GR Tier vs GR Time: An Analysis

General Discussion
42

I think that’s a good idea, though before we do that, I think we ought to run a little thought experiment:

What’s the worst we can screw up this process, with/without video?

For instance, let’s say I take my Frenzy Barb into a 120, which I can usually beat within 7 minutes. For the first 2 minutes, I just stand there, not doing anything. After that, I start to actually play the rift. It’s a great Festering, with easy elites and swarms. About 4 minutes in, I find a Conduit, but as soon as I trigger it, I just run away from the mobs again, and it doesn’t kill anything. When the Pylon runs out, I go back to killing the mobs, spawning the boss at around 9 minutes. I fight the boss till they are at about half health (~20 seconds, tops), then run away again, and just wait till the 13 minute mark to finish the kill.

So, how would we evaluate this dumpster fire of a run, either with or without video, and what conclusions would we draw in either case?

That’s a serious question. But, what I’m kind of suggesting is that there are cases where the input doesn’t really work with this analysis, and other cases where the output either doesn’t work, or is kind of irrelevant.

I think the analysis is really only worth doing for clears that are on the leaderboard. And the higher they are on any particular board, the better the analysis will work. Anything that is close to “rank 1” of any sort (overall, for class, for build, below X paragon, etc) will kind of inherently filter out a lot of the dumb garbage I included in my silly example. So, looking just at “leaderboard examples” kind of cleans up the input.

And, in terms of output, I think this analysis won’t really provide any worthwhile info about speed runs. I mean, the way we value speeds is based on consistency, and all this analysis can show us is that, in one particular case, maybe that player could go 3 tiers higher, while still falling below the time limit of whatever we consider “speeds”. And, that’s basically useless info. For pushing, though, it’s generally all about the single great clear. A set of 5 results of 20:00, 20:00, 20:00, 20:00, 4:37 is not just better than results of 13:00, 13:00, 13:00, 13:00, 13:00, even though the total time is higher in the first case, it is enormously better.

I know that, in a sense, there is no difference between pushing and speeds, but here we kind of come back to the leaderboard. A random 120 completed in 4:37 contains a huge amount of variables. It was probably done in one key, and the maps/mobs/pylons/boss might be good, bad, or some good, some bad… we just have no idea. 150 in 4:37 has to be the product of a long endeavor- many keys spent, and most of the conditions refined to a state of near-perfection. Because otherwise, somebody else would have done better.

I’m laying this all out there in the hopes of trying to steer back towards my original intent, which was to wind up with a tool that helps us analyze and differentiate good clears. I don’t think that analysis needs to be (and, it obviously cannot be) perfect, in order to still be useful.

More and more of the leaderboard has gotten crammed into 150, and some foolish people will treat any two builds that can do 150, under any conditions, as having the same power. Obviously, that’s very dumb, but even those of us who are a bit more “with it” kind of have our eyes glaze over when it comes to looking at pages and pages of 150 clears.

I mean, we all kind of have a “gut sense” for the difference between a GR 142 clear and a GR 145 clear. And, at this point, most of us understand that to go up each GR level, if you add 17% damage, and 2.3% toughness, the next level will play basically the same as the one before it.

But, what’s the difference between 150 in 9:32 and 150 in 7:44? My gut doesn’t have a damn clue.

43

Most of the numbers I share around PTR/patch time are purely based on my own experience playing the build itself, or a closely related build. I find it hard to guess without any play experience. What Rage is trying to do might make those guesses a bit better if we can get some sort formula out of it.

Taking a guess at the maximum clear is like looking at a heat map. The hottest point is like 5000 keys, perfect map, perfect gear, perfect play. Reality lives a few pixels around that point.

When I share max clears for DH builds, i do it within 500 keys because unobtainable goals are discouraging to players. Anyone who has some skill and really tries to reach the goal can do it, and those who are willing to invest the keys can actually beat the expectation and get satisfaction knowing they did better.

Will this need to be considered in the predicted max clear? Probably not — but maybe after the number is decided, survey the player for the map conditions and adjust the clear by +3GR, +2GR, +1GR according to the quality of what the player was served.

44

What i see is a web app that asks for a time, GR level, number of floors, map quality (select menu), pylons (checkbox), and rift conditions with some tool tips to direct the user it means elite affixes, mob type, guardian, pylon sequence etc (very poor, poor, average, good, great, perfect)

The time and GR level work per the standard formula in your post, the other questions just adjust the output.

No conduit… that is +2GRs, poor conditions another +2GRs, etc.

1 Like
45

That sounds like it could be a great tool, even if it’s not perfect. Though some of these menus/checkboxes require significant insight on the part of the user… for instance, in a “map quality” select menu, how would we distinguish a good Desert from a bad Festering? Because some Festerings can be as bad as noodley cave maps.

46

The tooltips need to be very educational.

Regarding the calculations, I actually remove Time initially…

GR Scaling
1 1
2 1.17
3 1.3689
4 1.601613
5 1.87388721
… …
100 5,628,599.048
101 6,585,460.886
102 7,704,989.236
103 9,014,837.407
104 10,547,359.77
105 12,340,410.93

If someone clears GR100 in 700 seconds I run the following:

(900/700) = 1.285714286 ← may not be perfect

To get the effective GR I reference the table, do a lookup, then multiply this value by the corresponding value:

1.285714286 * 5,628,599.048 = 7,236,770 (GR101)

This is like your effective Power, I think your original post had this type of math in there. Based on those inputs, we’d start to multiply those other inputs against 7,236,770… (we’d come up with logic based on how many pylons found – oh bad pylon luck… 2 of 4 found, no conduit…, channeling is next best, maybe the Conduit is worth 1.33x and Channeling is worth 1.05x = 1.4x) This brings you to GR104.

The map, rift conditions are also bucket multipliers… if we are slick, we have the user reconstruct the rift. Use a repeater element to create each floor. With selectors for maps, mob type, etc. We’d assign values for each and calculate the multipliers that way. We could calculate how much traveling the user did on average based on the maps they’ve built.

This isn’t going to work for all types of builds, single target killers like Invoker only cares about how close elites are together, so we could make the map/conditions optional. or setup a matrix based on the set chosen, which disqualifies mob type multiplier in the map building and rewards multiplier when there are less floors.

The biggest work is figuring out the multipliers based on real tests. I would probably use Challenge Rifts to figure out what a Conduit is worth, by skipping it and comparing times.

47

If one takes all popular D3 builds atm and makes relations between these (weaker/stronger scaling showed with Math) then those builds, which maximum is below or equal 150, could potentially offer more data for those having maximum above 150.

Further, testing with P800 (other points off), might again add data due to lowering the maximum reachable GR level.

So, let’s say one builds such “map of builds”. Then trying a new build and relating it somewhere on that map will possibly discover you the maximum potential of that build.

Yes. After all, those most interested in such analyses are those who push. The question is - would it be worth the time spend to make something like this for a game like D3? If one has fun in the process - sure, why not. But otherwise it’s a waste of time since D4 is coming and D3 will be put to rest when that happens.

48

Haha, yeah I do have fun with this kind of stuff, though perhaps I am just a masochist!

Lol, this is getting complicated! And it assumes the user has basically as much info about the rift as though there was video. At a certain point, maybe it’s a lot easier to just say, “send us video, we’ll analyze it”. Because with video, a lot of the difficulties basically disappear. You can get very solid estimations of “move time”, instances where higher incoming damage might kill the player, uptime/downtime of Squirt’s (if used), close calls on “cycle” damage (i.e. CoE), Conduit effectiveness, length of boss fight, etc.

And, to be used as I initially figured, the tool would also have to have selections for if you only know what can be gleaned from the time and the in-game build/clear snapshot.

A web tool that incorporates all that would be a great time-saver for adept users. But to become an adept user, I think “average Joe” would need to spend significant time with an in-depth tutorial.

Are we talking seriously about this? Developing an app is not really my schtick, but if that’s something in your wheelhouse, we can keep talking about the framework. I mean, we can keep talking either way, but are we just spitballing?

49

Mathbarb=masochist…

1 Like
50

Yeah, this is why I decided to sort out the various “varieties” of time, and scale them differently. Of course, in addition to “Regular time”, “Conduit time”, and “Stricken time”, I would now also add “Move time”, which should generally not scale up as the tier increases.

With video, you could get a pretty clear number for move time. Without it, you could either make different allowances for different builds, or just use a sort of overall figure. I think something like “60 seconds, or 20% of total, whichever is higher” would work well in most cases.

At high power/low tier, the whole rift is basically move time, right? You don’t so much have fights with your enemies as just run right through them, which still completely overkills them.

I guess that raises the problem that you could say, “well, let’s see… GR 1, eh? Did it in 60 seconds flat! That’s all move time, doesn’t scale up at all. So, +149 tiers, this build could obviously do 150 in 60 seconds!”

Obviously not. Maybe you’d have to add a certain flat amount of time for each tier, with that amount discounted, depending on how much Regular time there is. So in the complete absence of regular time, you add the whole amount for every tier, but that amount scales down as Regular time increases, reaching 0 at maybe 4 minutes of Regular time.

Or, if you’re sticking to clears on a leaderboard, you probably don’t need to worry about this stuff- “60 seconds or 20%, whichever is higher” would probably work fine.

51

Those guys don’t click the ‘advanced’ checkbox, they get less precise numbers.

Haha, let’s make it a TV series called RiftBusters.

1 Like
52

Haha, that might actually be a pretty funny show… I’d watch it!

Tinne, if you ever want to message me off-forum, feel free. My full battletag, minus the # sign, at google’s mail service.

53

This is your slightest problem. You’d need tons of input data (not just info from GR clears, but also characters’ stats) to be able to determine the maximum of every build. If you cut the DPS to minimum for each build going above 150 where possible without affecting the build so that it now reaches its maximum below 150 and after that you scale this with the DPS, and you add all the other builds below 150 to make a map/matrix of all builds, you’d still need input data for new builds.

That’s the issue here - you can’t “solve” this aka just take all the variables from a new build and the algo to throw its max GR number. And because you’d always need new input data, it would be easier to just farm keys and play the game ROFL.

What they could have done with Challenge Rifts is:

  • Allow tier progression above 150
  • Put snapshots of popular builds/clears or such voted by players
  • Keep all snapshots for future play and their respective LBs forever

This would have given us precise data about builds with players pushing those they enjoy most as much as they want. As of now we would have hundreds of CRs we could choose from and push. And we would be playing, not thinking of absurd ways to solve the unsolvable.

54

Right, of course. I’m still not looking at this as an exercise in determining “absolute max clear” for builds, just a reasonable minimum that might be added on top, based on certain clears.

Trying to figure out the actual maximum for any build… that way lies madness!

55

What would be the practical implications of stating a precise minimum?

For a guy doing GR1 it probably won’t be a surprise he can go beyond GR9.

In the GR100-120 range players aren’t that much into pushing.

In the top end this might save the players some keys and time allowing these to skip some levels for pushing, but it would still be far less usable than finding a maximum, which would require enormous dedication and approaching everything more like a study than something for fun. And if someone decides to dedicate much effort and money to a project like this he better go after constructing the optimal aRPG itemization than a pushing table for D3.

56

Well, it’s a useful tool for looking at builds, especially those that can reach 150. It’s clear it isn’t going to be a perfect tool, but imperfection is a long way from uselessness.

Think of, I dunno, a hammer. It can drive nails well, and pull them pretty well. It can tap a big board into place, a little at a time. If you tie a piece of string to it, it can be a decent plumb bob.

But it’s useless for driving screws, or taking them out again, or for adhering things together without puncturing them, or for changing the color of something. Even the things it does do, another tool can do better- a nail gun for driving nails, the right pair of pliers for pulling them out. None of that makes it useless.

Your “No Paragon Leaderboard” was much the same- not a perfect solution to the issue- since paragon can jack up your damage in ways semi-independent from mainstat (like reaching a toughness threshold that lets you swap a defense ring for CoE). But that doesn’t make your approach uninteresting, or somehow a worse solution than not accounting for paragon, at all.

At the end of the day, I think even the approach I outlined in the OP, without any of the helpful additions + amendments proposed since then, is better than assuming either “all builds that can do 150 are the same” or “builds that can do 150 are different, but there’s no quantifiable way of distinguishing them.”

Imperfect, but still useful.

57

Very good read, but something I always wonder about the 15 minute timer.

Can we safety say,

  1. Every Build as the potential to clear GR150 within 15mins and if so what would the situation be?

I would like to see some internal testing from Dev and how they came up with this 15 mins and GR150 as the measuring stick for the game… Was there a magic number in paragon that each class/build need to reach to be able to complete this feat.

58

Definitely not! There are plenty of builds that couldn’t clear 150, even at 20000 paragon.

59

My point is not that it’s not useful rather that those people interested in pushing probably won’t end up using such tool more than once. We have to see of course how far this project will go.

15 minutes are good design for a quick game no matter what genre is your game. That is the main drive here.

For GR150… They reached it with power creep and realized if they add more power creep and GR levels the gameplay experience on below tiers will suffer from that. It’s suffering even now. GR levels should not go above 100 so that upgrading remains fun.

60

500 keys :slight_smile: Great… and that is if all the moons and stars align…

61

That’s fair enough. I certainly don’t forsee people flocking in droves to make use of this analysis. I’ll probably use it semi-frequently, though, and I wouldn’t be surprised if a few other people, like dmkt and Tinne, use it too.