Atonement is a real moving target right now, having changed substantially at least four times in the past couple of weeks. As a result I want to go a little easy on the state of the current numbers, which aren't all that bad anyway, and delve into the deeper problems with the entire mechanic.
Just like I always do when I want to know how good something is, I made myself a spreadsheet with a simulator for the casting rotation and ran it using a variety of different stats. One purpose was the figure out the relative strength of Smite vs. Heal healing. Another purpose was the see how stats compared for Smite vs. Heal healing, and whether some stats were significantly stronger for one than the other. I am comparing a stand and cast single target healing throughput. The difference between the two builds for AoE healing definitely gives a greater edge to smite, since other than the smart healing Atonement on the melee the two builds do the same thing.
I'm going to consider throughput and mana considerations completely separately. How and whether these can be meaningfully combined into a composite weighting for stats is something I think a lot about, but for these purposes, it's not terribly illuminating to attempt to combine them anyway.
I used the following two talent builds: Heal and Smite
Now you may think I'm crazy for giving up Soul Warding in favour of say Borrowed Time or Train of Thought, but I honestly don't care that much for Soul Warding and think it shouldn't be a talent at all, a topic I'll write about later. Obviously this could be changed if fights actually warrant Soul Warding.
So in a raw comparison, the Smite build does about 11% more throughput than the Heal build. Also, the Smite build has easily enough mana to carry on through a five minute fight chain casting; about 56k more mana than it needs, actually, meaning it can easily substitute some smites or AoE heals for Flash Heals if warranted. Heal on the other hand, has basically just the right amount of mana to scrape by on the maximum mana efficiency rotation, and any extra healing it does with Flash Heals or AoEs will have to come out of its Heal budget. Of course on a real encounter you would have moments where you were moving or waiting for the boss to change forms, so you would have some room for those things.
11% more throughput and an extra 11k mana per minute to spend seems to make the Smite build the clear winner. That's even ignoring the 4.3k dps the smite priest provides, which is presumably a noticeable fraction of an actual dpser (though less than half of one). Of course these things can simply be adjusted. The healing value of Atonement could be lowered and the cooldown of Archangel could be increased and it would fix the problem, right?
Turns out, not really. I don't think there are any values that you could set for the talents and spells involved that would make this build make sense. The problem is scaling.
Heal scales at about 3/7 of spell power for a 2.5 second cast while Smite scales at 5/7 of spell power for a 2 second cast. Smite, therefore, scales at about double heal. Even when you credit heal with Empowered Healing and Grace, Smite still scales about 60% better with spell power. With a full stack of Evangelism, which you have more often than you don't, smite is about double Heal again.
Why can't this be fixed by lowering Atonement to healing only 50% of Smite's damage rather than 100%? It's because the base amounts are so radically different. The base amount for Heal is about 4 times that of Smite. If you reduce Atonement by half then the Smite build falls way behind the Heal build for healing. About 89% of Smite's heal comes from the spell power. For Heal, just over half the heal comes from spell power. If you want to reduce the scaling by half for Smite then you have to reduce the total throughput by 45%.
The mana problem is probably considerably more troublesome. Archangel restores 15% of your maximum mana on a 30 second cooldown. That's a talented shadowfiend times three. Could we simply reduce the amount of mana that archangel gives back? Again, virtually no matter what you set the number at the scaling problems are enormous. Intellect is about 35% better at providing mana for the Smite build compared to the Heal build. To equalize the scaling for mana from Intellect we would need to drop Archangel to 4% mana return, but then the Smite build would be unplayable to begin with. Instead of a 56k surplus it would run a 58k deficit over 5 minutes. There really is no sweet spot. Set Archangel at 8% and the Smite build doesn't have enough mana to play now, but will surpass the Heal build in mana by the final tier.
As things stand I think you can make arguments in favour of either build. Atonement is inevitably going to have some range issues, though there may end up being fights where using the proper ranges can ensure it heals only the tank when you want it too, which would be a plus. Smart targeting better than non-smart targeting for all heals except for heals you want to consistently spam on the tank which is is probably a little worse for. Strength of Soul is actually a very good talent that you might miss in some situations, and while both builds can move a couple points to pick up Inspiration, the Smite build would have a much worse uptime on it since it is not triggered by Atonement. I think Smite is a little overpowered, but not drastically.
By the final tier of raiding content Smite will be drastically overpowered compared to Heal, and there won't really be a choice. But fixing this would mean making Smite unplayable until upper tiers.
Just for curiosity's sake, the stat scaling values have been pretty consistent with all my other thought experiments. Spirit is a little under half an int for mana for Smite or nearly two-thirds for Heal (spirit gives the same amount, but is weaker relative to Int when Int is better). For throughput compared to Intellect critical strike rating and mastery rating are both terrible. coming in at 35% and 24% respectively for Smite and 42% each for Heal. Haste is worth more than half an Int of throughput for both builds but ends up costing a decent fraction of a Int worth of mana. Haste actually came out very strong for the Heal build, probably because more Haste means more shields for that build. But since mana is more of a consideration for that build and the other stats are also relatively stronger I think it's a bit of a toss-up. For the Heal build I would tend to ignore which stats my gear has since they are all the same in terms of raw power, though I might tend towards Spirit if I find I'm running low on mana, obviously. For the Smite build I'd focus on Haste since nothing else is even worth looking at and it has a surplus of mana anyway.
I really love the fact that they are making Smite a build Unfortunately the crazy scaling discrepancies between healing and dps spells are going to make it a really wonky one. I also don't doubt that other healers are going to start complaining that priests get to do dps while healing and they don't. Personally I'm going to enjoy the ride while it lasts, though I have a suspicion that it might not even last until release. On the other hand, getting rid of evangelism/archangel/atonement would mean needing to put other points in the discipline tree, and I really don't think they are up to thinking more points up. I guess they could just lower the effectiveness of Power Word: Shield by 20% and put in a talent that increases its effectiveness by 25%. That's what half the discipline tree is anyway.