Dev Letter: Balance Framework
Hello everyone,
Today we’re going to touch on what might be the most sensitive topic of all: Balance
[h2]Do the devs even play? [/h2]
One of the more common questions we receive is on the subject of whether or not we actually play Eternal Return.
The easy answer: yes. We understand the importance of in-game experience, and we play a lot. Taking a stroll through our office, you’ll see multiple people in-game at any given time.
Like those of you who have put in countless hours in Season 1, the team here has the same passion and competitive drive. The time and experience from this allows us to provide improvements that are only possible with a deep understanding of the game.
Our balance team consists of passionate players with pro gaming experience in other MOBA games. One of them even grinded for like 2,500 games.
The balance team lead DEMANDS that every member of the team be above a certain tier or put in a certain amount of games, but members of the balance team come up with excuses like “having a family” or “sleeping”.
We also have 2 in Titan and a couple Platinum and Diamond players. It’s not a lot, but looking at it as a whole can spark some concern in the balance team.
When going over a patch, you might wonder if we love and play the game, why are things adjusted as so or why was your main nerfed. Our team is simply too small to perfectly balance a game that has never been created before. Admittedly, we lack in some places.
We went through a lot of hardships through season 1. We’ve learned a ton and grew just as much. We already have tons of ideas and ways to improve for season 2.
[h2]On the ‘opinion’ of balance[/h2]
Looking at some of the stats from our pre-season, we can see that Hart is usually killed by Yuki while Lenox most likely dies from Rozzi.
The place with the most deaths is Chapel, but Shoichi mostly dies in Forest. Jackies usually die in Uptown or Dock.
Data also changes based on Ranked Tier. In Gold or lower, Chiara dies mostly to Luke. In Diamond or higher, however, the percentage of Chiara dying to Yuki or Fiora was over 5x higher. Even if we put pick rate into consideration, the difference still remains.
Due to the differences of routes, the ‘opinion’ of certain characters you meet in the game drastically changes.
Usually, characters with weak early game become stronger as the game goes on. Players have the chance to stop these characters early into their routes, but characters who do not share certain routes would believe that those characters are “too strong” - considering they only meet them once they’ve hit their power spike.
The routes a character takes determines which characters they meet and when they meet them. Using the same routes as you climb the ranks will cause you to meet different characters in different locations based on your tier.
Simply put, Eternal Return games result in a difference in opinions -- and data.
When we follow debates in the community or discussions between high tier players, there are always different outlooks and thoughts.
The team even considered adopting a “No feedback on mains” policy in balance discussions. This is especially a problem when a team member’s opinions are different from the data!
Balance is always a hard problem and there is much to fix. No competitive multiplayer games will be perfectly balanced. Discussions about various characters also tend to spiral due to the ‘Difference in opinion.'
Because of this, we want to talk about what we think are structural problems based on the feedback we got from season 1.
[h2]Season 1 Balance Patches[/h2]
During Season 1, we used our official homepage to post statistics weekly and usually based balance patches on that data.
We held a bi-weekly minor and major patch. The major patch featured structural changes, while minor patches featured adjustments based on the data.
Although it is important to look at statistics from different perspectives, these are the stats that we valued most over others:
[h2]Problems and improvement with the compilation and disclosure of statistics[/h2]
We heard a lot that our patches and statistics were sometimes conflicting.
A part of it was caused by the “Difference in statistics and opinions,' but we thought there were other problems that needed solving.
First, some problems seemed to be caused by differences in data views. Our patch drafts and balance analysis usually start around Sunday. We use the data from the previous 3 days for the patch. Statistics released on Monday include Sunday. Overall statistics rarely change significantly, but high-elo statistics often shift a bit.
For higher-elo statistics that are more meta-sensitive, the short window itself lacks value. When there is a change, we need time to study it and plan how to respond. The short cycle of presenting data led to us being impacted by pre-existing data.
In Season 2, statistics will release bi-weekly on Monday before major updates. This will nearly triple the number of statistical days, giving us more reliable and precise statistics. We can improve our consistency with additional time!
In the week without posting statistics, we'd like to release stats from new perspectives, i.e regional statistics, character statistics, starting area comparisons, team mode comparisons.
Ex: Top 5 Best/Worst Spawn Points (All Tiers)
(Factory, archery, temple, uptown, avenue)
(School, beach, chapel, pond, forest)
[h2]Patching with statistics[/h2]
In MOBAs, you typically examine win rate and pick rate. However in ER:BS, which is a Battle Royale at its core, kill average and placement average also massively impact gameplay and MMR in big ways.
Due to the amount of data points, we should observe stats that are slightly over or below average with more caution, rather than coming up with a quick solution. In season 1, we believe that we were too focused on the equality of results.
Based on this, we would like to implement the new ‘Eternal Return Balance Patch Framework” - allowing us to track characters that need nerfs or buffs based off of statistical outliers.
We want to establish some statistical standards, and balance characters that meet those statistics.
We’ll evaluate some additional statistics for the ‘Eternal Return Balance Patch Framework’ when Season 2 starts.
If it's not at a level that “breaks” the balance framework, we’ll try to reflect the play experience and characteristics rather than just balancing to equalize statistics.
[h2]Patch Cycle[/h2]
The short patch cycle per week had the advantage of quickly fixing balance problems, but we also received a lot of feedback on how to improve.
For minor patches, we will now avoid large-scale adjustments and only adjust for hotfix level balances that far exceed the 'balance patch framework.
[h2]Patch Note Communication[/h2]
We left a lot to be desired in the patch notes as well. We omitted things for small changes or didn’t give totally adequate descriptions. It’s tough to get patch notes translated into multiple languages and have the context make sense for each language.
However, going forward, we will work on sharing more details in the patch and take your thoughts based on that feedback to prepare for the next patch. We will try our best to address both individual aspects and the direction of larger categories.
There were also cases where we noticed a problem early but lacked statistical validation, or needed to wait a little longer because we noticed the meta shifting or something needed development time to improve functionality.
By increasing the time to communicate more about these areas, we will add "Special Targets of Attention" in patches that are focused on the current trend, and a "Preview of Next Patch" for elements that are scheduled for improvement.
Also, we might add some pictures for easier readability!
Releasing a ‘Patch Note Summary’ to make it easier to understand what characters and weapons have changed in this patch. (The following is a layout sample, and has nothing to do with actual patch)
[h2]Problems with balancing content[/h2]
Most feedback is about the content of the balance patch. There were issues with characters, but there were also more structural problems; the most mentioned parts were combat and the final zone.
Characters with good mobility showed great strength in the final restricted area resulting in a good win rate but average kills, which then led to these characters receiving adjustments rather than needed buffs.
We will continue to balance combat and work on the final restricted area system so that any character can reliably take home the W.
Some players expressed feedback on experiencing a sense of deprivation due to nerfs based on statistics from tiers above or below them.
This is a problem that is difficult to solve. There’s limitations to balance any game such as never having useless characters in both upper and lower elo, characters being too powerful in a particular tier, or even characters with wild power spikes.
Characters that are only powerful in specific elos are the most difficult cases. These require structural improvements and are difficult to solve in a short period of time, but still must be adjusted to circulate the meta and stop a character from dominating.
In addition to the overall ability to engage, there are times when matchups between characters become too unfair for players or players felt that a character was just a “stat stick”.
Some characters with play mechanics are too simple to be solved by simply adjusting numbers instead of focusing on their fundamental problems.
Some of the problems with this combat structure were addressed including some improvements to the formula of damage concerning skill amplification/ENAD/critical hits. However, the engagement structure of Eternal Return should be more complex and robust.
The update at the end of April is planned to change the capability of some low-utilization traps so that traps can be a factor in skirmishes, and introduce an EMP drone that can temporarily disable traps as well.
Adding item skills is an important goal to enrich combat so that we can have fun overcoming matchups through skill and outplays. At the same time, strengthening legendary items and mid-game elements will change the pace of combat and increase objectives to fight for.
We're going to continue to improve the functionality of characters with kits that are too simple in major updates. We're also going to add little bits of techniques for skill expression!
Example of EMP Drone:
[h2]Balancing new characters[/h2]
In the case of new characters, there was often a disparity between feedback and statistics.
One of the key problems was matchmaking in normals after character release. Top players who usually played ranked hopped into normals. This was improved in the last update by segmentation in normals.
In addition to the matchmaking problem, you all found ways to make new characters oppressive on release. We will set a stricter standard for the balance test process related to new characters in the future. We will also establish a hotfix standard for new characters in the Balance Patch Framework so that there can be hotfixes as soon as necessary.
Our team extensively tests new characters, but in just like 4 hours, the collective games you all play on a new character eclipses our total time.
[h2]Team Mode Balance Improvements[/h2]
Instead of simply adjusting the damage per mode according to performance, we may also adjust the required stats according to the character's role in team modes.
Through the addition of item skills and structural improvement, various characters can be adjusted with more care to showcase various forms of combat.
We often felt that separate explanations were needed for team modes in patch notes, 'Team mode balancing' segments are newly created in patch notes to explain team mode balancing issues, and reasons for adjustments.
[h2]Addressing learning curves and "Structural early-game" issues[/h2]
We also saw a lot of feedback for early game issues.
Gameplay in the early-game consists of a fun game of “who makes purple weapons first”. This is linked to the starting area and a bit of luck. This also sucks for new players when they get in matches with smurfs and lose early against someone that built an Agni in like 11 seconds.
[h2]Adding Survival Elements Early in the Game[/h2]
Rather than eliminating RNG itself, it is more important to address elements that transform RNG into unpleasant experiences. We will make spawn points as fair as possible and improve it to enable players that fall off early to get back into the game by using intuition and skill.
[h2]Improving Learning Curve[/h2]
Regarding looting, we plan to solve factors that can be a barrier due to inconvenience of the game, while preserving the difference in performance caused by individual skills. (May be subject to change.)
Shaded targeted items.
Added additional emphasis when acquiring all targeted items in the area
Emphasized target item combinations when possible
ALT to identify unidentified boxes and select overlapping items
Displaying targeted item tree
Adding an overview of targeted items that you haven't gotten yet.
Adjusting locations and numbers of boxes in Archery Range, Alley, Hotel, Avenue (Scheduled to expand to all areas in the future)
Display teammate routes in team modes
Aligning target items in the box to the left, and mark items that are only lootable in this area during the looting process (Scheduled for first major update in May)
[h2]Ending Remarks[/h2]
In order to relieve any frustration, we wanted to think hard about our future direction and demonstrate how it is getting better, rather than just doing random stuff that we’re not even sure about.
We're working hard to listen to all the feedback you give us on patches throughout all our communities. With such a small team, we rely on all your feedback to help us move in the right direction.
Although players may not be perfectly satisfied with each patch, the goal of the Season 2 balance patch framework is to make patches and adjustments that you believe are needed.
We’re still growing and admittedly we still suck in a lot of places, but we’re working our asses off to get better and be the game you deserve.
Thank you for always sticking by with us.
[h2]Summary[/h2]
[h3]Season 2 Changes[/h3]
Today we’re going to touch on what might be the most sensitive topic of all: Balance
[h2]Do the devs even play? [/h2]
One of the more common questions we receive is on the subject of whether or not we actually play Eternal Return.
The easy answer: yes. We understand the importance of in-game experience, and we play a lot. Taking a stroll through our office, you’ll see multiple people in-game at any given time.
Like those of you who have put in countless hours in Season 1, the team here has the same passion and competitive drive. The time and experience from this allows us to provide improvements that are only possible with a deep understanding of the game.
Our balance team consists of passionate players with pro gaming experience in other MOBA games. One of them even grinded for like 2,500 games.
The balance team lead DEMANDS that every member of the team be above a certain tier or put in a certain amount of games, but members of the balance team come up with excuses like “having a family” or “sleeping”.
We also have 2 in Titan and a couple Platinum and Diamond players. It’s not a lot, but looking at it as a whole can spark some concern in the balance team.
When going over a patch, you might wonder if we love and play the game, why are things adjusted as so or why was your main nerfed. Our team is simply too small to perfectly balance a game that has never been created before. Admittedly, we lack in some places.
We went through a lot of hardships through season 1. We’ve learned a ton and grew just as much. We already have tons of ideas and ways to improve for season 2.
[h2]On the ‘opinion’ of balance[/h2]
Looking at some of the stats from our pre-season, we can see that Hart is usually killed by Yuki while Lenox most likely dies from Rozzi.
The place with the most deaths is Chapel, but Shoichi mostly dies in Forest. Jackies usually die in Uptown or Dock.
Data also changes based on Ranked Tier. In Gold or lower, Chiara dies mostly to Luke. In Diamond or higher, however, the percentage of Chiara dying to Yuki or Fiora was over 5x higher. Even if we put pick rate into consideration, the difference still remains.
Due to the differences of routes, the ‘opinion’ of certain characters you meet in the game drastically changes.
Usually, characters with weak early game become stronger as the game goes on. Players have the chance to stop these characters early into their routes, but characters who do not share certain routes would believe that those characters are “too strong” - considering they only meet them once they’ve hit their power spike.
The routes a character takes determines which characters they meet and when they meet them. Using the same routes as you climb the ranks will cause you to meet different characters in different locations based on your tier.
Simply put, Eternal Return games result in a difference in opinions -- and data.
When we follow debates in the community or discussions between high tier players, there are always different outlooks and thoughts.
The team even considered adopting a “No feedback on mains” policy in balance discussions. This is especially a problem when a team member’s opinions are different from the data!
Balance is always a hard problem and there is much to fix. No competitive multiplayer games will be perfectly balanced. Discussions about various characters also tend to spiral due to the ‘Difference in opinion.'
Because of this, we want to talk about what we think are structural problems based on the feedback we got from season 1.
[h2]Season 1 Balance Patches[/h2]
During Season 1, we used our official homepage to post statistics weekly and usually based balance patches on that data.
We held a bi-weekly minor and major patch. The major patch featured structural changes, while minor patches featured adjustments based on the data.
Although it is important to look at statistics from different perspectives, these are the stats that we valued most over others:
- Contesting overpowered early game builds
- Priority 1 - Kill average from all elo, Kill frequency by time
- Priority 2 - MMR gain average from mid to low elo
- Priority 1 - Kill average from all elo, Kill frequency by time
- Overall game balance changes
- Priority 1 - MMR gain average from high elo
- Priority 2 - Kill average from high elo, ranking average, MMR gain average from mid to low elo
- Priority 1 - MMR gain average from high elo
[h2]Problems and improvement with the compilation and disclosure of statistics[/h2]
We heard a lot that our patches and statistics were sometimes conflicting.
A part of it was caused by the “Difference in statistics and opinions,' but we thought there were other problems that needed solving.
First, some problems seemed to be caused by differences in data views. Our patch drafts and balance analysis usually start around Sunday. We use the data from the previous 3 days for the patch. Statistics released on Monday include Sunday. Overall statistics rarely change significantly, but high-elo statistics often shift a bit.
For higher-elo statistics that are more meta-sensitive, the short window itself lacks value. When there is a change, we need time to study it and plan how to respond. The short cycle of presenting data led to us being impacted by pre-existing data.
In Season 2, statistics will release bi-weekly on Monday before major updates. This will nearly triple the number of statistical days, giving us more reliable and precise statistics. We can improve our consistency with additional time!
In the week without posting statistics, we'd like to release stats from new perspectives, i.e regional statistics, character statistics, starting area comparisons, team mode comparisons.
Ex: Top 5 Best/Worst Spawn Points (All Tiers)
(Factory, archery, temple, uptown, avenue)
(School, beach, chapel, pond, forest)
[h2]Patching with statistics[/h2]
In MOBAs, you typically examine win rate and pick rate. However in ER:BS, which is a Battle Royale at its core, kill average and placement average also massively impact gameplay and MMR in big ways.
Due to the amount of data points, we should observe stats that are slightly over or below average with more caution, rather than coming up with a quick solution. In season 1, we believe that we were too focused on the equality of results.
Based on this, we would like to implement the new ‘Eternal Return Balance Patch Framework” - allowing us to track characters that need nerfs or buffs based off of statistical outliers.
We want to establish some statistical standards, and balance characters that meet those statistics.
- Characters with high win rate and an above average pick rate
- Characters with overly high pick rates in high elo
- Characters with below average MMR gain rate
- Characters with high average kills in mid to low elo
- Hotfixes for new characters who are above a certain threshold of average kills and placement for 1~3 days after release
- Other elements that need to be statistically changed
We’ll evaluate some additional statistics for the ‘Eternal Return Balance Patch Framework’ when Season 2 starts.
If it's not at a level that “breaks” the balance framework, we’ll try to reflect the play experience and characteristics rather than just balancing to equalize statistics.
- We're going to work hard to address aspects that feel oppressive or too easy and improve structural problems.
- Pre-season data improvements are underway. More in-depth statistics about character matchups, areas, routes, etc. will be reflected to supplement basic statistics.
[h2]Patch Cycle[/h2]
The short patch cycle per week had the advantage of quickly fixing balance problems, but we also received a lot of feedback on how to improve.
- Due to the short patch cycle, players can’t even adapt, causing the game to flow in a different direction from what the patch intended. Opinions on character mechanics often changed over two to three weeks, but by that time it had already been patched.
- The patch cycle was shorter than the normal development period, so the timing of the problem and the improvement of the function did not always match,(e.g., Li Dailin’s silence/slow change, Silvia’s (R), Hart’s E targeting change, etc.)
- We read feedback that minor patches had too much content and large adjustments, so the game wasn’t stable or we just rolled the dice and patch notes were random.
For minor patches, we will now avoid large-scale adjustments and only adjust for hotfix level balances that far exceed the 'balance patch framework.
- We don't want to cause a weekly shift in meta so that the carefully written player guides don't go out of date or the value of the routes that players worked hard on drops.
- The goal is not to allow balance patches to determine the meta of the game, but to allow players to research and form the meta. A lead from behind approach.
- We will put more content and effort into major patches so that we can focus on fundamental problems a bit more and make a perfect game.
- However, if something GAMEBREAKING pops up, we will not hesitate to patch it quickly.
[h2]Patch Note Communication[/h2]
We left a lot to be desired in the patch notes as well. We omitted things for small changes or didn’t give totally adequate descriptions. It’s tough to get patch notes translated into multiple languages and have the context make sense for each language.
However, going forward, we will work on sharing more details in the patch and take your thoughts based on that feedback to prepare for the next patch. We will try our best to address both individual aspects and the direction of larger categories.
There were also cases where we noticed a problem early but lacked statistical validation, or needed to wait a little longer because we noticed the meta shifting or something needed development time to improve functionality.
By increasing the time to communicate more about these areas, we will add "Special Targets of Attention" in patches that are focused on the current trend, and a "Preview of Next Patch" for elements that are scheduled for improvement.
Also, we might add some pictures for easier readability!
Releasing a ‘Patch Note Summary’ to make it easier to understand what characters and weapons have changed in this patch. (The following is a layout sample, and has nothing to do with actual patch)
[h2]Problems with balancing content[/h2]
Most feedback is about the content of the balance patch. There were issues with characters, but there were also more structural problems; the most mentioned parts were combat and the final zone.
Characters with good mobility showed great strength in the final restricted area resulting in a good win rate but average kills, which then led to these characters receiving adjustments rather than needed buffs.
We will continue to balance combat and work on the final restricted area system so that any character can reliably take home the W.
Some players expressed feedback on experiencing a sense of deprivation due to nerfs based on statistics from tiers above or below them.
This is a problem that is difficult to solve. There’s limitations to balance any game such as never having useless characters in both upper and lower elo, characters being too powerful in a particular tier, or even characters with wild power spikes.
Characters that are only powerful in specific elos are the most difficult cases. These require structural improvements and are difficult to solve in a short period of time, but still must be adjusted to circulate the meta and stop a character from dominating.
In addition to the overall ability to engage, there are times when matchups between characters become too unfair for players or players felt that a character was just a “stat stick”.
Some characters with play mechanics are too simple to be solved by simply adjusting numbers instead of focusing on their fundamental problems.
Some of the problems with this combat structure were addressed including some improvements to the formula of damage concerning skill amplification/ENAD/critical hits. However, the engagement structure of Eternal Return should be more complex and robust.
The update at the end of April is planned to change the capability of some low-utilization traps so that traps can be a factor in skirmishes, and introduce an EMP drone that can temporarily disable traps as well.
Adding item skills is an important goal to enrich combat so that we can have fun overcoming matchups through skill and outplays. At the same time, strengthening legendary items and mid-game elements will change the pace of combat and increase objectives to fight for.
We're going to continue to improve the functionality of characters with kits that are too simple in major updates. We're also going to add little bits of techniques for skill expression!
Example of EMP Drone:
[h2]Balancing new characters[/h2]
In the case of new characters, there was often a disparity between feedback and statistics.
One of the key problems was matchmaking in normals after character release. Top players who usually played ranked hopped into normals. This was improved in the last update by segmentation in normals.
In addition to the matchmaking problem, you all found ways to make new characters oppressive on release. We will set a stricter standard for the balance test process related to new characters in the future. We will also establish a hotfix standard for new characters in the Balance Patch Framework so that there can be hotfixes as soon as necessary.
Our team extensively tests new characters, but in just like 4 hours, the collective games you all play on a new character eclipses our total time.
[h2]Team Mode Balance Improvements[/h2]
Instead of simply adjusting the damage per mode according to performance, we may also adjust the required stats according to the character's role in team modes.
Through the addition of item skills and structural improvement, various characters can be adjusted with more care to showcase various forms of combat.
We often felt that separate explanations were needed for team modes in patch notes, 'Team mode balancing' segments are newly created in patch notes to explain team mode balancing issues, and reasons for adjustments.
[h2]Addressing learning curves and "Structural early-game" issues[/h2]
We also saw a lot of feedback for early game issues.
Gameplay in the early-game consists of a fun game of “who makes purple weapons first”. This is linked to the starting area and a bit of luck. This also sucks for new players when they get in matches with smurfs and lose early against someone that built an Agni in like 11 seconds.
[h2]Adding Survival Elements Early in the Game[/h2]
Rather than eliminating RNG itself, it is more important to address elements that transform RNG into unpleasant experiences. We will make spawn points as fair as possible and improve it to enable players that fall off early to get back into the game by using intuition and skill.
- Improved imbalance between spawn points
- Buffing White, Green food
- Increased starting HP, Reduced HP per level
- Improved experience of looting
- New Map Objects Available for Survival
- A special ping when someone nearby builds a Purple or Yellow weapon.
[h2]Improving Learning Curve[/h2]
Regarding looting, we plan to solve factors that can be a barrier due to inconvenience of the game, while preserving the difference in performance caused by individual skills. (May be subject to change.)
- Looting enhancements already reflected in the game
Shaded targeted items.
Added additional emphasis when acquiring all targeted items in the area
Emphasized target item combinations when possible
ALT to identify unidentified boxes and select overlapping items- Looting Enhancements and Starting Area Improvements scheduled for end-April updates
Displaying targeted item tree
Adding an overview of targeted items that you haven't gotten yet.
Adjusting locations and numbers of boxes in Archery Range, Alley, Hotel, Avenue (Scheduled to expand to all areas in the future)
Display teammate routes in team modes
Aligning target items in the box to the left, and mark items that are only lootable in this area during the looting process (Scheduled for first major update in May)[h2]Ending Remarks[/h2]
In order to relieve any frustration, we wanted to think hard about our future direction and demonstrate how it is getting better, rather than just doing random stuff that we’re not even sure about.
We're working hard to listen to all the feedback you give us on patches throughout all our communities. With such a small team, we rely on all your feedback to help us move in the right direction.
Although players may not be perfectly satisfied with each patch, the goal of the Season 2 balance patch framework is to make patches and adjustments that you believe are needed.
We’re still growing and admittedly we still suck in a lot of places, but we’re working our asses off to get better and be the game you deserve.
Thank you for always sticking by with us.
[h2]Summary[/h2]
- The devs do actually play!
- Eternal return has a "Difference of Opinion" in which balance is completely different between players depending on your character/route/tier.
- Season 1 balance was “acceptable” but could have been SO MUCH better. We suck, but we’re trying to get better.
[h3]Season 2 Changes[/h3]
- Stabilizing statistics by releasing stats bi-weekly and getting more granular.
- Simplify minor patches and avoid weekly meta changes due to balance patches
- Introduce "Eternal Return Balance Patch Framework" and add context on changes that might go against our system
- More comments and context in patch notes
- Add 'Special Targets of Attention' / 'Next Patch Preview' / 'Non-Regular Disclosure of Statistics'
- Add Item skills and upgrade characters to create a more complex combat system and promote more skill expression
- Improved new character and team mode balancing process
- Mitigate learning curve and improve early-game balance