test content
What is the Arc Client?
Install Arc

A deeper look into scaling & suggestions

elwing#6559 elwing Member Posts: 29 Arc User
edited May 2019 in Player Feedback (PC)
So as we all know, Mod16 brought us two new concepts: “counter stats” and “scaling”. I tried to get a deeper insight into both concepts at the current state of the game. I decided to note my stats for all (epic) dungeons, skirmishes and available epic trials for several characters with different item level, stats and player level.

I compared the following characters:
  • Ligula, a lvl 80 Rogue with around 22k Item level. She is my main and I'm working on having good/balanced stats for end-game content. Still overcaped at most stats except CA but hey, working on it... She has orange Alabaster weapons.
  • Maxwell, a lvl 70 Warlock with about 14k Item Level. He is just some alt I've done some campaigns. I gave him spare enchantments, mostly brutals, so he reach 14k Item Level to test stuff. There is now “big plan” behind his stats. He has purple Wootz weapons.
  • Maxwell, same lvl 70 Warlock, but with no enchantments and around 11k. Item level. Still purple Wootz weapons.
With all three characters I visited all available content in the game and noted my offensive stats.

You find a summary of my stats in this spreadsheet: https://docs.google.com/spreadsheets/d/1ynPsCICOVIuDVqnCFJKr0tdc7iOzjLMdyeAZHC6xfPE/edit?usp=sharing

I've calculated the %stats out of player and enemy stats and the expected value of a hit with 100 Magnitude and 100% CA uptime based on the formulas on Janne's homepage (https://jannenw.info/pages/mechanics16/formulas) (At this moment, the calculations aren't in the spreadsheet because I just copied my OpenOffice data – I consider to fix that at some point). I didn't take into account gear % damage, boons, weapon enchantments etc. because that is not the point here. I hope I didn't make any major mistakes in my calculations – if so feel free to correct me.

The suggestions I will derive out of the data are the following:
1) The scaling for MSP isn't right at the moment.
2) Skirmishes aren't scaled in a meaningful way. The scaling needs to be adapted.
3) Looking a the implications of the current scaling, a re-structure of the random queues is necessary.
4) Scaling of weapon damage has a massive impact on the overall damage.
5) Scaling brings players closer together – closer than it should be.

To be able to compare the outcome in the different places with different characters I plotted some data I found interesting. I try to discuss this urgent issues based on the implications of the graphs.

1) When you look at the damage modifier, you see a clear structure for the intermediate dungeons (MC to FBI) – they all have the same damage modifier. As you can see in the graph, there is a problem with MSP, it has (for no obvious reason) a lower damage modifier than the other dungeons. Because MSP is logically a part of the other named dungeons, its scaling need to be fixed.



Edit: I forgot to clairfy what is meant with "damage modifier". It is everything stat-wise combined what will buff your damage.
Simply put: damage = weaponDamage * MagnitudeModfier * damageModifier.

2) No if we look at the skirmishes (MotH to MBH). There is no clear structure visible. Of course Master of the Hunt has a way lower enemy rating (6125) the other have a rating of 7000. So you would expect that they are in some way similar. But they are not. Even worse, for this content it is possible that a lower level / item level character has a higher damage modifier. That isn't right.

3) Now for the “higher” content (Tomb, Cradle and CR). They have obviously a minor harsh scaling close to no scaling at all. So if you take a look at the overall picture, based on the scaling the following re-structure of random queues is meaningful (at least for me). Move Tomb, CR and Cradle into the expert queue.

4) Now let's talk about the weapon damage. I can understand you feel the urge to scale the weapon damage too. But if you look in the effects of that scaling, you can see that powerful weapons are reduced massively what has a direct influence to the overall damage additionally to the scaling of the stats. In the QA stream was mention, that “double scaling” was a mistake. But with scaled weapon damage you provoke a double scaling again. With scaled weapon damage the content is harder than before the patch (with the lower scaling). Please consider to either not scale weapon damage (because your damage modifier is already scaled) or at leas scale them less harsh so that it is meaningful to have better weapons.




5) My 5th point is that you scaling brings players closer together that it should. I see positive sides on the idea that players, that play together at different progression points in the game shouldn't be too far apart from each other. Like Ligula is doing trice the damage than Maxwell. But where is the benefit in investing into my character if it is only slightly better (or even worse) than a lower geared character? In scaled content Maxwell, a character I put now effort in, with random enchantments, is doing 2/3 of Ligula's damage. The closer together damage is an outcome of harshly capped weapon damage, and harshly scaled other stats (for example Power and Armor Penetration).







I hope my calculations help the community to get a deeper insight into scaling at the current state of mod 16. Additionally I hope that my descriptions help the developer team to fix some obviously non-fitting scaling.

Edit 2: I've added some colors...
Ligula Trickfoot
Post edited by elwing#6559 on

Comments

  • oremonger#9999 oremonger Member Posts: 213 Arc User
    @elwing#6559 Thank you for taking the time to put this together. I hope they reply to you.
  • nitocris83nitocris83 Member, Cryptic Developer, Administrator Posts: 4,498 Cryptic Developer
    Thank you @elwing#6559. I'll make sure this post gets passed on to the rest of the team.
  • dupeksdupeks Member Posts: 1,789 Arc User
    edited May 2019
    Excellent write-up. Please note that Armor Penetration and Accuracy are particularly important for damage output, and investing in them to meet cap is the best return on investment for stats.

    It's fascinating that in scaled content a higher geared character drops to well below ArPen cap compared to a lower geared character. As a result, anyone playing that character will be frustrated / want to have specialized gear adapted to the scaling in order to still meet these two important caps.

    The current implementation of scaling forces players to choose between carrying 2 sets of gear or running painfully suboptimal setups in scaled content.

    Edit: Is the damage modifier assuming full effectiveness of Armor Penetration and non-deflected hits?

    If yes, then that further explains why scaled dps feels less effective, since this graph isn't capturing the impact of not fully mitigating defense or deflection. Not sure on this point though.

    Edit2: The above graph already takes into account counterstats.



    Edit: I forgot to clairfy what is meant with "damage modifier". It is everything stat-wise combined what will buff your damage.
    Simply put: damage = weaponDamage * MagnitudeModfier * damageModifier.

    Post edited by dupeks on
  • elwing#6559 elwing Member Posts: 29 Arc User
    I plotted the graphs for the other offensive stats: Crit Chance, CA Severity and Deflect chance of mobs. They additionally underline my points from the first post.





    Ligula Trickfoot
  • elwing#6559 elwing Member Posts: 29 Arc User
    dupeks said:

    Edit: Is the damage modifier assuming full effectiveness of Armor Penetration and non-deflected hits?

    The damage modifier - as I use it - is calculated out of the expected value of a hit (Ref: https://jannenw.info/pages/mechanics16/formulas)

    (1 + power/100000) * resistanceIgnored * [1 + critChance * CritSev + CAUptime * CASev] * (1 - targetDeflectChance * 0.5)

    "resistance" and "targetDeflectCance" take into account what Armor Penetration and what Accuracy you have. If you are interested, take a look at Janne's page how the several parts are calculated - it's explained well there.
    Ligula Trickfoot
  • dupeksdupeks Member Posts: 1,789 Arc User

    dupeks said:

    Edit: Is the damage modifier assuming full effectiveness of Armor Penetration and non-deflected hits?

    The damage modifier - as I use it - is calculated out of the expected value of a hit (Ref: https://jannenw.info/pages/mechanics16/formulas)

    (1 + power/100000) * resistanceIgnored * [1 + critChance * CritSev + CAUptime * CASev] * (1 - targetDeflectChance * 0.5)

    "resistance" and "targetDeflectCance" take into account what Armor Penetration and what Accuracy you have. If you are interested, take a look at Janne's page how the several parts are calculated - it's explained well there.
    Got it, I misunderstood.

    The implication is then that the higher gear character could improve damage output in scaled content by reallocating stats back to ArPen and Accuracy. Which brings us back to the idea that it's optimal to have a separate set of gear for scaled content.
  • adinosiiadinosii Member, NW M9 Playtest Posts: 4,294 Arc User

    From this data I can clearly see that a BiS Level 80 toon does as well or slightly worse than a 14K Level 70.

    That sounds...about right, as far as regular gear is concerned. However, in practice the BiS 80 player will have access to more gear that either does not scale, or scales "better", so he will probably stay ahead.
    Hoping for improvements...
  • micky1p00micky1p00 Member, NW M9 Playtest Posts: 3,594 Arc User
    adinosii said:

    From this data I can clearly see that a BiS Level 80 toon does as well or slightly worse than a 14K Level 70.

    That sounds...about right, as far as regular gear is concerned. However, in practice the BiS 80 player will have access to more gear that either does not scale, or scales "better", so he will probably stay ahead.
    Not as long as those are simple caps. There are few things that are not scaled, but those won't bring much to the table overall.
  • micky1p00micky1p00 Member, NW M9 Playtest Posts: 3,594 Arc User
    edited May 2019
    There is an issue with how things are now, players are not supposed to hit the caps, that is what the counter stats are for, yet it doesn't work, so there is the capped now, not capped scaled issue, but anyway, In general if the powers that be will ever read, I've suggested this before:

    There are simple and working ways to do non caped, non linear scaling.

    Some key aspects we want from the system:

    1. To maintain the linear relationship between the stats, if you have twice power than ArP we want to keep that ratio.
    This will maintain correctness vs the counter stats without players reallocating stats

    2. We want scaling to be 'proportionally' correct, if Player A has more power before scaling over player B (pA > pB), then F(pA) > F(pB) where F is the result of the scaling. Or in short, a BiS level 80 player will not find themselves weaker than a new level 70. And progress preserved.

    3. Adjustable non linear function, something easy to replace or to tune if needed.

    4. Non-linear so power creep will not be an issue for the foreseeable future.

    5. Adjustable per zone, the 'scaling bar' should be set per zone


    So our overall system will work based on total IL, it will take a zone scaling 'border' for example 14k for some zone, bellow that number no scaling will occur, the scaling multiplier is 1.

    Here two examples, one more simple to understand, the other more elegant:

    If the player total IL is above that number we do the following:

    ScaledIL = ScalingBarrier + ScalingFunction(PlayerIL - ScalingBarrier)
    divider = PlayerIL / ScaledIL

    Where the scaling function is a non linear function derivative between 0 and 1. For example it can be x^(1/a) for a > 1, it can be
    even an asymptotic function which will create an eventual cap.

    In this example we will pick x^0.8

    So our 22k player, just became 15325
    with a divider of 1.435

    We divide each stat by our divider and we are done.


    Second similar example will be:

    Scaling Divider = ScalingFunction((PlayerIL / ScalingBarrier ))

    Here again we pick the same function, and for 0.8 the methods will almost align for these numbers so

    Divider is 1.435
    22k player becomes 15325k and we do the same per stat and divide each by the divider.


    (For other values, for example 0.9 the results will differ in those methods)

    https://www.desmos.com/calculator/xb3kw98h4e


    This general idea, perhaps not perfect, but meets the above requirement, and with a diminishing enough function will last for a long time before adjustment is needed.
    More so, there are hard caps on most stats anyway, so the scaling by caps is just not needed. And in any case, adjustment is easy, just change the 0.8 to 0.81 (for example) and viola or change the target barrier, and it's done.
  • photeus#7983 photeus Member Posts: 34 Arc User
    I suspect the scaling isn't so simple as numbers adjustment. I think it's the original scaling system (which was barely noticeable) layered on top of the recently developed AI system to scale gear/ enchants by ilevel.

    I also think that the coders can't modify some things without breaking the engine. That's why there is more focus on individual instances and creatures. They can't change anything on player end (other than order of which system scales first or turning off monster-spawns-at- player- level completely).

    People have been complaining about it making the players weak. But the devs concentrate on the monsters.
  • gripnir78gripnir78 Member, NW M9 Playtest Posts: 374 Arc User
    micky1p00 said:

    There is an issue with how things are now, players are not supposed to hit the caps, that is what the counter stats are for, yet it doesn't work, so there is the capped now, not capped scaled issue, but anyway, In general if the powers that be will ever read, I've suggested this before:

    There are simple and working ways to do non caped, non linear scaling.

    Some key aspects we want from the system:

    1. To maintain the linear relationship between the stats, if you have twice power than ArP we want to keep that ratio.
    This will maintain correctness vs the counter stats without players reallocating stats

    2. We want scaling to be 'proportionally' correct, if Player A has more power before scaling over player B (pA > pB), then F(pA) > F(pB) where F is the result of the scaling. Or in short, a BiS level 80 player will not find themselves weaker than a new level 70. And progress preserved.

    3. Adjustable non linear function, something easy to replace or to tune if needed.

    4. Non-linear so power creep will not be an issue for the foreseeable future.

    5. Adjustable per zone, the 'scaling bar' should be set per zone


    So our overall system will work based on total IL, it will take a zone scaling 'border' for example 14k for some zone, bellow that number no scaling will occur, the scaling multiplier is 1.

    Here two examples, one more simple to understand, the other more elegant:

    If the player total IL is above that number we do the following:

    ScaledIL = ScalingBarrier + ScalingFunction(PlayerIL - ScalingBarrier)
    divider = PlayerIL / ScaledIL

    Where the scaling function is a non linear function derivative between 0 and 1. For example it can be x^(1/a) for a > 1, it can be
    even an asymptotic function which will create an eventual cap.

    In this example we will pick x^0.8

    So our 22k player, just became 15325
    with a divider of 1.435

    We divide each stat by our divider and we are done.


    Second similar example will be:

    Scaling Divider = ScalingFunction((PlayerIL / ScalingBarrier ))

    Here again we pick the same function, and for 0.8 the methods will almost align for these numbers so

    Divider is 1.435
    22k player becomes 15325k and we do the same per stat and divide each by the divider.


    (For other values, for example 0.9 the results will differ in those methods)

    https://www.desmos.com/calculator/xb3kw98h4e


    This general idea, perhaps not perfect, but meets the above requirement, and with a diminishing enough function will last for a long time before adjustment is needed.
    More so, there are hard caps on most stats anyway, so the scaling by caps is just not needed. And in any case, adjustment is easy, just change the 0.8 to 0.81 (for example) and viola or change the target barrier, and it's done.

    This would be correct BUT its not for a very simple reason - the method you proposed would work on defensve stats (and still not that much depneding on gear bonuses) but It wont work t all for dps.

    First of all its weapon damage dependant - so you can adjust pure stats but that element alone makes huge difference - not just its not linear but rng based but also it doeas change a lot - pretty much with every mod- just look at difference with older weapon sets and new from ME - same stats - different main hand - huge difference, and on top of that you still have some bonuses/buffs fro gear (Kuno of the bear) - so in short system you offer wont work.
  • micky1p00micky1p00 Member, NW M9 Playtest Posts: 3,594 Arc User
    gripnir78 said:


    This would be correct BUT its not for a very simple reason - the method you proposed would work on defensve stats (and still not that much depneding on gear bonuses) but It wont work t all for dps.

    First of all its weapon damage dependant - so you can adjust pure stats but that element alone makes huge difference - not just its not linear but rng based but also it doeas change a lot - pretty much with every mod- just look at difference with older weapon sets and new from ME - same stats - different main hand - huge difference, and on top of that you still have some bonuses/buffs fro gear (Kuno of the bear) - so in short system you offer wont work.

    Sorry, I don't see the issue. Set bonuses can remain as they are, they are not scaled now, and won't be under the proposed change.

    Weapon damage is not random, it's a stat. The weapon damage minimum and maximum.
    The jump from old weapon damage to new around ~4k weapon damage was a m16 change to normalize the classes, but otherwise there should be a linear 'connection' between weapon IL and the weapon damage. If needed, the divider can be adjusted for weapon damage specifically without much issue - mostly by using the weapon IL as the 'pivot'.

  • elwing#6559 elwing Member Posts: 29 Arc User
    I wanted to do a follow-up of my post after the changes done in scaling. So I collected the data again with my two characters (well, my characters have a bit more item level now, but that shouldn't be a big problem for the general idea). I added the new data into the same spreadsheet: https://docs.google.com/spreadsheets/d/1ynPsCICOVIuDVqnCFJKr0tdc7iOzjLMdyeAZHC6xfPE/edit?usp=sharing

    My main points with the scaling (before the big patch) were the following:


    The suggestions I will derive out of the data are the following:
    1) The scaling for MSP isn't right at the moment.
    2) Skirmishes aren't scaled in a meaningful way. The scaling needs to be adapted.
    3) Looking a the implications of the current scaling, a re-structure of the random queues is necessary.
    4) Scaling of weapon damage has a massive impact on the overall damage.
    5) Scaling brings players closer together – closer than it should be.

    If you compare, for example, the Damage Modifier and the Expected Value of a Hit, point 1) and 2) definitely improved!

    Comparison of the Damage Modifier:


    Comparison of the Expected Value of a Hit:


    MSP is in line with the other dungeons, even has a less harsh scaling.
    Also the Skirmishes are scaled the same way. There is a meaningful relation between the item level you are scaled to and the stats.

    For point 3) I can see a reason now to pot CR in the Advanced Queue, not Expert. Still I don't think this is a good choice with regard to player experience, but it's not totally unreasonable.

    4)Players are still quite close together, even with much different Item Level (and sense in choice of enchantments...). But the ridiculous situation in skirmishes is solved, so overall it's quite ok.

    To summarize, scaling is in a much better state now. Of course there is still room to optimize, but I appreciate your approaches and I hope my data collection helped to improve the scaling.
    Ligula Trickfoot
Sign In or Register to comment.