I should belatedly add that, while those links go into great depth, a relatively shallow tool-level understanding of transformations, vectors, and other processes can go a very long way in game development and physics engine usage.
You don't need to be able to manually scribble a projection matrix in order to use a Matrix helper function to create one. If you understand the purpose of a projection matrix- to transform from view space into clip space- the numerical details are secondary. Understanding how the matrix and vectors form a system of equations is a nice aid in understanding how such things are constructed, but you don't need to get bogged down.
Here's a non-exhaustive list of helpful tools and tricks you might see repeatedly:
-Dot products. Are these vectors facing the same direction (dot product > 0)? What is the angle between these two normalized vectors (Math.Acos(dot product))? How fast is this object moving in normalized direction vector V (Vector3.Dot(V, entity.LinearVelocity))?
-Cross products. What is a vector in the direction perpendicular to vectors X and Y (Vector3.Cross(X,Y))? What's the area of this triangle with edge vectors X and Y (Vector3.Cross(X, Y).Length() / 2)? What is the normal of this triangle with edge vectors X and Y (Vector3.Normalize(Vector3.Cross(X, Y))? What is the angle between two normalized vectors (the dot product can do it too, but the cross product's version is slightly different in its domain and usage)? How does the order of cross product parameters affect the result (Vector3.Cross(X, Y) == -Vector3.Cross(Y, X))?
-Properties and identities associated with the dot and cross products can come in handy, particularly in optimization. The Scalar Triple Product is seen fairly regularly. Knowing how familiar scalar operations like multiplication interact with dot products is very helpful for intuiting correct behaviors. And many, many other little things.
-Projections and primitive geometric operations. Orthographically pushing points onto a plane (remove the component of the position along the plane's normal). Intersecting a ray and a plane. Distance between lines (it might appear at first like it requires minimizing with a pair of parametric equations, but the application of the cross and dot Tools makes it simple
).
-Rotation matrices are orthonormal bases. They have up, down, left, right, forward, and backward vectors. You can use those vectors to rotate around an object's local X axis or move along an object's local Y axis and other fun stuff. The matrix's Up vector is the value you'd get if you transformed Vector3.Up by the rotation matrix, but the property can read it right out of the components of the matrix!
-The hierarchy of common transformations. Rotations, translations, combined rigid transformations, and less commonly scaling, shear, linear transformations in general, affine transformations (translation with linear transformation). The little tricks you can perform when you have guarantees about what a transformation is can simplify work quite a bit. For example, instead of inverting a pure rotation matrix as if it were a general matrix, transpose it!