The knowledge you need to come up with the score is the same knowledge you need to balance the game. Just because
you’ve devised a simple sounding deliverable doesn’t
mean the problem is simple.
The whole challenge of balancing a game is that you’re forever wrong about that score.
Underestimating the score of a single unit (or a more complex combo of multiple units) is precisely how you accidentally create an imbalanced metagame where that unit/combo can be spammed.
That makes sense. Doesn’t that then seem like a perfect problem for a neural net type system to solve? Watch a bunch of games, figure out empirically how effective different units and combos are and try to come up with a scoring rubric that is incrementally more accurate than before.
I am not trying to trivialize the problem at all. I don’t know much about it and that’s why I’m asking why it isn’t simple.
A large issue with this is the fact that certain armies of different strengths can only exists in certain time-frames, so even if you could perfectly calibrate for "this army combo is as strong as every other", the fact that one unlocks parts of these at different times makes this approach totally unrealistic. Since the unlocking of army combos is extremely rooted in economy, map layout/resources, power spikes of opponents ect.
> That makes sense. Doesn’t that then seem like a perfect problem for a neural net type system to solve?
In theory. However if you change the balance then the mega game will change as well and you have to throw away the training data and start anew.
Also it’s very important to not change too many stats at the same time because then it will be hard to learn the game and the new meta will take too long to emerge.
I think I’m starting to see the picture. Since we so far don’t have an AI capable of playing as well as the top human players, in order to figure out what’s out of balance you have to watch a sufficient amount of PvP games. And then you rebalance and must start the process from scratch. And unless you hit the perfect balance by accident you will always need to keep rebalancing.
The only other way I can think to do this would be to either change units incredibly quickly to not allow players to get too far into any particular strategy development before the unit is retired. Or introduce things like variations between units: two different marines might have different stats to not allow their 0.2 DPS advantage to add up, etc. Or I guess introduce acts of god type situations that might throw off a specific strategy. Napoleon supposedly lost at Waterloo partially because of a stomach upset that required him to keep running to the latrine and unable to focus on the battlefield. Imagine if you suddenly got fog of war over the area of combat because of “broken communication”. Basically more variability that would require the players to not concentrate on a single strategy and instead having plans A through Z for various eventualities. I’m not sure if that would make for a more exciting experience or not but might be fun to experiment with.
Some multiplayer games (like hearthstone) introduce a lot of randomness and quick variability to keep the game entertaining for new players. Even as a newbie you can win say 30-40% of games against skilled players. However, this also means that skill is only shown in long term and thus you attract a different crowd.
Much like in VS fighting games, people like that time and training translate to good skill.
The whole challenge of balancing a game is that you’re forever wrong about that score.
Underestimating the score of a single unit (or a more complex combo of multiple units) is precisely how you accidentally create an imbalanced metagame where that unit/combo can be spammed.