Might be really hard to actually calculate the minimum. There seems to be too many build specific factors. It seems much easier to just estimate it.
If you want a build and time independent formula for minimum, you just need to select worst case scenario.
Worst case scenario might be a speed build that doesn’t scale well. “Normally” speed builds wouldn’t need to be considered in maximum levels, but as people are running 2 minute 150s, they are using speed builds. Bad scaling speed build could easily have problems even with 17% increase per level. For example if someone kills rg with stacked dmg, it only scales well until stacks are lost.
I think global minimum scalability factor just cannot be calculated. Easier and more reliable just seems to be doing more or less educated guesses per build.
While this analysis can work somewhat, I think it misses an important factor of survivability especially regarding very fast clears.
When you can clear a rift in 5-6 minutes, you can afford to go full glass cannon, hope to dodge stuff, and 1/2 rotate elites. When you get to 13+ minute clears, the usual way to play is to make bigger packs (because you can’t kill elites fast enough anymore). These bigger packs lead to having to either avoid or tank more damage, which means you can’t do as much DPS as when you clear way faster.
Some builds can tank either way (say roland crusader, super tanky, doesn’t matter), but some other builds will struggle, like inna for example. Inna is quite squishy. When you don’t have enough dps to kill an elite by itself, you kinda tend to drag it to another pack. This other pack will attack you, which means you have to dodge and you’re taking even longer to kill the first elite, all while the first elite also attacks you, potentially breaking your squirts.
To go even further, some clears you’ll see are, idk, 12 minute clears that were done “chill”, which means you don’t play for absolute best time, you play for certainty of not messing up. Example of those would be, “idk, this elite will take me 3 rotations more to kill, I could skip it and hope to do the same amount of progress in 2 rotations but 3 rotations will give me enough time to kill the boss anyway”. Of course, in your analysis, you consider those 2 different lines of play to be “different rifts” (ie, if he had done the 2 rotation clear, he would have cleared faster and it would change the outcome), but because of that you can’t rely on actual clear times on the LB without video “proof” because for all you know the clear could have been made faster and the player just didn’t bother.
I was just about to mention toughness. The monsters get a lot harder. That’s one of the issues I’m facing now with pushing the crusader S24 leaderboards. It’s forcing me away from using better items that increase damage to boosting survivability.
That’s a good point about toughness. Obviously, dying a lot would change the outcome significantly. I should amend the OP to mention that. And I should change the phrasing to indicate that it’s not necessarily certain that one would get the clear X tiers higher based on such and such a time, just that you certainly have the damage to do so.
As for the point about “chill” play, I would say that doesn’t really matter. This method doesn’t estimate the maximum number of tiers higher you could go, it’s a minimum. So, when based on a “chill” rift, that would certainly factor into the minimum, too.
Going back to the example in the OP of the Monk clear in 4:37: just based on that sample, it’s certain he had the damage to go at least 7 tiers higher… unless incoming damage killed him. If he was playing that “chill”, then yeah, altering that would definitely improve his time… but 7 tiers would still be the minimum.
Edit: I put a note at the top of the OP mentioning the incoming damage issue, will incorporate that into the OP later.
Just 3 GR levels made me go from using the LoN ring set with bane of the trapped and hexing pants to unity, justice lantern, string of ears and the lod gem. talk about a difference.
I think you summed it up well and even took it a step further by wanting to know what happened in the rift with Conduit.
Like Abella, Kikaha and others said, the whole play dynamic changes according to the clear time. When your clear time is as low as 4:37, your gear and technique are likely to be very different than if your time is 14:37. You might skip on EW/CoE and go for F&R/Squirts… your conduit might only down 4 elites instead of 7. You might not take a stricken etc.
Ok, so we’ll use this rule for extrapolating how much higher the build can push and just assume no optimizations are done to gear and the play decisions remain the same.
Want to hear something funny before I jump into it…. ironically on my recent LoN Bomb clear, I accidentally pulled the Conduit with only 1 nearby elite, which went into a Power pylon that I grabbed anyway…. I was about 8 min ahead of the timer by next map, no real strategy there… it was the best outcome after like 700 keys of GR144 attempts and I think it might be the fastest/highest clear for LoN/LoD Bomb in my paragon bracket. In retrospect, I don’t think pulling all elites to the conduit would have produced a better time.
You’d really have to know:
Time to Guardian
Guardian Kill Time
Conduit Time still has player damage. In fast clears the player damage often exceeds the conduit output. You have to make those 60 seconds actually Conduit + Player damage, as the GR increases Conduit Time will yields less progress.
I will try to write more once past this holiday, been getting the place and food ready for family since yesterday, if you celebrate too have a Happy Thanksgiving. Now to extrapolate how many kCal I’m going to eat…
Yeah, in some cases the best strategy on average isn’t optimal under a specific set of circumstances. Sounds like you stumbled into some good luck!
That’s a good and interesting point. I initially started thinking about this endeavor in terms of pushing… but the top “pushes” are also quite fast (150 in as little as 4:37).
If your damage per minute (just that you inflict, not from a Conduit) were always the same through a whole rift, then in a 15:00 clear, you’d only be dealing about 6.7% of your player-inflicted damage during the minute you have a Conduit. But in a 5:00 clear, it would be 3x that much, 20%. Of course your damage isn’t perfectly even like that, but you get the idea.
This would be a pretty rough estimation, but you could change the “Conduit” time scaling to be based on the overall time, with the time scaling higher, the lower the “initial” time (i.e. the time of the original clear we’re looking at). I’ll give that some thought and see what I can come up with. If you have any suggestions I’d love to hear them.
Fair enough, if your goal is only to estimate a lower bound on grift based on clear time, it does work (± the toughness issue ofc)
I also want to add that you’re taking damage as a linear value throughout the rift. Most builds play around CoE, some do a lot (since you’re a barb main, you should know about the wonderful raekor hota gameplay where you charge until you have many charge stacks then unleash everything on fire coe). Sometimes, increasing a gr by even a single level can result in much higher time increases because instead of idk, 2 rotating a pack it takes 3 rotations. Boom, went from 32 seconds to 48 seconds.
It’s particularly bad for very fast clears with squirts, because if you one rotate a pack, you can usually keep your squirts up which doubles your damage. But if you can’t, you lose squirts and now you do half the damage. Which means in certain damage brackets, you can go from 1 rotation (with squirts) to 3 (without squirts) with only like 1 or 2 grift level increase.
It’s particularly obvious when you check actual speed clears (from gr129 to 131 on my monk, it feels like I need more than twice the time), but even for some push it can have effects when you rely on snowball effects like archon stacks.
I think that in the case of “just CoE” or “Raekor CoE” where your damage comes in discrete (but generally consistent) waves, you would end up seeing a sort of “syncopation” in how the time increases. Like in your example, where +1 GR adds 50% extra time (32 → 48 seconds), the next +1 GR would likely add 0% extra time. I think the extra time would ultimately (more or less) converge with the extra time added in a build that did the same overall damage, but continuously, rather than in phases. For instance, in a series of 4 “+1” GR increases, the “cycles” build might see +50% (2–>3 cycles), +0%, +0%, +33% (3–>4 cycles), or a total 1.99x the starting time, and the “consistent” build might see +17%, +17%, +17%, +17%, or a total 1.87x the starting time.
If you had video of a clear, then you could go through and check each separate engagement to see where, in that original case, they might have had a close call on killing a group of enemies, in which case you’d know that at +1 GR, they’d need an extra cycle to finish those enemies off, and you’d add in the correct amount of additional time.
If you don’t have video, then it’s tricky… but I still think you can make reasonable estimations. You know that the most the time can increase (just from this effect, of course) from a +1 GR increase is +100%, which is a case in which everything previously took one cycle, but now takes two. The least the time can increase is zero: you still kill everything in the same number of cycles.
And, the more cycles everything took to begin with, the lower the volatility of the increase, and the higher the likelihood of the increase. If it took 1 cycle to kill this mob before, and at +1 GR it takes either 1 or 2- i.e. an increase of either +0% or +100%- that’s quite volatile. But it’s also more likely than not that the time won’t go up at all.
For instance, let’s say you do 100 damage per cycle, and we know you killed this group of mobs in 1 cycle in the base rift. That means the mobs have somewhere between 1 and 100 life. If their actual life is anywhere between 1 and 85, you’ll still kill them in one cycle, even with 17% additional life (85 * 1.17 = 99.45). So the actual likelihood of that increase is pretty low, around 15%.
But if it took 5 cycles to kill the mobs before (500 damage), then at +1 GR it will still take at most just one additional cycle to kill them (500 * 1.17 = 585), and at the least, zero extra cycles, an increase of either +0% or +20%- much less volatile. But the odds of the increase are much higher, since if those mobs had anywhere between 428 and 500 life, it will take that extra cycle to kill them (428 * 1.17 = 500.76). So, there’s a 73% chance you’ll see an increase.
Since the size of the increase goes down as the chance of the increase goes up, I think what you’re really seeing is an increase in the uncertainty of smaller +GR increases, with much less uncertainty in larger +GR increases. And the level of uncertainty is going to be higher in cases where the number of total cycles is lower.
In that 4:37 Monk clear, for instance, there were only 13.85 CoE cycles (maximum, in a succeeded rift, is 45 for a 5-element class, 56.25 for 4-element, 75 for 3-element). But, having a really low number of cycles also means you also have a lot of extra time. So, it may be that to go from GR 150–>151, there’s a small but significant chance that it would take him a large amount of extra time: +50% or more. But even if it took him twice as much time, that would still be an easy clear. And to go +7 GRs would still be quite likely, since those “jumps” in the time would even out.
These effects would matter a lot, as you point out, in speeds, where the point is not just “to clear the rift”, but to do so both fast, and consistently. After all, nobody cares if you cleared GR 120 in 58 seconds one time, but it usually takes you 6 minutes to do it. There’s no leaderboard for speed clears (below 150, anyway). But for pushing, failing miserably 10 times and then breaking the world record is not just a success, but an outstanding success.
Throwing Squirt’s in there definitely adds another wrinkle, (whether or not it’s on top of damage cycles like CoE), since it links your outgoing damage to your incoming damage. For builds that just take Squirt’s in the hope that they’ll find a Shield pylon, I think you could pretty much ignore the Squirt effect, since with a Shield, you will always take 0 damage and always have the damage buff while the pylon is active, and basically never have the buff otherwise. But for builds that use a shield-generating power, and try to keep Squirt’s up 100% of the time, you could definitely see a big drop in outgoing damage, as the incoming damage gets higher.
Anyway, thanks for the comments and the Thanksgiving wishes (even though you don’t celebrate… you are French, yes?)
Hmm, let’s see. So, I guess the endpoints of this relationship are:
The Player does basically all the damage (20000 paragon player with godly gear, full 150 gems + augs, playing Inna Monk, with a Power Pylon also active, in a GR 1). In this case, as you increase the GR level, the “Conduit” time will (essentially) increase at the same rate as “Regular” time, +1.17% per level.
The Conduit does basically all the damage (0 paragon player with all white items in a GR 150). In this case, as you increase the GR level, the “Conduit” time will (essentially) not increase.
And, as you move from the first of these extreme examples, towards the second, the rate of increase will slow down, eventually reaching ~0% increase.
With video, I think you could get a very solid estimate of what % of the total damage the Conduit was doing, and it wouldn’t be too much of a problem to work out its contribution. Without video, perhaps you could just make an estimation based on the total time - “Stricken” time (so, all the time other than fighting the boss)? That itself could be hard to estimate, but let’s suppose for a moment we had a pretty good rough estimate of the boss timing.
So, let’s say a 480 second (8:00) clear, with a 90 second boss fight. That leaves 390 seconds of “non-boss” time. You do 100% of your player-inflicted, non-boss damage in those 390 seconds, or .256% average, per second. In 60 seconds of Conduit time, you’d do 15.38% of your total damage. So in this instance, maybe I’d scale up the Conduit time by 2.61% (15.38% of 17%)?
That seems like it’s maybe too low. I think part of what’s missing here, and from the analysis as a whole, is time spent just moving around. After all, when playing, say, a GR 1 with any regular build, your damage might as well be infinity, but you don’t clear the rift in a time that might as well be zero (0.01 seconds or something)- it still takes you 60 seconds, or however long you need to spend moving through the rift, hitting the mobs, picking up globes, etc.
And by the same token, when “upgrading” a GR, time spent moving around also wouldn’t generally increase at all. Let’s say that in a clear, the player played level 1, which took 3 minutes, skipped level 2, which took 30 seconds, and played level 3 + killed the boss, which took 4 minutes (7:30 total). If playing this rift at +4 tiers, it wouldn’t take 30 seconds * 1.17^4 = 56 seconds to skip that second level. It should still only take about 30 seconds.
This is why I included “moving around” as one of the major GR activities, in the OP. I meant to consider this further, but kind of got sidetracked.
Once again, if you have video, this would be pretty easy to figure out- just see how much time is spent just moving, rather than fighting. But that’s tough to estimate, without video… unless somebody has some ideas about how to do so. Tinne, I’ve seen you lurking about! Any thoughts on this particular matter, or on any other part of this analysis?
Here’s the current list of “flies in the ointment”:
Incoming damage, which at a higher tier might kill you, increasing the time.
Intermittent outgoing damage, i.e. CoE, which might significantly impact the time (increase by either more OR less than the “expected” 17%), especially in small-number tier increases.
Inconsistent outgoing damage, linked to either incoming damage (Squirt’s) or outgoing damage/kills (Archon stacks, Rampage stacks, etc), either of which can increase the time.
Scaling of Conduit time, depending on the amount of damage the player is doing.
Estimating “move time”, which should scale up very little, or not at all, as the tier increases.
All of these difficulties can be mitigated A LOT if you have video. But, can we come up with reasonable estimations for some or all of these, that apply either “broadly” (i.e. to all builds) or “specifically” (i.e. the estimations are tuned to the particular characteristics of different builds)?
Good talk! I like the excerpt from that novel, maybe I’ll have to check it out.
Point taken, you are probably right. The estimation will be close enough, overall that it shouldn’t usually be too prejudiced for one build, or against another.
I guess there’s a “reasonable” set of adjustments-what you could do without going too deep into the weeds. Based on the issues I highlighed before, those might be:
Add possible +/- of maybe 30% to the answer, i.e. if we were saying “+7 tiers” before, now we might say, 5 tiers, +/- 2.
I don’t think this will actually have very much impact, in terms of yes/no completion. For smaller times, it might move the time up a good bit, but you’ll still easily clear. For larger times, it’s much less likely to change the outcome. So I think maybe I just let this one go.
There are a few really major items or abilities that cause this issue, most of all Squirt’s. The easy fix is probably just to apply an additional +/- 30% to builds that try to maintain Squirt’s uptime. So “+7 tiers” would now be “3 tiers, +/- 4”
Just discount the Conduit effect, based on the share of Conduit time, i.e. in a 1 minute rift, Conduit time scales like Regular time (+17%/level). And in a 15 minute rift, it doesn’t scale up at all.
Just count an average % of time as “move time” (which doesn’t scale up) for every rift, like maybe 20% of the total (3 minutes out of every 15).
The relative difference between 1 and 2 rotations is bigger than the relative difference between 3 and 4 rotations. CoE rotations will have a bigger impact the faster the rift / fewer rotations on average you are taking on each pull.
I notice that a lot when doing farming runs. That spot where a troublesome elite or the RG jumps from taking one rotation to >1 rotation is pretty significant. As you push higher, that impact shrinks.
It’s basically just a discrete/“stairstep” function with jumps instead of a continuous function. The higher you go up the “staircase” the closer it will approximate the actual difference were it a continuous function. For small x, (x+1)/x can be much larger than 17%. For larger x, (x+1)/x can be a pretty close to 17%. For even large x, you start looking at (x+2)/x instead (i.e. it takes you two extra rotations instead of 1). Point being, there’s a range where the discrete increase in CoE rotations doesn’t deviate that much from the 17% difference in enemy HP, so it’ll be reasonably approximate to the actual increase.
I think something like that is reasonable. I do think it will vary a bit by build (some builds have an easier time continuing to attack while dragging a pull, for example). Might have to go through a few videos of different builds and see if there’s much variation in move time. “Elite killer” builds might have a higher proportion of move time for example.
Cool. What do you think about plus/minuses for possibility of dying (every build) or shedding lots of damage (Squirt’s)? Also, Conduit “downgrade” based on rift length?
conduit vs rift length is hard because on one hand it’s a lesser percentage of the rift spent with conduit but it’s a more important part of the clear when it kills 7 elites that you would never kill without cond
on speed clears (-5min) you basically outdamage the conduit so it’s not as key
Right. Part of this analysis though is that “nothing changes”. So, y’know, if somebody did GR X before, either they did it in that time with a Conduit, or in that time, without it. We aren’t adding or subtracting the Pylon, just tweaking the value up or down a bit (if present) or ignoring it (if not). But yeah, in a 2 minute speed, the value of a Conduit is basically 0.
Ultimately, I think this method is a lot more useful for clears that are on a leaderboard, rather than just random speeds we knock off every day. For instance, while it may have quite a low time, that Monk 150 in 4:37 is sort of fundamentally different from some particular “speed” run of 130, say, in the same amount of time. For one thing, it’s the world #1 clear, meaning a lot of positive factors have to have converged to make such a thing possible- good map, mobs, pylons, boss, etc. If that wasn’t the case (if it was easy), that player, or some other player, would have done better.
We had a coefficient in our No Paragon Rankings dependent on each class influencing the times. You can apply such concept for each build that deviates from the general case.
Whatever formulas come out of this should be play tested. It is easy to run rifts 125 through 140 a few times over with different builds, recording the clear times and quick notes of what happened.
If a few people did this we’d have enough data to plot some lines and combine the findings with the theory. It’d also be amusing to see the intermittent losses on the upper end clears.
How do you want the clears documented? A table, introduced by citing the build, with these columns:
GR Level
Clear Time
Pylons Triggered (Co,Po,Ch,Sp,Sh)
Number of Floors
Notes (3 deaths, bad maps, etc)
Do you care if it season vs non-season, shouldn’t matter right? Other requirements: no gear changing throughout, follower must have Flavor of Time?
You know what’s funny? Some Blizzard dev mentioned years ago they “cut paragon” themselves when they are testing numbers internally… So, instead of fixing Challenge Rifts with tier progression and builds determined with voting from players so that they have the perfect data, they had made a formula on their own…
