In their book, The Woodpecker Method, Axel Smith and Hans Tikkanen give these percentages of decisive games decided by tactical mistakes:
The percentages in the three rating bands below GM level are based on 32 randomly chosen games in each category
"32 randomly chosen games in each category" seems a rather small sample size for determining this information. Does anybody have more comprehensive results based on (much) larger sample sizes?
As I dipped below 1800 a few years ago, I would also be interested in the 1600-1800 category.