Howdy everyone and Happy New Year!
First off, I'm bumping the update schedule slightly -- instead of posting in the middle of the month, we'll post at the end. New Year seems like a good time to make that change. Also, my entire family was sick two or three times so I needed the extra two weeks :)
I’m not in the testing set, so this is based off my experience in DQ1. While a good approximate rule, I don’t think it is quite right to have dps per cost higher for high Tiers. Considering a dps efficiency, I’d usually suggest (very) slightly undershooting as safer. Basically, I think change “at least” into “almost”.
“Pushing” (playing at near the minimum power possible for the level) really hard does force you back to DPS per PSI in DQ1 pretty strongly. But, especially in NG+ where there is enemy heal and stuns and AOE, there is a sharp priority on damage density as well as total DPS. Being able to dump a ton of power into one spot really matters some times. (It can even fake lane coverage at times if everyone is functionally gatekeeping.)
Another way it pops up in DQ1 is weapons for the mooks of a class. If you’re doing say the linear run trying to take on the first Wrenna level in NG+ at only Azra 19, you might have 4-5 archers. But since your berserkers are higher DPS, most of your psi is going there, and you just don’t care about which bow your 4th archer has. (In fact, I’ve been poking that again trying 4 archers and 4 berserkers because I’m hoping knights with knockback + AOE Wrenna might crack sheep. So I don’t even have a 5th one at all.) Since you are psi limited on who you summon, you might well “overpay” for a weapon trying to have the best one you can on your first few units down.
Not all of these factors apply to the DQ2 the same way, but maybe new ones do. A sufficiently advanced measure of “effective DPS” could even wrap them all in. But I’d guess that, slightly overshooting the lesson of Niru, you might be discounting the value of the big unit hammering the critical point - if the high Cost/Tier are actively more efficient, you might get the low Tier people being ignored unless it is a pretty straight (and maybe dull) spread lanes check map. While if they are close but slightly less effective on raw power, you get questions about concentration vs. efficiency.
Most of the same considerations exist around boost vs. more units too. One of the fun and very subtle bits to DQ1 is how much those tiny margins can add up. And often those efficiency/strategy decisions hinge on exact map design, which is almost the holy grail for this sort of game IMO - all your systems are interacting with each other at the point of play.
Especially if any of the other flexibility (coverage, cc, density etc.) gets mis-valued when balancing with DPS as the main balancer for Power Tiers, you could easily get “no brainer” strategy choices. I think a lot of the detailed balancing happens in a polishing type step, but having the high tier “overdriven” on dps feels like it might risk needing more than just a bit of a detune unless DPS was really all they did.
Finally, a special note - if you have character slot limits, and the ability to swap them at points, and could change the tier composition of your party that way… and finally, have levels that produce a whole lot of summon resource over time, you’ve made a problem. Now one of your limiters is character slots, and a Tier 3 is just flatly better always for that, basically by definition. In theory even the perception of that limit could nudge a lot of players. I don’t know if all those parts are even in play, but if enough of them are that could be a fairly fundamental concern.
Sorry it got a bit long as a response - just a few thoughts that hopefully help to read and then put at the back of your mind. And again, as a rule of thumb keeping similar DPS/Cost seems exactly right: calculating the animation frame lockouts to get actual DPS values is what prompted me to read the XML files of DQ1 - because it really is that important a metric. All my commentary is opinion on the fine details of that with the understanding the core observation sounds quite right to me.