You may be right in criminal cases, where the client usually knows for sure whether they "did it." That's not my area.
But what you say is 100% not correct in civil cases, particularly patent cases. My clients will ask me dozens of times over the course of a case what their chances of winning are, and each time I give them my best professional opinion at the time. Some clients even use decision tree software that forces me to assign a % likelihood to every individual issue in the case. The software then spits out a "value of the case" based on the various outcomes and the likelihood of each.