What is a good approach to identifying a pattern and a way to solve it?

(Without machine learning)

So I have a game that I am creating a bot for. It is essentially a matching-tile game, but much simpler. There are several different colors you can score with, each with their own score values (e.g. orange=1, pink=2, etc) You can score a specific color if it meets 2 conditions: 1. it has to have at least 4 of the same color adjacent to it. & 2. It has to lie within a "score" slot, of which there are 6 on the left and right of the "board" respectively. Every time you score, it removes those slots and a few new, randomly selected slots are added to the board. You can either score, or swap slots. Swapping two slots adds a few new colors to the board as well.

So basically I'm trying to design an algorithm that plays this game very well, but I'm not sure how to approach it. My initial thought is to have a model that analyzes the existing board's state and ranks it. Then consider every available option, and rank its potential result, and then choose the option with the highest rank. In practice though, this doesn't hold up very well. I am not sure how to determine an appropriate rank for an existing board. What should be prioritized the most?

Anyway, this might be a rather broad question, but I was wondering if anyone had any design ideas for the concept.