I think the trick is, don't rely on any character as infallable. Three-toeds are one of the most variable turtles out there, both in color and shape, which makes sex determination by morphology difficult in some specimens. All of the other U.S. varieties of this species seem (to me) to show a little more gravity in their dimorphisms, although easterns can occasionally be confusing as well.
The good ole concave/convex plastra, tail length, and eye color characteristics do hold up sometimes, but not others. There are plenty of females out there that have eyes that are much "redder" than those of many males. Similarly, males can have totally white eyes (excepting the pupils); which category (brown/red) does that fit into?
Claws on the hind limbs are longer in some males, carapaces are less aerodynamic-looking in some females, but, all of these are characteristics that don't hold up with enough consistency to prove reliable.
The easiest way to tell is, once you've seen enough of them, you get a "feel" for gender determination, and, you still make mistakes. The best approach is likely to take as many of the (aforementioned) characteristics and weigh them against one another collectively rather than relying on any one or two alone.
To make matters even more difficult, even the turtles don't always initially judge a conspecific's gender properly (although actions such as below can sometimes represent phases of male/male combat). The two below did have it right.
