A 1/10 chance that a member of any given breed exhibits any given behavior based on breed doesn't seem unreasonable given a) how granular the behaviors are (the

A 1/10 chance that a member of any given breed exhibits any given behavior based on breed doesn't seem unreasonable given a) how granular the behaviors are (they include "eating grass"), so there are a large number of specific behaviors that might all fit a larger pattern but which any given dog might or might not exhibit, b) that they included mixes, who you would expect to be less likely to fit a breed's specific behaviors, and c) the 9% predictiveness is on top of a base rate of exhibiting the behavior regardless of breed.