Every time someone mentions IQ tests the conversation immediately goes to matrix puzzles and shape rotation. But I have seen breakdowns of actual clinical assessments and they seem way more varied than that. What are the different types of tasks on a legitimate IQ test, why were they chosen, and is there a coherent theory behind what gets included or is it just whatever seemed to correlate with being smart?
Real IQ batteries are much broader than the matrix puzzle reputation suggests and the task selection follows actual theory. The CHC model identifies distinct cognitive abilities through factor analytic research and modern assessments are built around it. You typically get verbal tasks like vocabulary and analogies, fluid reasoning tasks like matrix puzzles and figure weights, working memory tasks like digit span and sequencing, and processing speed tasks like symbol coding under time pressure. Each type is included because it loads on a known broad ability factor and adds unique variance to the g estimate. RIOT IQ covers 15 subtests across 6 CHC domains which is broader than most clinical batteries and includes reaction time measurement that traditional tests often skip entirely.
The matrix puzzle reputation exists because fluid reasoning is the most g loaded task type on most batteries 
But building a test entirely out of matrix puzzles would give you a narrow and unreliable g estimate. The variety across different task types is what makes the overall score meaningful.