I first really understood this idea through World of Warcraft, particularly the talent tree system that was revised heavily in the Cataclysm expansion and then even more drastically revised in Mists of Pandaria. Due to the unpopularity of Cataclysm, the company began publishing blogs by the design staff to explain many of the changes they’d made to the system in that expansion. Players were now putting in some 31 points into these talent trees (down from 70) selecting minor stat upgrades and accessing new abilities, so there were a lot of possible combinations.
The original talent tree
The trimmed down talent tree
The game’s community and through its fan sites published what they had determined through their own number crunching what were the most effective point allocations. Anyone who played the game mildly seriously was aware of these and followed the pre-existing builds, leaving only novice players out of the loop and under-performing.
At this point Blizzard had acknowledged that this old model of talent tree was simply not viable, it was a failed idea and impossible to balance. The reason being was that even though there were so many choices to make, they were not meaningful choices. The best builds would always be calculated by others for you, so there was no need to make a decision. If you chose to deviate, you only retained the right to make the wrong decision, the sub-optimal decision. That was the opposite of fun.
In Mists of Pandaria, the talent trees were overhauled and pruned even further, with only 6 talent points. Instead of 3 separate trees some 7 levels deep, there was a single tree 6 levels deep. At each level, players made 1 selection, choosing 1 talent from a selection of 3. Each level was grouped around a similar idea, so it presented different ways to approach the same idea, meaning the choices were easier to balance with fewer options and the emphasis on creating different approaches meant you could pick the one that suited your play style.
Current talent tree
This new style isn’t perfect as there are always mathematically optimal builds, but the numerical difference between the right and wrong choices is often negligible and only the top players pretend to see the 1-2% difference in performance. In other cases the choices compare apples to oranges, instead of giving a choice of damage boosting abilities, you choose between 2 unrelated support abilities – do I get a small movement speed boost or the ability to stun an opponent. In those cases, there is no right or wrong answer, so the choice retains meaning. Alternately, because each choice performs in a different way, they excel in different situations, so the game encourages you to swap these around for specific boss fights, treating your talents not as some set in stone set of bonuses, but as a toolbox you reach into to adjust for each situation.