Assuming you're talking about analytical geometry and not rare things like that D in AGC: Library functions are a must. A lot of fails in geometry problems are due to not knowing a correct formula or missing special cases and standard stuff to compute like intersection of segments appear very often. There's just a few algorithmic ideas (e.g. "rotating calipers") and they aren't all that hard.

A common trick with non-convex inputs is reduction to convexity. If you can replace a polygon or set of points by their convex hull, it has some nice properties that you can use.

Apart from that, play around with drawing everything related to these problems to get a feel for it. That's one nice thing about geometry, you can gain a lot of insight into it with pen and paper.

Not really an answer, but the tip that helped me a lot (maybe you know it already): use matrices to formulate and solve all necessary equations. That makes math much cleaner and handling various cases (like number of solutions) easier quite a bit.

Hey, I like this book. https://vlecomte.github.io/cp-geo.pdf it explains a lot of things and has many codes explained. Also there are some interesting post in codeforces about computational geometry. Maybe you can find them and learn from them. Good luck.

You could create a template with basic operations over lines, circles, polygons, segments and points. If you create it by your own you will learn a lot, then you may try to test and improve it.

Read "geometry concepts" tutorials in TC https://www.topcoder.com/community/competitive-programming/tutorials/

There are a couple of things you have to know:

1) how IEEE floating-point works, and especially, when you can use a naive formula and still get the exact answer, vs. when you have to implement formulas using integer/rational arithmetic;

2) a few basic geometry formulas, for example:

• segment-segment intersection
• segment-circle intersection
• polygon area
• is a point inside a polygon?
• lines tangent to two circles
• circles tangent to various things (three points; three circles; two circles and a line; etc)

Fair warning: many formulas you'll find online are not fully correct or robust (they ignore some corner cases, e.g.). It's worth looking up high-quality solutions to these basic queries (or even better, if you have time, deriving them yourself), writing them down in your reference documents, and then practice using them.

3) a few basic algorithms and techniques, for example:

• angle sweep
• convex hull
• rotating calipers

These, combined with careful thinking about corner cases and some on-the-spot derivation, will get you 75% of the way there when it comes to geometry problems. For example, "Airport Construction" from WF 2017 requires only the basic formulas and some small insights (i.e., that the longest runway must pass through two polygon vertices), but is hard because you have to be very methodical and precise to solve the problem robustly (handling cases like multiple colinear vertices, non-convex island shapes, runways coincident with some of the island edges, etc.)

Of course, there are plenty of other algorithms that show up every now and then, and it's cliched but true that the best way to get better at geometry problems is to simply practice more problems, taking note of any algorithms or formulas you wish you would have had in advance.

I learnt a lot from this blog

You might want to start with the Geometry section of this website.