I also could have explored the more pure-DPS appropriate situation of a raiding tier split between all three specs, MS/OS/OS, where you have one mainspec and two offspecs of different trailing ranks, but I didn’t think of it at the time, and I will may update later when I want to do that amount of work.
— Me, Multispeccing Artifacts, 12 July 2016
I’m glad I decided to double-check myself.
Thinking about tri-spec practically
In my previous post, I concluded that while dual-speccing is definitely viable compared to investing in a single main artifact, tri-speccing and quad-speccing may not be as viable, depending on player preference of grinding Artifact Power (AP) and mechanical game offerings of such amounts of AP. I wasn’t wrong per se, but in that post I was also describing a singular and very general case of tri-speccing. To be specific, I was describing the “worst case” scenario of tri-speccing, where you effectively kept all three specs equal in artifact trait ranks — in other words, where every spec was your mainspec (MS/MS/MS).
Practical tri-speccing doesn’t often happen that way. The offspec (OS) is usually inferior to the mainspec, and the third spec is inferior still to the offspec. Furthermore, the third spec isn’t always a constant presence; you might be happy with two specs for some time, and then a change or pro/regression in activity causes you to add that third spec.
Again, as a pure DPS myself (Warlock), the viability of tri-spec in Legion is a significant concept that I personally want to explore. Moreover, tri-spec is probably the broadest application of Legion’s new freedom of speccing feature. (Quad-speccing is Druid-only and dual-speccing isn’t new for us, despite being the only option for Demon Hunters.) It also applies to other cases outside a pure DPS — e.g.:
- the third tank who normally maintains a double DPS spec setup (Warriors/Death Knights)
- the swing healer (Elemental/Restoration Shaman) who PvPs in their third spec (Enhancement)
- a mainspec PvWhatever endgamer (Balance Druid) who does the other PvContent endgame in their second spec (Restoration) and does retro/solo content for transmog in yet a third spec (Guardian)
I had concluded that the viability of any multispeccing at all decreases significantly once you finish the main artifact tree of any given spec and hit the hidden traits; thus, I’m going to focus my results and conclusions here on the main artifact tree (34 traits). The data in my spreadsheet goes up to the full 54 traits (34 + 20 hidden) if possible, so you can check that out if you want more.
The work for this post builds off the work I did in my previous post, Multispeccing Artifacts. You can go read it if you like, but hopefully I’ve written this one such that you won’t have to (but maybe you’ll want to!). This post in particular explores the details of Artifact Power expenditure when tri-speccing, while the previous post deals with the impacts of Artifact Knowledge as well as more general cases for dual-speccing, tri-speccing, and quad-speccing compared together.
I have a new spreadsheet particularly for tri-spec analysis. It cribs a little bit of data from my previous spreadsheet, which was itself based off & adapted from spreadsheets from Kerriodos of <Temerity> US-Windrunner and from Kib#2498 on Discord.
I found a small typo in Kerri’s sheet, where rank 46 was said to cost 5115000 AP, but was actually 3115000 AP. I corrected this in my latest spreadsheet (for this post), but it doesn’t have much effect on my conclusions (the effect it has is that equivalent ranks past trait 45 may be adjusted down by 1 rank).
Thanks to Ranor for proofreading/reviewing this post for words problems, and generally being a sounding board for whether my math & explanations were clear enough for a layperson.
Thanks to Binkenstein for mentioning difference graphs to me so I would quit trying to show off a bunch of overlapping scatter point-lines on a zoomed graph.
Thanks to various people who told me about where they were discussing my previous post. Like Arielle, who mentioned it on TankCast, so I thought of third tanks with third specs. Or like Dravvie, who said there was an awesome discussion-argument on her guild forums over who could really put in the effort to tri-spec, so I thought about, really, what would the effort look like if I did partial tri-speccing? Could I beat my original conclusion?
The first step to analysing tri-speccing in detail is to define it in detail. While I don’t intend to analyse all 7140 combinations with repetition of 34 traits in three different artifact specs ( (34+3-1)!/(3!*(34-1)!) = 7140), I can look at some reasonable and representative cases of tri-speccing.
I’m defining all of my setups as tri-spec setups, as they can help us compare impacts of AP spending from both the OS and the 3rd. My terminology starts to solidify now, in order to keep the various sets clear and distinct from each other.
- MS — MainSpec, usually defined as the spec with the most/leading number of traits purchased in its artifact
- OS — OffSpec, usually defined as the spec with the second most number of traits purchased in its artifact
- 3rd — third spec, usually defined as the spec with the third most/least number of traits purchased in its artifact
- MS/OS/3rd — The slashes when talking spec setups separate spec terminology. I usually refer to them in this order, so the “3rd” spec is the spec after the second slash.
- MS-3, MS-6, etc. — I represent constantly trailing specs in terms of how far behind the mainspec they are, so they can change as the mainspec is progressed
- 0, 13, 16, 26, etc. — I use plain numbers to represent a static number of traits in a spec, usually the third spec. The static nonzero numbers are pulled from my previous post where I looked at the general rank point where AP is cheap (13) or where Warlock specs gain approximately 2 & 3 gold traits (16 & 26, respectively).
Click the picture for a bigger picture.
I started with my original graph, without the quad-spec and with the axes trimmed. Both the x- and y-axis only go to 35, because I’m focusing on the main artifact tree progression. I thought about zooming the graph by moving the axis lines more, but even then, the points would still be so close together. I was going to add four more cases of tri-speccing, which would overlap each other a lot. I needed a better graph.
Enter the difference graph. We’re going to take the points of each multispec setup and subtract the points of the 1x Artifact line, and then graph the results. The general shapes of the lines remain the same; think of it like rotating the graph clockwise about 45 degrees. That 1x Artifact y=x line becomes y=0, and the other multispeccing setups become lines with areas on top of it. I left in the full trait progression on this graph, just to emphasize the jump in AP for the multispeccing spent after trait 30 in your mainspec.
Click the picture for a bigger picture. Lines:
- Green (top) — 3x Equals, or MS/MS/MS
- Blue (middle) — 2x Equals, or MS/MS/0
- Red (bottom) — Dual-spec trailing, or MS/(MS-3)/0
- Black (y=0) — 1x Artifact, or MS/0/0
Now that rank 13-14 breakpoint is really obvious, since all the graphs suddenly dip down to a 1-2 rank difference at trait 13. Also, it’s obvious that each setup starts to take off significantly after 32 or so, which is when you’re about to finish a mainspec and start investing in hidden traits.
The impact of partial offspeccing
We’re adding that third spec, so it’ll be above a two-spec, but it’s not a full tri-spec, so it’ll be less than that tri-spec. Thus, a partial tri-spec would graph between the dual-spec equals and the tri-spec equals. Right?
Wrong! In fact, the only place where the trailing tri-spec crosses into the area between the equals dual-spec and equals tri-spec is way the hell up at trait 40+. Otherwise, it’s actually less than the equals dual-spec for most of the time.
Click the picture for a bigger picture.
- Lighter colors = equals multispeccing
- Darker = partial multispeccing
- Yellow = Dual-spec
- Blue = Tri-spec
My second hypothesis is correct — that the partial tri-spec lies between the full dual-spec and the partial dual-spec — and it should have been obvious. I built the partial tri-spec setup based off the partial dual-spec setup. The first spec is MS, the second spec is MS-3, and the third spec is 0 (partial dual-spec) or MS-6 (partial tri-spec). The lowest the partial tri-spec could be is MS/MS-3/0, which is my definition of a partial dual-spec. The equals dual-spec being an upper bound should also have been obvious due to the way the AP ranks significantly increase in value as you up. Rank 32 by itself is bigger than all of 1-26, for example.
Thus, partial offspeccing can give you a substantial artifact progression increase in a second or third spec at the cost of an increase of 1-3 traits for a single mainspec.
Stopping places versus dumping
Is it better to partially tri-spec up to a point and stop? Possibly — again, this depends on your preference of degree of competence in whatever number of specs you are spending in.
Click the picture for a bigger picture.
- Blue = partial trailing in both dual-spec & tri-spec varieties
- Yellow = Third spec stops at a given point, becoming trailing dual-spec thereafter.
“3rd Easy” (trait 13 in third spec) and “3rd 2xGold” (trait 16 in third spec) have about the same effect, while trailing tri-spec and “3rd 3xGold” (trait 26 in third spec) are close together. I’d likely draw the conclusion that if you were going to try to have three well-progressed specs, you’d probably want to partial tri-spec the whole way up to 33 in your main spec (so that’s 27 in your third spec). If you want to stay more equivalent to a partial dual-spec later in your mainspec’s artifact progression, you’ll have to be OK with a half-filled third spec.
If you look at dumping 1-X AP into a third spec all at once, it might be easier to look at how much of the next rank that amount of AP would cost you.
|Rank where a 1-X AP dump
is approximately worth a percentage of the rank’s AP cost
|Percent of the
Essentially, it’s really easy to dump 1-13 into a third or fourth spec, even when you’re in the middle of your mainspec progression. You could reach for a second gold trait with a 1-16 or 1-20 dump while in your mid-to-late-20s of a mainspec. However, doing a 1-X dump over 25 or so gets harder to visually convince yourself that you’re not wasting your mainspec’s next rank of AP in this other spec, until you’re already in ranks where you ought to be devoted to a mainspec because you’re in the hidden traits.
It’s obvious that it’s not recommended to dump 1-X AP into a second, third, of fourth spec until you’ve passed at least X ranks on your mainspec.
If you are going along dual-speccing just fine, but for some reason need that third spec fairly well progressed (e.g., Main Tank quits game & third tank now gets sudden promotion), then it’ll be worth a significant amount of AP to get that trade-off.
Partial tri-speccing is viable when compared to fully equal dual-speccing, and only slightly more expensive (approximately +1 equivalent single artifact rank) than partial dual-speccing.
When it comes to filling the third spec, players have a few options. Keeping a constant third spec running, even if inferior to the other two specs, is recommended if the player wishes to have three well-progressed artifacts. Dumping many ranks of Artifact Power is not recommended for keeping a secondary, tertiary, or quaternary spec well-progressed, unless in a case of sudden need.
Dumping Artifact Power into a third spec up to approximately half tree progression (17) is doable while midway progressing in a mainspec artifact without a strongly negative, wasteful feeling, whether through numerical analysis or through the visual experience bar in-game. Dumping large amounts of Artifact Power gets less convenient and requires player- and situation-specific decisions on the trade-off of one more rank in the mainspec versus what a well-progressed second, third, or fourth spec would bring to the player and to the content being attempted.