Really nerdy thread inc
People often make claims about power levels – “Arthas was far more powerful than Bolvar,” for instance.
We mostly all know that these claims are kind of silly, but they are fun, and they can be useful for some discussions. I thought for this thread I’d look at some of the challenges to having these arguments and then look for some ways to deal with those challenges.
Let’s start with why these claims are hard:
One defeat does not a law of nature make
At high levels of play, the difference between the best player and the second best player might be that the best player wins 60% of the time. That’s why a lot of competitive leagues have a “best 3 out 5” system for their finals.
So, if Jaina defeats Khadgar once, how do we know that Khadgar didn’t just roll a 39 on a 1d100 with a DC of 40?
Power is asymmetrical, so power inequalities aren’t transitive.
Just because rock beats scissors, it doesn’t follow that rock will beat paper.
For example, suppose Thrall defeats Garrosh. Now suppose that Garrosh defeats Anduin. Does it follow that Thrall can defeat Anduin? Not necessarily. Shaman vs Warrior != Shaman vs Priest != Priest vs Warrior
Power is unstable
One interesting trend with chess masters’ ELOs over a career is that they tend to increase, then peak, then decrease. Chess gets more competitive and old players’ abilities start to decline.
WoW is a universe with a LOT of power inflation. Because of this, we can’t necessarily infer that because Azshara was obscenely powerful in her prime that she’s still at the same level thousands of years later, either in absolute or relative terms.
Context Matters
“Home field advantage” is a statistical reality in a lot of sports.
For example, say we’re talking Jaina vs Khadgar. OK, is this fight on a boat in the ocean or in Karazhan? Does Jaina have the Focusing Iris? Is Jaina currently in a peacecraft or VENGEANCE mindset?
Since the difference between THE BEST and THE SECOND BEST is often so tiny, these can make a big difference.
The devs can do whatever they want
Broxigar can mess with Sargeras if the devs feel like it. Not much to say or do here, but just something you have to acknowledge if you mess with this nonsense.
And now for some suggestions for getting around these:
This post is a lot longer than I’d intended, so I’m just going to throw out som bullet points and maybe expand on these later if it seems worth it.
- Create tiered hierarchies. Instead of arguing about whether Malfurion > Tyrande, maybe we can say that they’re too close to call – but they belong in a tier above Nathanos and below Sargeras.
- Speak in terms of probability. Speaking of Nathanos, maybe we could say that Tyrande has a 99% chance of defeating Nathanos based on their relative feats and combat history. This leaves open the possibility that Nathanos could win if he’s obscenely lucky.
- Use combat history to create those probabilities: For example, I think there’s a roughly 80% chance Arthas would defeat Bolvar. Why? Because if you look at their win/loss record, and then you plug those W/L records into a win probability calculator for sports, then you get about 80% for Arthas winning, about 20% for Bolvar.
- Assign weights to data based on estimated power level and recency: For example, I could make the above comparison more accurate if I gave more weight to wins vs world-level threats than to wins vs lower level threats. Similarly, more recent wins should count for more than wins in the distant past because of the tendency for power levels to change.
- Control for relevant variables: If you’re comparing Anduin vs Garrosh on an EK battlefield, for example, then as much as possible give weight to data re Warrior vs Priest while the priest has homefield advantage.
What thoughts have everyone else? What are some ways we could make power level comparisons more useful? Are there problems with power levels I left out?
Also: The real take-away from all this is that most of the time we have very little data on how an x vs y fight would go, so most fights are probably less predictable than we tend to think.