The Truth about PVE scaling and Roster Levels

Eddiemon

fmftint wrote:

Back again with more truth.
The last time Meet Rocket and Groot ran the average of my top 4 characters was 125, this current run my top 4 average is 166.
Here are the results

Surprising to me, trivial nodes started lower than previously, all other nodes were significantly higher

LEVELS MATTER

With your test, how did you manage your intervening PvE events to ensure that you didn't scale up or down as a result of historical performance?

Given that you levelled your characters up one would expect that your performance curve also increased during the intervening events.

This could be expected to push your opponents levels up. How did you allow for this when determining that only the levels affected your opponents.

Additionally do you have any reason to believe that D3 didn't tweak the opponent levels? Given that some went up and some went down that would be a more logical inference than one where your characters' levels caused opponents to both rise and increase at random increments.

Unknown

Levels do matter. So does scaling based on performance and community scaling. But character levels do play big role, and you can easily see that by top pve alliance having mostly 94s. If it was constantly going up, a 2* players at one point should either hit point where he can't do nodes, or hit 395 opponents. But it doesn't happen. I'm playing nearly half year going for average of t10-t20 finish and my opponents levels, while high, stayed relativly same between events (until I leveled few chars a bit up).

Look from other angle - there is something in the code that caused level 1 enemy become level 6 enemy by leveling multiple characters to 270. That's nothing. It makes no difference. So in that case, why is there line of code that makes it happen? You could delete it and it wouldn't matter, yet for some reason it exists.

What people seem to miss is that scaling doesn't have to be additive, but almost certainly is multiplicative.

Here is my personal theory based on the node shown on picture:
Venom in this picture have base level of 0. Depending on your top 3 characters in your roster, he gets bonus levels, say +1 for lvl50s, +2 for 94s, +4 for 166s and +6 for 270s. So he will be level 2 for 94 roster and level 6 for 270 roster.
Personal scaling is a multiplier that increase and decrease depending on your performance. Why? Because as multiplier, it makes sense that increasing roster level change single digits on nodes. It have a cap, as it's impossible even for 94s to ever reach level where they can't compete for top spots. Let's assume that cap is 50 and accounts reached it by constantly winning top prizes.
2 * 50 = level 100 Venom for 94 roster. 6 * 50 = level 300 Venom for 270 roster.
Then you have stuff like community scaling and scaling based on you clearing node, but those can be smaller multipliers (like node level *1.1 for every clear).
Whether this is the way it works can't be said for certain. One thing is sure though - if you have 2 people constantly winning events, one with 94s, one with 270s, their nodes won't differ by 5 levels.

Unknown

I agree. I hate the "bad science" of the original post. Take a completely unrealistic scenario and make a whole load of nonsensical inferences based on the bad data.

An analogy would be building a PC with a processor, hard drive and completely insufficient RAM and saying the performance is **** - then doubling the processor and finding performance is still ****, then making the deduction the processor speed doesn't impact PC performance.

If you gave a MPQ test system to 5 of the brightest brains here, then they could probably work out the entire algorithm in a day, but much better to drip feed misinformation.

orbitalint

ephyzephy wrote:

I agree. I hate the "bad science" of the original post. Take a completely unrealistic scenario and make a whole load of nonsensical inferences based on the bad data.

An analogy would be building a PC with a processor, hard drive and completely insufficient RAM and saying the performance is **** - then doubling the processor and finding performance is still ****, then making the deduction the processor speed doesn't impact PC performance.

If you gave a MPQ test system to 5 of the brightest brains here, then they could probably work out the entire algorithm in a day, but much better to drip feed misinformation.

Great analogy - I take issue with the idea of direct vs indirect impact of roster levels on PVE scaling because this is a system working together to create scaling rather than isolated parts working independently.

Do levels directly impact PVE scaling? Evidence presented points to "not really." However, performance on a node definitely plays a factor in scaling on the non-trivial nodes that are both community and personally scaled. And you better believe that the general ability to clear nodes without taking damage impacts scaling, which roster levels indirectly plays a role. Heck, I could increase my scaling with Winfinite and it had nothing to do with their levels BUT scaling increased because it thought I didn't take a "fair" amount of damage in that match to warrant my current scaling or whatever other metric they use.

However, on average, 166 characters take less damage as a % of total health and have stronger abilities vs a lvl 94 character and a similarly leveled PVE node. So, the system accommodates that by increasing the scaling after it recognizes the 166 roster because of the lack of damage taken or whatever and in theory, does so until it finds equilibrium or level 395. So saying roster strength plays no role ignores this second fact completely.

The problem, based on accounts, is that there are scaling thresholds or something that you can't go below once you cross. It ends up being where leaning on 270's exclusively made the system think you deserved the 395 treatment exclusively because of how you were able to clear nodes with relatively no damage a majority of the time. XF/4hor made it easy but now you can't intentionally do anything to get below that new floor you set.

I honestly don't know what to do for those people, you can't reset everyone's scaling because a portion of the people would game the system knowing a little bit more about how it works than when they first started but it drives away people because it is so much work to always clear 395 hard nodes.

fmftint

Ok Thick as Thieves again, some of the nodes have changed order but the flavor text has remained the same so a comparison can still be made.

Last run my T4 average was 161 this time event/weekly boosting has made my T4 average 171


Family Reunion I
29/29/29______________28/29/28

Family Reunion II
40/35/34______________40/31/31

Family Reunion III
56/56/56______________54/54/54

Family Reunion IV
120/120/120___________122/123/123

Family Reunion V
161/161/161___________173/174/174

Family Reunion VI
No Data_______________162/163/162

Family Reunion VII
No Data_______________182/183/183

Captain America
85/85/85______________103/103/103

Storm
108/109/109___________115/115/115

Iron Man
No Data_______________98/98/98

Trivial nodes were within a few levels, Easy and Normal as expected went up. Again LEVELS MATTER

orionpeace

Yeah, can a mod please edit the title of the thread to something closer to the actual truth?

Something like "A Terribly Unrealistic and Skewed View of PvE Scaling and Roster Levels: The Lie We Want You to Believe".

Thanks!

loroku

XandorXerxes wrote:

Please be careful not to conflate a few different things.

The original post is only relevant to the effect leveling your characters has on your PvE opponents. When you start talking about how higher level characters deal more damage and thus level enemy goons, that's a distinct effect that this post isn't addressing. Jamie and IceX are pointing out that if you use Ares, OBW, and (2*) Daken to take on PvE, then leveling your Punisher to 166 isn't going to affect PvE substantially (if at all). If you start using Punisher in PvE, then you will bring other factors into play. They are saying that there's no point in saving your ISO and keeping your Punisher at 94 (especially if you're not planning on using him in PvE).

...

So yes, be aware that using higher level characters in PvE will cause your enemy levels to increase. However, the act of simply leveling those characters does not.

Yeah, sorry IceX, sorry JM, but the larger point about how this game works is still: if you have lower level characters, your enemies will never get too high a level. You can say with certainty that the actual level of the characters doesn't affect starting enemy level - awesome. That point is 100% worthless for PvE, because the truth is that by PLAYING the game with high level characters, you DO increase scaling. However it works, the end result is still the EXACT SAME: higher level characters leads to higher scaling.

The ONLY way this information is useful is to say that if you have EXCLUSIVE characters for PvP, feel free to level them up and it won't break PvE.

Unknown

I think the Heroic PVE during the Ultron events with ridiculously low scaling showed just how much of an effect community scaling has. Every heroic that doesn't introduce a new character tends to have much lower community scaling.

Unknown

Jamie Madrox wrote:

So there's how much the roster increases your scaling. 5 levels on a 1* character. The rest of what you see when you enter the node with your roster is based on your personal scaling and community scaling. So level up those characters and have fun.

Is that supposed to be a proof or a joke? It's pretty easy to have a formula that have, say MULTIPLICATION somewhere in it -- so the diff is just 5 levels while your other personal bump is close to 0, but as soon as it has reasonable value the same roster-based diff can create 100 or 500 levels worth of difference.

Certainly it could just stay at that 5 too -- just the experiment provides no real info to jump on. If IceIX was actually involved he could just share the formula and we would be actually wise. (Yeah, I know no chance.)

Also if the roster DOES play as much as making that 5 levels of impact at max strength -- if it is that much only throughout the next logical question would be why is it in the formula at all? With all the fuss and rumors around, if 5 lvls is all the intended bump why not scrap it and declare clear NO IMPACT from there?

One could just put it in the "braindead design" box with all the rest of MPQ, while others just keep their reasonable doubt -- either way the intended "no fun for you" state is maintained.

TxMoose

since nobody know how scaling works, i'll just say this - mine has gotten wacky since ultron.

fmftint wrote:

Last run my T4 average was 161 this time event/weekly boosting has made my T4 average 171

Again LEVELS MATTER

fmf - my average boosted was lower (avg of top4 is 168.5 boosted), but keep in mind that includes an undercovered storm 2/1/5 at level 198, my Lthor at 166, imhb at 160 and obw at 150. BUT, my levels of the last 3 nodes were higher than yours - IF was 180, daken was 200, and the goon node was 240. that is just dumb to start there with my roster because after what I listed there, pun/squirrel/mystique are boosted to the 140s (none are useful), loki at 140 (used him until he died against IF/falcon) and the rest are 104 down. I actually used a 104 cap/104 Cyclops/94 mnmags to take down the 240 goon node at the end and didn't take any damage, so I guess it needs to be 280 next time - will get to a point any cd ability will one shot any of my team, so it becomes 1) don't take any damage at all or 2) wipe out and start over if you have enough health packs - no middle ground. why the last 3 nodes are set up this way is beyond me. but I'm about to be scaled out of pves. not fun any more.

so, back to my boosteds - thor and obw are legit covered characters, but the other 2 are undercovered and other than storm's black, not very useful. if boosted levels matter (who knows), recently released 1-2 cover characters that are boosted 90 levels shouldn't count against you. if levels truly don't matter, then something else is wrong. but I'm sure it is....

....working as intended.

GothicKratos

Here's how I understand it (and does not reflect how I feel it should be, etc);

Champion Level of your highest three Champions affect the starting levels of your nodes. The 'top three' consideration does include the Featured Character Buff, as far as I can tell.

You're actual scaling, as in how your nodes level up across the event do so in two similar manners; personal scaling and community scaling.

Your personal scaling goes up or down depending on how efficiently you do in your node clears. If you do really well, you're going to get harder nodes. If you do really bad, you're going to get easier nodes.

Your community scaling is similar, but it's based on a playerbase metric - so if a lot of people are repeatedly clearing the same nodes, then that node is going to get harder, but if a node is not getting a lot of clears (or subsequently, a lot of wipes) the difficulty goes down.

This seems to fall in line with my own experiences, as well as players like Bee or Colog, whom are both relatively vocal about how good their scaling is, despite having full rosters and placing extremely well in basically every event ever.

====

My understanding aside. is this the "right way" do to it? I think some things need changed, I really do, but I think they have the right idea.

Personal Scaling is how it should be, in my opinion. Firstly, on a game theory level, it encourages creating a deep roster of Champions, and no just gathering the creme of the crop. Secondly, on a logical level, let's say you've got a bunch of tasks (nodes) you need to get done and you only have so many employees (Champions/Health Packs) to get it done - what're you going to do? You're going to send people on to do the tasks they do most efficiently and pass the rest off to people whom will do it better (i.e. shifting easy nodes off on 1*/2* Champions and loading the 3*/4* nodes off on your A-Team). You're not going to send in the new guy to do the complicated tasks, nor are you going to send in the veterans to do the menial tasks. Perhaps the algorithm for how much it scales up or down could use some adjustment, but the system itself is on point, in my opinion.

Community Scaling is bum. It should go as soon as possible. Just because the community is having an easy time with a specific node doesn't mean that I am, and you're making it even harder for me to pass something that might just be too hard for me anyways. Besides that, it's too multiplicative with standard personal scaling. Your progress shouldn't be gauged or gated by the communities.

There's clearly a 'secret' value ("MMR" as most people coin it) that makes your personal scaling somewhat persistent over events. This, in my opinion, needs to be reset, and it should be reset relatively frequently. Quarterly sounds fair to me - you don't want it to be too frequent or else people will just not care and keep playing how they are now but without real penalty - but you also don't want it too long either otherwise it just gets out of hand. Outside of that, I do feel for the folks that drove their scaling up prior to this information being common knowledge, and it seems it is quite difficult to drive back down without completely throwing a few events, and I feel like that's just not something that should be asked of a user.

[EDIT]: Just a small bit I forgot to add. Rubberbanding should come back. The biggest problem with rubberbanding is that it completely screwed people outside of the US. Since there is Time slices now, that problem should be mostly alleviated. It should be small. It should decay as each sub gets closer to an end. It should completely dissolve within a certain window of each sub closing (an hour? two?). It should continue to decay as you get closer to a certain group of players (top 10%?). However, this is a blessing in disguise for casual players, which seems to be what they want to cater to. It doesn't really hurt the "hardcore grinders" at all, because they'll still grind their hearts out and they'll still win until they miss a couple refreshes in a row.

There's my two cents. Nobody is being deceived. Nobody is being lied to. There's not a conspiracy.

fmftint

GK, I've been using the T4 average because we know that is the metric they use to scale TUs earned/received, it only seemed to make sense that starting levels would use the same metric

simonsez

GothicKratos wrote:

Rubberbanding should come back.

We have 8 hour refreshes now. If someone can't find the time to play once every 8 hours, we don't need to help them catch up when they finally decide to play.

Born2DieNPvP

I swear the more I read about this issue the less I understand. And that's not a knock on anyone who's posted info. I can't shake the feeling the devs have decided there is a certain way they play the game that they find fun and if we don't that's too bad. On their Q&A vids they make it clear they want people to explore different combos of heroes because that's what they like to do. What if I don't? What if I want my XForce to punch 3 nice holes in someone's face over and over while LThor and Doom play wingman? Should I be punished for that with nerfs and insane scaling because my team is actually effective? I, like everyone else, was once a new player and got stomped on in PvP and PvE. Comes with the territory in any game. It didnt make me whine and rage uninstall. It motivated me to play more so I could build my characters and get new ones. I loved every minute of it for the most part. And when I FINALLY get a team built up I've been working on forever through countless hours of grinding and real money, I got about 3 weeks of some of the most satisfying game playing ever of any game. I could finally hold my own at the goals I had set for myself. And i was excited for future team combos i had in mind. I didn't regret the hours and money I had sank into the game and planned on investing even more. Then like a nuclear bomb... MMR, Scaling, Nerf after Nerf with unfair compensation, health buffs, etc. My team is now so much **** because XF is nerfed but he also is so high level that my scaling is monstrous. I refuse to go broke buying health packs and boosts after every other increasingly longer fight. I have saw brilliant ideas on here how to address all these things but it's met with silence. I don't expect to rank in the top 11-50 every time. I know not everyone can win. But I do like to think that if I invest time in my characters I will actually be able to feel a sense of progression, not punishment for actually doing well. And I think that's at the core of everyones frustration. A lot of people still remember when that was what drove us and what made this game so much fun. It's not a case of all the vets just getting burnt out because of the age of the game, it's because it is no longer the same game. And, try as they may, are hard pressed to define just what it has become. People wouldn't be so emo if they weren't somewhat vested in this game's obvious potential and want to see it succeed.

GothicKratos

simonsez wrote:

GothicKratos wrote:

Rubberbanding should come back.

We have 8 hour refreshes now. If someone can't find the time to play once every 8 hours, we don't need to help them catch up when they finally decide to play.

I can agree with that. I simply like the idea of light rubberbanding because it allows life to be life. Sometimes once every eight hours isn't a possibility and being able to hit progression rewards still would be a nice gesture.

fmftint wrote:

GK, I've been using the T4 average because we know that is the metric they use to scale TUs earned/received, it only seemed to make sense that starting levels would use the same metric

I said three above because there was a CS response that claimed it was top three. Who really knows lol.

Oldboy

Do the devs (or anyone else for that matter) really know how scaling works?

Is like asking

Do cascades happen "randomly" for the AI?

6 pages of inconclusive discussion posts. I strongly suggest the devs reset all enemy levels in PVE to base levels in the next event and let us evaluate how our performances in the event affect scaling. We can then compare it here for a more meaningful discussion

TL;DR: nobody knows. Go do something else instead of this

Unknown

There was one 270 character in my entire top10 in the HB pve. The vast majority of players in every top5 pve alliances have low level max cover rosters, and they do well every single event, so please don't tell me they're all taking 25 minutes to clear a node, and don't tell me average roster level doesn't matter. Clearly keeping your roster at a low level has a significant effect on scaling. As it stands I have to go into the final grind with at least 6h on the clock if I want to compete, do nothing else while I'm grinding the nodes, and I still fall out of top competition because there is always someone who can start the final grind with 2-3h on the clock, so they have a higher amount of possible points available to them. I am hard locked out of competition, and it's been making playing the events a real slog, and unfortunately, since they only release characters in those, an obligatory one.

Unknown

One way to look at how broken scaling is that it took someone like Thor 4* or X Force for the max rosters to consistently come out ahead over someone with a low level roster, and it was not a big margin. As broken as those two characters are, they're only slightly more powerful than the scaling advantage. Having a specially built roster is likely at least equal to having 50% of a broken character that you can use that a max roster cannot. Yes the max roster will win when we get to use 100% of a broken 4* or 3*, but lately D3 is getting pretty good at about squashing these things, and note that it isn't particularly that hard for someone with a specially built roster to max out a 3*.

Linkster79

So to sum up. Level characters as much as you want to, just don't use them efficiently. Is this right?

I absolutely refuse to believe that any game developer would come up with a system that punishes efficient use.

Unknown

Linkster79 wrote:

So to sum up. Level characters as much as you want to, just don't use them efficiently. Is this right?

I absolutely refuse to believe that any game developer would come up with a system that punishes efficient use.

I don't think scaling has factored how much damage you've taken for a long time. If it did, it'd be very easy to unwind your scaling because you don't even have to try to end up with taking a stupid amount of damage fighting level 300 guys. I'm pretty sure the original bit about how scaling isn't dependent on your roster's levels is one of those Jedi speak thing because scaling is really based on the character you used and in theory, you could simply never use anybody you leveled up the entire time because the first thing anyone would do when they level up their best characters is to not use them in PvE.