In most evaluations we are guided by the material situation, but how do you keep track of it in a complicated position with a large tree and many captures? Soltis recommends the simplest way is at the end of the variation count up what’s left on the board.
Sounds simple doesn't it? Soltis gives several examples, one by Alekhine (!), where mistakes in counting pieces were made. Another problem exists when the sequence of moves involves a material imbalance because in those situations much depends on the specific nature of the position rather than on a simple piece count. It’s so complicated Soltis devotes 60 pages to the problem, more than any other subject in the book.
Here’s a good example from one of my correspondence games. Should black capture the P with 16…Bxb2 or capture the B with 16…Nxf4? Or, simply move the attacked R? I suggest setting up the position and trying to run through all the possible variations then, if possible, checking your analysis with an engine. There are some lines which result in a material imbalance, so that adds to the difficulty.