TLDR Summary:
A mid-level 3-star player should take roughly 100 total damage per enemy level to keep personal scaling in check.
You can level up your roster without affecting your personal scaling too much if you are willing to take more damage in PvE.
Scroll down for fancy data plots.
Introduction:
The mechanics of PvE scaling have long been a subject of debate in the community. An abundance of anecdotes coupled with a dearth of data have generated much confusion and anxiety among the playerbase. Despite many appeals to clarify this issue, the developers have been reluctant to provide much detail about the scaling system. Perhaps it is too complicated to explain easily, or there is a fear that more knowledge of the system may allow players to exploit it. Regardless of what engenders this reticence, it seems like the scaling system is something that players will just have to figure out for themselves.
The mixture of personal and community scaling in PvE made it difficult to untangle their effects in the past. When community scaling was disabled in late June however, I was encouraged to start looking at personal scaling in more detail, so I began collecting various data in a spreadsheet at that time. The hope was that even without any detailed knowledge about the underlying mechanics of the system, one could still obtain some phenomenological results to guide gameplay decisions. The advice offered above in the summary is one such result.
Nomenclature:
Some terms will be defined below to avoid confusion:
-
Node level is the level of an enemy in a PvE node. This level does not normally deviate by more than one between the three enemies in a node, so I will address each node as if it had a single level.
-
Initial node level is the starting node level in a sub event before the player has completed any battles. Many sub events do not reveal all nodes to the player before several battles have been completed however, so it is not always possible to determine all initial node levels accurately.
-
Lower/upper/essential nodes are the three categories of repeatable nodes that can appear in PvE events. There are at least three nodes of each type in a sub event. Lower nodes are the first nodes visible to players. Their level is capped, so they will typically have the lowest level of all nodes and are usually labeled by the game as having trivial or easy difficulty. Essential nodes are slightly higher in level and often labeled as having easy or normal difficulty. Upper nodes are unlocked only after playing the lower nodes, and are generally labeled as normal or hard difficulty but can scale up to deadly as well. The usage of lower/upper/essential nodes is meant to avoid any confusion that may result from designating nodes by their fluctuating difficulty level.
-
Community scaling is the adjustment in node levels that occurred automatically over time in response to how all players in a particular time slice are performing in each node. These adjustments to node levels were generally quite small individually, but they accumulated continuously over the entire course of PvE events and were the primary drivers of extreme node levels in the past. Community scaling was thankfully disabled in late June.
-
Personal scaling refers to the adjustment in node levels that may be triggered by completing battles. These adjustments tend to come in relatively large chunks. A level 120 node could easily gain or lose 15 levels in a single tick for example.
-
PvE skill rating is a single hidden rating for each player that is supposed to represent their performance in PvE. We can be reasonably certain that this rating exists because the developers have referred to it themselves, and it would also be the most expedient method to implement a scaling system in practice.
-
Scaling hell is the result of a negative feedback loop triggered by the current scaling system. Increased node levels may force players to deploy their best teams and use boosts, AP stealing/denial, or stuns to prevent increasingly dangerous opponents from ever firing off a single power. If the player is successful, the game will usually underestimate the difficulty of the battle and further scale up the node level.
-
Scaling ticks are the discreet jumps in node level that are triggered by community or personal scaling. For personal scaling, there appears to be a fixed order in which nodes will scale up or down. This is consistent with a model where each node has a different set of break points that determine when it will scale up or down based on the skill rating of the player. See the code block below for an illustration of this model:
current skill rating of player ↓ Low Rating <-------------------------▼------------------> High Rating Node A: <------95-----------102-----------109--------> Node B: <---------------120-----------132------------> Node C: <---------142---------158-----------175------> Actual node levels due to current skill rating... Node A: 102 Node B: 120 Node C: 158 Next scaling tick that can be triggered by higher skill rating... Node B: 120 -> 132 Next scaling tick that can be triggered by lower skill rating... Node C: 158 -> 142If this is correct, then nodes that are supposed to be in ascending order of difficulty can appear to be out of order because the player’s skill rating may just happen to sit above or below some breakpoints. This would be consistent with the observation that the level of upper nodes can often appear out of order when it would make more sense for them to appear in ascending order of difficulty.
Methodology:
Past experience indicates that playing to one’s full potential would merely provoke a rapid inflation in node levels, with no respite until one finally gets trapped in scaling hell. Playing MPQ like any decent gamer would thus generate a poor data set, as one may not trigger enough scaling downticks to draw any useful conclusions. To improve data quality, I adopted the more unnatural play style of trying to maximize damage taken in battle while still securing a win. We can be reasonably sure that damage taken is the most important factor in personal scaling since the developers have repeatedly stated so in the past [1] [2] [3] [4] [5], and it agrees with the experience of many players as well. To help reduce possible interference from roster levels, I also tried to avoid leveling up any characters during the course of a PvE event.
The following information was collected for nearly every PvE battle I fought since community scaling was disabled.
- Date and time of battle.
- Even, sub event, and node name.
- Number of enemies in node.
- All node levels in a sub event, both before and after playing each battle. I used the maximum level for convenience, but also tried to ignore enemies whose levels were at the minimum of their star rating.
- Total level of the highest level team that could be assemble for each node.
- Level, maximum health, initial health, and final health for each character that was deployed in battle. All temporary healing was ignored and only true health levels were recorded.
- Number of loaned characters available in each node.
Not all information was collected for all battles however. I did not start recording a detailed breakdown of my team’s level and health until about a third of the way through, and data for the total level of my best team for each node was only added around the halfway point. Information that could reasonably be construed to embody player performance but was too tedious to collect was ignored as well (e.g. turns used in battle, matches made, AP collected, powers used, etc.). Despite these shortcomings, some simple but useful conclusions could still be drawn from the collected data.
Data Plots:
Below is a plot of the total health lost in battle versus node levels. Red and yellow points indicate the battle lead to an increase in level of some node in the sub event, whereas green and blue points indicate a decrease. If a battle generated multiple scaling ticks, they will be recorded as multiple data points. The dashed line marks the threshold for 100 health lost per node level. It is not the best line that could be drawn to separate the up scaling versus down scaling data points, but it is simple and easy to remember, and appears to be sufficiently accurate to be used as rule of thumb. Keep in mind that this data was gathered with a mid-level 3-star roster, over a period of time in which the average level of my 3-star characters have probably increased by 30 or so to reach the current state of my roster.
Some salient features in the plot above are as follows.
- The vertical spread of points around level 50 corresponds to the three repeatable lower nodes in each sub event. The initial level of these nodes were generally at 34/43/55 or 43/55/70 for me, and never scaled up or down in an sub event. The level of these nodes may correlate well with the initial level of other nodes in the same sub event, so they may be a useful bellwether in this respect. As I did not set out to investigate initial node levels however, it will take some effort to extract the relevant data to verify or reject this connection.
- The blob of points around level 100 corresponds to the essential nodes. This dense smear of data points in this region is a result of my essential node levels drifting up from around 100 to 110 over time.
- The data points from level 145 and onwards correspond to the normal/hard nodes in each sub event. Similar to the trivial nodes, I only encountered a few distinct sets of starting levels. The vertical patterns appear when I am able to complete many battles within a narrow level range by keeping personal scaling in check.
- While most battles (362) resulted in no change in node levels, there were still many more that lead to up scaling (83) than down scaling (26) despite my attempts to generate more scaling downticks. This discrepancy could have several implications. If personal scaling during an event cannot reduce node levels below their initial values, then any data I gather will be naturally biased towards up scaling. Should that not be the case, then my PvE skill rating ought to be drifting upwards over time due to this imbalance, yet my initial node levels have remained relatively stable over time. This would suggest that there may be some natural decay of the PvE skill rating in between sub events, or that initial node levels are simply not influenced by the skill rating.
- Only one node ever scaled above level 200, which was two or three ticks above its initial level. This occurred due to me blitzing nodes at the end of a sub event in order to secure some much needed hero points, so players who are not always in a rush can probably control their personal scaling in a similar manner and keep all node levels within a tick or two of their initial values.
- The heavy dependence of personal scaling on health lost in battle can have undesirable effects on the metagame. A scaling system that looks primarily at damage taken will favor brute force characters that have no damage mitigation but can sustain a lot of punishment (e.g. Thor/Ares). Characters that rely on protect tiles, healing, stuns, or AP denial/steal, but cannot absorb much damage will in turn suffer greatly (e.g. Hood/OBW). This is one factor that also accelerates the negative feedback loop leading to scaling hell. Since many high level nodes cannot be defeated with brute force alone, players frequently have to deploy support characters that will greatly exacerbate their scaling afterwards.
- At level 161, there is one scaling downtick that appears below an uptick. These two data points were from entirely different events, and may reflect different scaling parameters between them. The crossing may also be due to factors other than absolute health lost influencing the outcome here.
It is possible to generate a plot for percent health lost in battle versus node levels too.
The plot above is a good illustration of why personal scaling can feel so frustrating to players. Emerging from a battle with 70% health lost would make most players feel like they just barely eked by a very challenging battle, yet the game can still punish players by further increasing node levels because it decided that they still played too well. This is a significant flaw of the current system where the scaling behavior can deviate greatly from player expectations.
Let me note as well here that the high health lost expected from players in normal/hard nodes is not merely an artifact of the unnatural play style I adopted to collect this data. While the lower nodes and essential nodes gave me some leeway to deploy suboptimal teams, I rarely had the same luxury for the upper nodes, and frequently used a strong team with boosts to get through them.
Consistency of Data with Past Claims about PvE Scaling
Claim: The initial level of a node is based on the highest level team you can use in that node.
Sources: Demiurge_Will, more Demiurge_Will, and MPQ customer support.
Verdict: Possible.
The data collected so far is at least consistent with this claim, but the difficulty of running controlled tests on initial node levels makes it very hard to verify or reject this claim conclusively. The top three characters that one can bring into battle are typically the same for each sub event except for the essential nodes, and even then the levels may not differ much if you have an evenly developed roster.
When I began gathering data in late June, my highest level characters were in the 120s, with most others sitting in the 90s. I have since leveled three characters to 150 for PvP, brought a handful of others into the 121-140 range, and leveled up the majority of my 3-star characters to 120 if they had enough covers for it. My relatively stable initial node levels over this period of time would seem to contradict this claim, except that the three strongest characters I can bring into any node has not actually changed much due to character boosting. It is thus possible for this claim to be consistent with the data gathered here.
Claim: Character level and boosts do not affect personal scaling.
Sources: IceIX and more IceIX
Verdict: Partially true.
Character levels do not directly affect personal scaling, but they can do so indirectly by improving player performance, and perhaps by influencing initial node levels. The claim that boosts do not directly affect personal scaling seems very reasonable to me, but I did not collected any data about their usage.
Claim: The truth about PvE scaling and roster levels.
Source: Jamie Madrox / IceIX
Verdict: Consistency requires rejection of claims made by Demiurge_Will and MPQ customer support.
This well intentioned post carried a very confusing message, so I cannot even paraphrase it properly. Nonetheless, the content of the post can still be reconciled with the data collected if one assumes the reported test was performed as follows:
- IceIX creates a new account and populates its roster with maxed level characters shown in the first screenshot.
- IceIX joins the PvE event. The initial node levels would presumably be set at a minimum if they depend only on player skill rating, and we assume that new accounts start with a minimum skill rating.
- IceIX fights through several battles to reach the node shown in the second screenshot. Although it is possible for IceIX to skip battles using developer tools here, that seems like too important a fact to neglect mentioning.
- During the fights not mentioned anywhere in the post, the skill rating of this new account would presumably have increased due to the easy battles, and may have triggered several upticks in node levels. Because these scaling ticks would likely be distributed across all the nodes in the sub event, the particular node shown in the second screenshot may only have experienced a single uptick in node level that increased Venom’s level from 1 to 6.
- The screenshots get sent to Jamie, but some communication error results in him pointing at the second screenshot and calling it proof of a maxed out account experiencing “base scaling”. This conflicts with the definition given for “base scaling” just two sentences before because IceIX is very likely to have fought through several other nodes in order to reach the one shown in the screenshot.
If the above description of events is correct, then the actual message conveyed by the post is that initial node levels and the magnitude of scaling upticks are not significantly influenced by roster levels under the given test conditions. Unfortunately, this message would also contradict the advice offered by Demiurge_Will and MPQ customer support, who claimed that initial node levels are determined by the level of the best team available at each node. Given the lack conclusive data, this conflict may require developer input to resolve.
Conclusions:
Below are a few things we can learn from the collected data.
- The amount of damage taken in battle is the primary factor in player skill rating, which in turn determines how node levels are adjusted during an event through personal scaling.
- It is possible to trigger scaling downticks on purpose by taking lots of damage in battle.
- A mid-level 3-star player will need to take roughly 100 damage per node level to keep personal scaling in check.
- Sturdy character with self damaging powers like Carnage, Sentry, Ares, and Juggernaut are invaluable in PvE due to the ease with which they can trigger scaling downticks.
- Flimsy support characters that generate protect tiles, heal, steal/deny AP, or stun enemies can inflate your skill rating and ruin your personal scaling if they mitigate too much incoming damage.
- The makeup of the enemy team in terms of heroes or goons has no obvious effect on personal scaling.
- It remains unclear whether initial node levels are determined by roster level, best team available, or player skill rating. Developer input is likely required to resolve the conflicting official information.
- Node levels have not been observed to drop below their initial levels through scaling downticks during a sub event. Though I did not test this by repeatedly suiciding my entire roster in battle, it is possible that initial node levels are also minimum node levels for the duration of a sub event.
- Players who are willing to exploit the personal scaling system’s reliance on damage taken in battle can probably relax the soft caps on their roster. Better PvP performance with a higher level roster still requires a sacrifice of deliberate poor play in PvE however.
