Reply To: Would the devs or members of this community be interested in a comparative…

#4185
guidon101
Participant

Wow, you are going somewhere with this! :)

Wanted to offer some constructive considerations. Maybe you already had all this in mind, and you were just starting with the RawNumbers approach as a base… but in case it adds any value:

As with all statistical approaches, I worry that quantifying things empirically may get convoluted and misleading due to many factors, if the other factors are not taken into account properly: for example, (1) contextual value and (2) temporal value and (3) purpose/playstyle

It seems right now you are going for an “economic raw BB-stat optimization curve”; what’s most cost-efficient with highest raw stats, is that right?

Also, when you create those variable labels, like “RawBonus” I was wondering how you defined them? Is it the raw sum of the numbers from background+trait, so you are treating +1 mdef = +1 rdef = 1% xp bonus = 1 raw point ?

In that case, I would consider coefficients/weights associated with each stat based on at least:
(1) contextual value: In my opinion, not all stats are made equal depending on context. +1 mdef is worth more to someone prone to be in melee vs. not; +1 mattack is worth more to a melee fighter than +1 ranged. In this sense, it becomes not cost-effective to pay for a high-RawBonus +20mdef/+20mattack Swordmaster, only to make it an archer (not that anyone would, because we all weigh all these other contextual considerations when we decide)

(2) temporal value (another kind of context): for example, with the +%10 xp bonus trait, it becomes completely worthless when you reach level cap; it’s value is closely associated with the early game development; so in terms of the RawBonus points economic model, you are paying $ for 10 RawPoints, worth nothing at lvl 11

(3) purpose/play-style: if you are going for 9 tanks/3 rangers, you may optimize on specific stats; 9 melee-oriented stats, 3 ranged-oriented; I would never just look at the Raw numbers for this reason; some stats are more valuable because of my play-style/purpose. In one game, I opted out of using Rally Troops and decided to manage fatigue individually; in that game, I never saw cause to put points into Resolve since I had no “rally captains”. Or, for example, if I want a “Perfect Focus” character, I would value high-Fatigue stats more

Basically, depending on context, a player would probably place a value coefficient for each stat (whether we do it consciously through hard numbers or subconsciously through intuition). I imagine that is hard to model with high fidelity, but that may be a fun and informative challenge and maybe that is your goal/point.

Taking the “experience based”/experiential “normative economics” approach is also very useful, and maybe more preferable/easier to understand/relate to for most people.

With the level of detail you are approaching, I very much look forward to where you are going and the final result, so take my comments with a grain of salt.