at the moment i make always ( not always always but almost ) the same mistake when the game is close like a draw .... and my opponent has a knight and maybe i have a Bishop or a Rook .... basically i make the king and the other piece in the wrong position and i lose the game that probably will be finished in a draw . 

well i really hope that someone can help me ... and sorry for my really bad english . Thanks  

Two basic ways to avoid knight forks like that:

1)Keep your pieces on opposite colors. Knights only attack squares of the same color.

2) Pieces 2 squares away diagonally from a knight require a minimum of 3 moves to attack, so are very safe.

King to b2 or c1 will result in a fork.
King to any light square is safe
King to d2 is the safest (knight needs at least 3 moves to check again).

In fact that configuration (2 squares diagonally) can be useful again. Note that Kc3 is safe because pieces two squares away diagonally are impossible to fork.

And a general tip about knights.
Knights alternate the color they attack on every move. So whenever your king is in check from a knight, if you move out of check diagonally, it's impossible for that knight to check you on the next move.

(p. 14)
   "How Can I Lose?"

   When you're at the board, you look for good moves--either
good moves for you to play or dangerous to you if he plays them.
It may seem like toying with fire, but you should also be looking
for bad moves. A good sense of danger, of knowing how you can
go wrong in a particular position, can help enormously.

Soltis, Andrew. 1979. Catalog of Chess Mistakes. New York: David McKay Company, Inc.

you'll get the hang of it eventually. They will always remain pesky pieces, especially with low time controls. Don't worry. You're in good company.

Hoo thanks for all the answers . 

This topic deserves more attention. The rules of thumb above are great first steps. And learning some of the theoretical endgames with a knight v pawn seems helpful.

But trying to anticipate where a knight could be in just two moves on an open board is difficult/confusing.