DEFCON 5: Defender vs Controller - Which is the Better Support AT?

[Dark Current]

With DEFCON 5's Verdict behind us, we're taking things to DEFCON 4... NATURE AFFINITY!

 

DEFCON Level 4 pits a Nature / Sonic / Energy Defender against a Dark / Nature / Dark Controller in a locked cage match to find out who is superior at supporting their team!

 

 

-----------------------------------------------------------------------------------------

 

Starting in the Blue Corner... an Aerial Guardian defender, Skreaming Tree!

 

 

MIDS Build: Screaming Tree - Defender (Nature Affinity).mbd

 

Build Discussion and Solo Showcase Video: RELEASING 12/29 6 AM

 

------------------------------------------------------------------------------------

 

And in the Red Corner, a Night Terror controller, Fey Wode!

 

 

 

MIDS Build: Fey Wode - Controller (Darkness Control - Nature Affinity).mbd

 

Build Discussion and Solo Showcase Video: RELEASING 12/30 6 AM

 

Place your bets now!

 

[Dark Current]

Skreaming Tree makes his debut punishing Council alien collaborators.

 

First part of the video presents an in-depth look at his Mids build.

 

Edited December 29, 2024 by Dark Current

[Dark Current]

Fey Wode has arrived!

 

 

Defenders across the shards are quaking in fear.

 

Edited December 30, 2024 by Dark Current

[Seed22]

As much as I hate to be a stick in the mud, it’s corruptor. That’s the best support, but if you’re talking strictly between troller and fender it truly depends on the performance of the team and how well they capitalize on said support.

 

A troller at or near 50 will mitigate the damage wholesale via solid CC, which the team could capitalize on for smooth times ahead.

 

On the flipside, a fender can mitigate the damage via overpowered buffs and debuffs that neuter the mob in their own way, allowing the team to continue in relative soft safety while contributing better damage outside of procs.

 

I’d have to give it to the fender given the current meta, if Corr wasn’t an option.

[Player-1]

Hello @Dark Current,

 

I appreciate your enthusiasm in comparing these two archetypes! This is something we do internally as well, but with a few more controls in place to limit randomness. Looking at what you have written, the goal appears to be comparing the "Support" side of each more than anything else. This is a great goal, but the way it is approached leaves a lot of room where different variables can impact the results.

 

You have started with comparing an Electric Affinity/Electrical Blast Defender and Plant Control/Electrical Affinity Controller, and now a Nature Affinity/Sonic Attack Defender to a Dark Control/Nature Affinity Controller. It appears that you want the Support powersets to go head to head, but the different Blast and Control powersets in the mix will inevitably change the outcomes for each round of comparison. Not to mention the power and epic pool choices. For example, if you leaned into Endurance Drain the Elec/Elec Defender may synergize much better than the Plant/Elec Controller and lock down entire encounters thanks to Electrical Blast, compared to an Elec/Sonic Defender which would not be able to do so compared to an Elec/Elec Controller.

 

If I were to run a similar comparison, powersets that either show the extremes with "Support" in mind (such as Dark Blast where you can apply a lot of -ToHit debuff) or ones that are very neutral with little to no secondary effects (such as Fire Control where secondary effects are limited) would be preferred. That would either be a showcase of the "best" of each Archetype, or put both on a more even playing field where only the Support powerset truly matters between them. If these comparisons were an Elec/Fire Defender, Nature/Fire Defender, Fire/Elec Controller, and Fire/Nature Controller, the playing field would be much more fair between the two to show how Blasts and Controls contribute to team safety alongside the actual Support. 

 

The other parts involving team members are more of a challenge as this opens up nearly endless variables to your comparison. The easiest example would be other Support players interfering with your metrics around safety. Say you had a Corruptor on the team with Force Fields while testing your Defender, and while testing the Controller you happened to have no other support. With the Defender, you would need some way to determine how much of the team's safety was solely due to your efforts versus the buffs from the Corruptor. With the Controller, you lack backup and it may negatively impact your safety rating comparatively. 

 

Teaming with the same players for each test, or creating an Architect mission where there are friendly NPC's to protect may be a better way to approach this as it would again limit the amount of variables that could impact your data. Running the same content as much as possible would also be critical as if one side gets a harder group of enemies or maps than the other, it will unfairly favor the one who had an easier time as being "better" through pure luck.

 

I look forward to seeing the rest of your results!

 

 

[Dark Current]

Hello, thank you for the feedback.

I appreciate the thoughtfulness and time taken to share your critique. As I’ve heard this and similar from a few sources, I think I need to clarify my position in regard to the design of the experiment. I am not disagreeing with any one point or saying anyone is wrong with their opinions - each flaw that has been pointed out crossed my mind at one point or another. I spent a good couple weeks before starting the tests puzzling over the variables and controls I was facing. I feel that my approach and the results I’m sharing with you are as honest and reliable as I can make them. To that end, here is my rebuttal to the critics:

 

Strengths of the Approach

1. Acknowledgment of Variability:

I clearly recognize the uncontrollable variables (teammates, enemies, missions) and am trying to mitigate their impact by collecting a large amount of data across diverse conditions. This is an accepted statistical approach, as large sample sizes often help "average out" noise, making it easier to detect meaningful signals.

 

2. Quantitative Analysis:

I am compiling data into sums, averages, rates, and indices, and applying statistical methods (e.g., R-squared, whisker-and-box plots, standard error), to leverage valid statistical tools to extract insights from the data.

 

3. Holistic Evaluation:

Comparing across multiple support sets (rather than just one) strengthens the generalizability of any findings I may uncover. It also acknowledges the complexity and variability inherent to archetypes in CoH - a highly dynamic MMO environment.

 

Challenges

1. Random Team Composition

Criticism: Different teammates, even with minimal incarnate powers and no other support sets, could drastically affect the outcome (e.g., a DPS-heavy team might make a support set look better than it is).

Response: This is a valid concern, but it’s mitigated by the breadth of the dataset. Over many missions, random variability in teammates will tend to "wash out" if the sample size is large enough. While this introduces noise, it doesn’t invalidate the experiment—it just requires a cautious interpretation of the results.

Supporting Principle: My approach aligns with the Central Limit Theorem: with a sufficiently large sample size, the effects of random variability diminish, and the mean of the sample (tested characters) approaches the true population mean (defenders and controllers in general).

 

2. Slotting and Secondary Powerset Variability

Criticism: The differences in secondary powersets (e.g., blast vs. control) and slotting could bias results because the synergy of primary and secondary sets differs for Defenders and Controllers.

Response: This variability reflects the nature of the two archetypes and is, therefore, part of the question I’m investigating. It isn’t a flaw but rather a feature of the real-world comparison I’m performing.

Supporting Principle: This aligns with the principle of ecological validity, which emphasizes that experiments should reflect real-world conditions when evaluating practical differences. Since no player uses identical secondary sets or slotting across archetypes, my testing mirrors actual gameplay scenarios.

 

3. Mission Variability

Criticism: Differences in mission goals, maps, and enemy factions could introduce noise that obscures the true differences between archetypes.

Response: While true, this variability reflects the dynamic nature of CoH gameplay. By analyzing trends across a wide range of missions, I’m testing defenders and controllers under diverse conditions, which adds strength to any findings I may make.

Supporting Principle: This aligns with the principle of robustness analysis, where varying conditions are intentionally included to ensure conclusions hold across a wide range of scenarios.

 

4. Training Bias from Repeated Missions

Criticism: Repeating missions could lead to learning effects, where I become better at running specific missions over time, skewing results.

Response: I’ve chosen not to repeat the same mission for this reason. I could, at a later date, include a few repeated missions as a control group to test for a learning effect.

 

Summary

My testing is not foolproof; however, it is based on four key principles from statistics and experimental design:

 

1. Central Limit Theorem:

As sample size increases, the mean of the data approaches the true population mean, and random variability (noise) diminishes. My large dataset (50 missions total) should provide a basis for detecting significant trends despite uncontrolled variables.

 

2. Ecological Validity:

The testing reflects real-world gameplay conditions, which include variability in teammates, enemies, and missions. This makes the results more applicable to practical gameplay rather than controlled, artificial conditions.

 

3. Robustness Analysis:

By including multiple support sets, teammates and mission types, the analysis is robust across diverse scenarios, ensuring any trends are not artifacts of a specific setup.

 

4. Statistical Significance:

If my statistical analysis (e.g., >95% confidence level) reveals significant trends despite variability, it strengthens the argument that the observed differences are real and meaningful.

 

I am not saying any of this to shut off dissent or concern, as I am open to both. Hopefully, I have made my case, and this response demonstrates that I did not enter this experiment blindly. It is my belief that the logical and statistical principles the tests are built on will allow me to detect any significant differences between defenders and controllers despite the uncontrolled variables I openly acknowledge are present.

 

Thanks!

Edited December 31, 2024 by Dark Current

[Bionic_Flea]

Great job in putting all this together and sharing your thoughts.

 

While I agree with you that having a large sample size can minimize noise and randomness, when you have sooooo many different variables I'd expect you'd need a sample size in the thousands to do so.

 

But hey, you do you! 

[Death2Tyrants]

Premise sounds interesting.

 

How much stick time do you have with each AT, Defender vs Controller?

 

So how are you picking what to pair with your support sets for each AT? How about power pools and Epics?

 

How are you selecting powers to keep or skip within a build?

 

Are you tracking the maturity of your teammate's builds (IOed and/or Incarnated) as a variable?

[RCU7115]

I would have run the test on each AT running the Posi TF's, Synapse, Penny Yin, Citadel, Manticore. Numi's TF's and  finish with an ITF.

I would also used the change name feature for evey TF so no one would recognize the toon along with changing the costume at the tailor.

To keep the names just make a new toon with the name and delete the lvl 1 toon after the test.

I also would have only run the toons with  either basic SO's or IO's. Once IO sets are put in the numbers actually get more skewed due to each toon having completely different bonuses.

Edited December 31, 2024 by RCU7115

[Maelwys]

10 hours ago, Dark Current said:

2. Slotting and Secondary Powerset Variability

Criticism: The differences in secondary powersets (e.g., blast vs. control) and slotting could bias results because the synergy of primary and secondary sets differs for Defenders and Controllers.

Response: This variability reflects the nature of the two archetypes and is, therefore, part of the question I’m investigating. It isn’t a flaw but rather a feature of the real-world comparison I’m performing.

Supporting Principle: This aligns with the principle of ecological validity, which emphasizes that experiments should reflect real-world conditions when evaluating practical differences. Since no player uses identical secondary sets or slotting across archetypes, my testing mirrors actual gameplay scenarios.

 

Whilst I agree that the influence of Random Team Composition and Mission Variability could in theory be drastically reduced to the point of negligiblility by running a sufficiently high number of trials... I think that you're missing a major issue here with Slotting and Powerset Variability.

 

You seem to be intentionally matching Support Powersets between the Defender and Controller ATs. Aside from a few outliers such as Dark Miasma and Dark Affinity, this is an achievable goal. However if you're also trying to factor in the effects of the Offensive powersets and Pool powers then in order to make it a fair test you would have to consider the effects of ALL POSSIBLE offensive Powersets - and ideally in combination with each Support set. Otherwise you can and will end up in a scenario where the Offensive Powerset that you arbitrarily pick for one AT will drastically outperform the Offensive Powerset that you arbitrarily pick for the other AT. And to a lesser extent the same is going to be true for pool powers.

 

As an example: Plant Control and Arsenal Control are both consistently top of the charts for Controllers. They have highly Procable AOE attacks, Short-Recharge AOE Confuses, a variety of CC and useful pets. Plant can even function as a Ghetto Healer with a Panacea Proc in Spirit Tree (and Arsenal's pet gets hardcapped resistances and Taunt). The performance ceiling of those sets is drastically higher than the likes of Ice or Symphony Control. However most Controller Primaries (with the notable exception of Illusion before Perma Phantom Army) can lockdown at least two full enemy spawn groups indefinitely.

Conversely, Dark Blast is about the only Defender set that can drastically improve team survivability, and when procced up it can also deal reasonable levels of Single Target damage (whilst Elec Blast can theoretically floor mob Endurance and Recovery, unfortunately only the toughest of targets will have their Blue Bars bottom out before their Red Bars eg. EBs and AVs - and these days those foes can still attack with zero endurance anyway!). All of the other Defender sets have at best a few middling low-Mag ST CC abilities, but bring wildly different levels of additional damage to the table (Ice Blast is good due to probability, Beam Rifle and Dual Pistols and Sonic Blast are good due to -res stacking, etc). Likewise, Defenders getting higher base scalars for debuffs/buffs means taking specific pool powers such as Weaken Resolve and Spirit Ward grants more benefit to them.

 

Aside from different raw base scalar numbers on pool powers; there is going to be a big difference in how specific pool powers mesh with different powersets. A Kinetics that can Combat Teleport into melee range for a split second to drop Fulcrum Shift and then immediately reappear at safe distance is going to be bags more survivable than one that relies on Sprint and Combat Jumping. And the same goes for any Defender with a PBAoE attack, especially a Nuke.

 

In short, unfortunately I really think there are too many variables here for you to "average out" unless you run an unfeasibly large number of tests (e.g. MANY THOUSANDS)... so as things stand I very much suspect that the only way you're going to be able to draw any useful unbiased conclusions here is if the variables are drastically reduced by standardising as much of the build (and the missions and level range and team composition) as possible.

 

However I appreciate doing that would not make for as much fun when playing your way through all these toons, and I'm sure the eventual writeup will at least be entertaining even if it won't be possible to draw many unbiased conclusions from the data set. So by all means continue to have at it.

[Maelwys]

I'm going to highlight something in particular here.

This is the build for "Ampere Avenger"; your Elec/Elec Defender:


Defenders have a low damage scalar and their Blast sets have minimal CC. The best way to get milage out of their Offensive powersets is to Procbomb them.

By not aiming for maximum damage output from your Defenders' blasts, you're unfairly favouring the Plant/Elec Controller even before the Top-Tier performance of Plant is considered; because Controller CC effects do not rely on specific slotting and neither does their additional damage from Containment (whilst your Controller's Carrion Creepers and Fly Trap would both really benefit from being Procced out - neither will *drastically* affect their Survivability or Clear Times unless they're soloing).

Your Elec Blast Defender's attacks are currently dealing damage like this (Assault + Interface active, but no Vigilance or Hybrid):

Charged Bolts: 84.79
Lightning Bolt: 132.7
Zapp (Quick): 185.2
Tesla Cage: 47.86
Ball Lightning: 84.71
Short Circuit: 75.14
Thunderous Blast: 316.3

Compare that to another Elec Blast Defender build like this one Defender - Sonic_Elec (Energize+TCage).mbd  
(again, Assault + Interface active, but no Vigilance or Hybrid and ignoring the effects of the Primary)

Charged Bolts: 99.76
Lightning Bolt: 239.4
Zapp (Quick): 304
Tesla Cage: 316.9 (more than 6 times higher!!)
Ball Lightning: 178.5 (more than 2 times higher and applies a -res proc)
Short Circuit: 134.5 (also applies a -res proc)
Thunderous Blast: 496.9

(They also have sufficient recharge levels that their single target attack chain is seamless, and the Nuke is up every ~51s rather than every ~116s)



That second toon's Elec Blast attacks can also drain the endurance of whole mobs in a few seconds and keep them at zero indefinitely if needed, but they will be contributing far more damage than the first toon as well as applying a rather decent chunk of -res. Even Tesla Cage (a power which is traditionally viewed as skippable for many top-tier Elec Blast builds!) will be doing far more for the team as long as you can keep your 'Static' stacks up.

I'm not trying to say that all the builds in your testing ought to be min-maxed and purpled out. But I wanted to highlight that the performance of the same Defender Blast set can very WILDLY whenever the character is built by two different people with two different goals; even if you completely discount any differences from things like Secondary and Pool Powers and Team composition etc. So at best you're going to be observing what the performance of one particular predefined character is, not what a powerset - let along a combination of powersets - on each AT is actually capable of.

 

Edited December 31, 2024 by Maelwys

[Seed22]

3 hours ago, Maelwys said:

I'm not trying to say that all the builds in your testing ought to be min-maxed

So this part actually is very important. If you want to test something like this, all of your builds NEED to be min/maxed. You can’t test the full potential of a set without first getting to its’ full potential.

 

Without minmaxing, the test are effectively heavily skewed and not painting an accurate picture of performance.

[Hjarki]

16 hours ago, Dark Current said:

My testing is not foolproof; however, it is based on four key principles from statistics and experimental design:

All of these mathematical tools are dependent on independence of your random variables - which isn't the case here. What you're attempting to do is akin to feeding a large number of chess matches into a statistical analysis to see which piece is most popular to move on a given turn - and expecting this analysis to help you win chess games.

 

While the graph of Chess has far more rigid connectivity than CoH, your confidence in your results is misplaced.

Edited December 31, 2024 by Hjarki

[Maelwys]

1 hour ago, Seed22 said:

So this part actually is very important. If you want to test something like this, all of your builds NEED to be min/maxed. You can’t test the full potential of a set without first getting to its’ full potential.

 

Without minmaxing, the test are effectively heavily skewed and not painting an accurate picture of performance.

 

Min/maxing is certainly one way to go about ensuring a level of consistency across multiple characters. But you'd need to establish a very clear set of build goals and priorities. Attaining a reasonable level of damage output without negatively impacting on Buff/Debuff enhancement slotting is doable; but after a certain point chasing additional damage needs to be weighed up against other beneficial things; such as +Defense/Resistance/MaxHP to ensure personal survivability. And I doubt that any builder worth their salt has never found themselves in the position where they had to choose between additional proc damage and additional global recharge; especially when an important buff is a few seconds short of perma!

 

Another way of going about it may be intentionally limiting yourself to only utilising specific enhancement types (e.g. no Incarnate Clickies, no HOs/Purples, or even "SOs only"). Although that'd be less valuable at showing "real world performance"... 😖

[drgantz]

For a long time, I've wanted to make a plant based support.  I've wondered if I should make Plant/NA Controller, or NA/Dark Defender. 

From the comments in this post, it appears that Def is better for a strong team while Contr is better for a weak team. 

I'll make a Def.  Thanks for the comments.

[Without_Pause]

Odds are far greater you get on a steamrolling team versus a struggling one.