Well, interpretation is always the problem with statistics, DeCypher. There are a few possibilities in the example I gave:
- The tech itself may be imbalanced.
- A descendant tech may be imbalanced.
- The tech may be necessary to access a flexible branch of the techology tree.
- The imbalance may exist, but be intentional.
Case 1 should be easy to diagnose. It would exhibit itself as a tech with a high selection rate across several hundred games followed by descendant techs with statistically acceptable selection rates.
Case 2 will show a string of imbalanced techs until a final tech, at which point it should revert to Case 1.
Case 3 is probably the most problematic case, because it means a large portion of the content is going unused.
Case 4 is often bad game design if it occurs at an early stage. Presuming people are expected to take a certain tech at a much higher rate than its alternatives, why spend time on the alternatives at all? Development resources are finite.
Consider a hypothetical tech tree:
There are a few problems here. A vs. B reveals a problem like Case 4. The B branch is being neglected in favor of the A branch. In the second tier of technolgoies, Aa is individually overpowered; the Aa1 and Aa2 descendant techs are both used evenly. Aa may be the tech which is causing the B branch to be neglected. Similarly, Ba is grossly overpowered compared to Bb, but doesn't seem to be the "terminal" option--Ba1 is also overused compared to Ba2. At a later tech tier like this, the 7:3 ratio is more acceptable than at the foundational tier, however, so this may be desired behavior in which Ba2 serves a special purpose... but Ba vs. Bb is still problematic.