How do they Balance Characters?

Artmastercorey

Smash Journeyman
Joined
Oct 31, 2005
Messages
394
#1
Anyone have any insight on how Sakurai and team test and balance the characters. Or how its done in fighting games in general? I always wondered because some characters Im thinking theres no way they cant see this character needs adjusting, then there are times a character has hidden potential I didnt see initially . I often wonder how do they acheive that balance.

There are so many characters in this game theres no way they can balance it all by themselves, so I wonder if they just recruit top players who play those characters and get their insight? Just thinking out loud.
 

Spiny Top

Smash Cadet
Joined
Jan 5, 2019
Messages
47
#2
I believe they use elite smash data, in particular win rates. If the character has win rate of less than 50% in elite smash they are eligible for a buff. More than 50, a nerf or whatever.

They can then monitor the new win rate percentage and see if it goes up or down.
 

Artmastercorey

Smash Journeyman
Joined
Oct 31, 2005
Messages
394
#3
Yeah I do beleive online data does play some roll in it. I wonder if before the game is even released, before it goes online if they just make bots fight each other or how they determine "ok this character is good enough".
 

Balance ssbu

Smash Rookie
Joined
Jul 3, 2020
Messages
14
#4
They’ve done an all right job but one ill be honest one of their methods to balancing is to give a character some cheap or cheese gimmick to save it from being stomped. Which ultimately unbalances the game in so many other ways.
 

Mario & Sonic Guy

Old rivalries live on!
Joined
Sep 18, 2007
Messages
22,122
Location
Mushroom Kingdom
NNID
TPitch5
3DS FC
5327-1637-5096
#5
The main problem with using online play as a source for making character balance changes, is that less viable characters can end up getting nerfed, even when such characters don't need nerfs to begin with. Just look back at Smash 4, and how the balance changes gradually made Dedede worse with each update.
 

StrangeKitten

Smash Journeyman
Joined
Mar 25, 2020
Messages
334
Location
Pokemon Stadium 2
#6
It seems like they're taking competitive into account in at least some measure. Dedede recently got buffs, and the only characters that have seen nerfs have been top tiers aside from K Rool and Dedede's Gordos, both of which happened early on before we had time to realize that these two were bad characters
 

Artmastercorey

Smash Journeyman
Joined
Oct 31, 2005
Messages
394
#7
The main problem with using online play as a source for making character balance changes, is that less viable characters can end up getting nerfed, even when such characters don't need nerfs to begin with. Just look back at Smash 4, and how the balance changes gradually made Dedede worse with each update.
Yeah, even before the game was introduced online when it was just the smash team balancing it, I wonder how they did it?

Like how do they balance speed for example? Like could it be every character has a certain amount of frames they can use, and you disperse them throughout their moveset in different ways until all the frames are spent. Like how do you know your character is too slow or not. How do you know your character's range versus speed, versus power, is evenly distributed is what Im curious about?

Like they felt a top tier (like pikachu, joker, palatina, ect), range, speed, and power were balanced , but at the very bottom of the tier list (someone like a kirby/ddd, ect) they also felt was balanced and had the speed, power, and range needed to compete? It seems like such a big misight and it makes me curious as to how they couldnt see it, or what methods they are using to test the balancing.

Perhaps it initally was just the overwhelming number of characters so they figured it can be patched later but Im not so sure if the character patches have totally addressed the low tier balancing in significant enough ways. Anyway back to my main point, Im just curious as to the method of how they initially balanced the characters in this game, or fighting games in general for that matter.
 
Last edited:

GhostM

Phantom
Premium
Joined
May 3, 2019
Messages
408
Location
Off the Radar
#8
Usually when they balance a character, they are tame when it comes to buffing or nerfing a character since they don't want to make them too powerful or too weak. When it comes to buffs, they usually implement tame buffs that strengthen a character's already good moves/options to make them better, but not outright broken. Same can go for nerfs, since they usually nerf the good options/good moves, but just tame enough so that they don't outright nerf a character to the point that they aren't usable anymore.

Easiest example for tame buffs is Captain Falcon from the latest patch. Raptor Boost was buffed earlier to have super armor and buffed to have combo potential, but the latest patch (v8.0.0) made the launch angle much easier to follow up with confirmed combos like up airs, and knees, and made the overall range much further when used grounded. They also made the meteor hitbox easier to hit in the air too.

Easiest example for tame nerfs was the nerf to Joker's Arsene in patch v7.0.0. Arsene was already a great tool since it greatly strengthen's Joker's attacks, but they reduced the amount of time Arsene is summoned for when receiving damage when Arsene was summoned in a 1v1 battle. They nerfed Joker's main good option, but not to the point where Arsene was useless.

The only character that has been nerfed slightly hard was Pichu, but not outright nerfed to the ground.
 

Mario & Sonic Guy

Old rivalries live on!
Joined
Sep 18, 2007
Messages
22,122
Location
Mushroom Kingdom
NNID
TPitch5
3DS FC
5327-1637-5096
#9
Yeah, even before the game was introduced online when it was just the smash team balancing it, I wonder how they did it?

Like how do they balance speed for example? Like could it be every character has a certain amount of frames they can use, and you disperse them throughout their moveset in different ways until all the frames are spent. Like how do you know your character is too slow or not. How do you know your character's range versus speed, versus power, is evenly distributed is what Im curious about?

Like they felt a top tier (like pikachu, joker, palatina, ect), range, speed, and power were balanced , but at the very bottom of the tier list (someone like a kirby/ddd, ect) they also felt was balanced and had the speed, power, and range needed to compete? It seems like such a big misight and it makes me curious as to how they couldnt see it, or what methods they are using to test the balancing.

Perhaps it initally was just the overwhelming number of characters so they figured it can be patched later but Im not so sure if the character patches have totally addressed the low tier balancing in significant enough ways. Anyway back to my main point, Im just curious as to the method of how they initially balanced the characters in this game, or fighting games in general for that matter.
Another thing that's worth noting is that the software updates that we've received have not touched any of the fighter mobility attributes. Who thought that it was a good idea to make Incineroar so slow that more than half of the roster can walk faster than its running speed?!

Anyway, when you look back at Smash 4, it at least had the decency to improve Kirby, Lucario, and Mewtwo in the mobility department. But for a price, Corrin ended up being made slower, which felt unnecessary when you consider that only his Counter Surge warranted nerfs.
 

Garo

Smash Journeyman
Joined
Apr 26, 2014
Messages
233
NNID
Garomasta
3DS FC
3308-4572-3157
#11
These are from a 2018 DenFamiNico interview with Sakurai, translated from Japanese to English:

Speaking on how Smash Bros. keeps improving and adding more content, we can imagine that most of the time was probably spent adjusting character balance.
Masahiro Sakurai, director: “Well, it’s not just that. Actually, for Ultimate the system was changed so that I don’t directly work with the character parameters myself.”

Wait, really?! We actually really wanted to know whether you were handling all the new characters’ abilities and balancing yourself or not.
Sakurai: “Well, it was abnormal up until now. (laughs) In fighting games, multiple project leads split the balancing work by characters between themselves. But if you do that, you can have problems where “only the characters balanced by this person are strong”, and the opposite can happen to “weak” characters. In the first place, for Smash Bros., there was nobody who could do those adjustments, so I’d end up working on the parameters myself.”

“From Brawl and onwards, I made a monitoring team to collect battle results, and used that data and proposed options to adjust character data. For Brawl, the team was made of around 4 people, and they’d basically finish at 4-player battles, so there would be a bias towards that. There was also a team like that for Smash Bros. for 3DS/Wii U, but it seems they weren’t experienced enough.”

“For Ultimate, we greatly increased the scale of this system for both the planning and monitoring sides. We also added a certain amount of people who were good at battling… No, even before, there were good people, but this time, it was done in a more organized fashion. Of course, I would look for any points to change every day, and make comments if there were problems.”


Still, 74 + 1 + 5 fighters…
Sakurai: “Although I didn’t balance the parameters myself, the staff would suggest things like, “I want to make it this way,” and I’d end up watching over the entire process. That said, even doing this, I still think that “a game is no fun if it’s fair”.

“To make things fair, you have to standardize many things. And when you do, every character ends up feeling the same… Although in the first place, I decide when character attack frames begin at the very beginning.”


Can you detail how you pinpoint something as precise as that?
Sakurai: “I take pictures and other stuff of the motions I want, then decide, “Using this pose and composition, the attack begins in this frame and ends in this frame.” I add these to the project plan at the very beginning. If it’s too strong or too weak, the startup frames are changed bit by bit… So while to a certain extent the balancing is left to others, the work before that regarding the image of the attack is my work.”

Source link: https://www.siliconera.com/sakurai-...alancing-at-all-in-super-smash-bros-ultimate/
 
Joined
Nov 27, 2014
Messages
8,319
Location
New Orleans
NNID
BGenius23
3DS FC
0662-2900-1492
#13
In fairness, Little Mac's playstyle can never truly be balanced, as he's meant to have an abysmal air game to balance out his strong ground game.
I understand that Mac is an extremely polarizing character and that the wrong change could break him, but I don’t think flat out ignoring him is the answer. It’s not just him, either. There are many characters that need way more attention than they’re being given. Doc, for example. They gave him a nice kill confirm, but when has Doc ever struggled to take stocks? That’s the one thing he does exceedingly well.
 
Joined
Apr 16, 2015
Messages
1,026
Location
Somewhere in this Universe
NNID
MrGameguycolor
Switch FC
7681-9716-5789
#14
I understand that Mac is an extremely polarizing character and that the wrong change could break him, but I don’t think flat out ignoring him is the answer. It’s not just him, either. There are many characters that need way more attention than they’re being given. Doc, for example. They gave him a nice kill confirm, but when has Doc ever struggled to take stocks? That’s the one thing he does exceedingly well.
In short:
 

Mario & Sonic Guy

Old rivalries live on!
Joined
Sep 18, 2007
Messages
22,122
Location
Mushroom Kingdom
NNID
TPitch5
3DS FC
5327-1637-5096
#15
I understand that Mac is an extremely polarizing character and that the wrong change could break him, but I don’t think flat out ignoring him is the answer. It’s not just him, either. There are many characters that need way more attention than they’re being given. Doc, for example. They gave him a nice kill confirm, but when has Doc ever struggled to take stocks? That’s the one thing he does exceedingly well.
What Dr. Mario really needs is improvements to his overall mobility. This is especially notable for his weak air speed, since that's one of the reasons why his overall recovery is poor.
 

GabeLincoln

Smash Rookie
Joined
Oct 26, 2016
Messages
3
#16
The main problem with using online play as a source for making character balance changes, is that less viable characters can end up getting nerfed, even when such characters don't need nerfs to begin with. Just look back at Smash 4, and how the balance changes gradually made Dedede worse with each update.
Well, the real problem is that the netcode is garbage, and with it being garbage. If the netcode was good, you'd have a world where the game behaves mostly as it does offline, and so, characters who benefit from things generally being harder to punish due to said netcode such as Dedede and Ganon would have significantly worse winrates. Although, nothing would help the unfamiliarity problem where bad characters win in a casual setting like online because of matchup knowledge.
 
Joined
Nov 27, 2014
Messages
8,319
Location
New Orleans
NNID
BGenius23
3DS FC
0662-2900-1492
#17
What Dr. Mario really needs is improvements to his overall mobility. This is especially notable for his weak air speed, since that's one of the reasons why his overall recovery is poor.
Agreed. Even the slightest mobility buffs would help him out a great deal.

Doc is also plagued with the particularly infuriating trait of a lot of his moves being unsafe on hit at low percents. The same moves that could potentially start combos for Mario are actually combo starters for the opponent in Doc’s case.
 

Thinkaman

Moderator
Moderator
Joined
Aug 26, 2007
Messages
6,230
Location
Madison, WI
NNID
Thinkaman
3DS FC
1504-5749-3616
#18
There's a lot of underbaked takes on this thread.

To answer the op's question:
  • Smash 64 and Melee were balanced exclusively by Sakurai
  • Brawl was balanced exclusively by a group of the 4 best players at Monolith, who would identify issues and submit multiple solution proposals to Sakurai to pick one.
  • Sakurai has stated he believed the system was sound but insufficient in size. Smash 4 was the same system at Namco but with 12 members.
  • Ultimate is a continuation of the same, with the insight that we believe top Japanese players Amsa, Earth, 9B, and Ranai are/were in the group. Sakurai became so satisfied with the results of this team that he no longer requires approval for the changes they propose.
We have no official word from the balance team's metholodgy. However, we can guess:
  • They DO watch supermajors. Sakurai has commented on this.
  • They seem to balance the game with a primary target of a very high level, slightly above that of your average tournament player.
  • They probably have a slight bias towards data from Japanese tournments, which they are more likely to watch due to both language and time zones.
  • They tend to make conservative changes.
  • They do not balance around win-rates. (Which would be terrible)
  • They (almost) never make meaningful balance changes to the same character in consecutive patches.
  • They seem to be very patient when it comes to making DLC balance changes.
  • They do not appear seriously interested in balancing the Miis to the rest of the roster--only vaguely to each other.
We are pretty sure online data is not significant to them except maybe to the first patch after launch. None of their changes correlate with any of the online data we have.

What DOES correlate is this:

smash_patch_8.0.0.PNG


This is a chart indicating the 8.0.0 changes on a comparison of OrionRank 3 (pre-COVID 2020 weighted tourney results) and smash.gg lifetime tournament usage data. If we factor in various known conditions, this trendline would have predicted the 16 purely-buffed characters in 8.0.0 with 91% accuracy. (It would have predicted Lucario and Bowser Jr. instead of WFT.)

Specific character commentary that comes up a lot:
  • :ultjigglypuff: was buffed significantly in 7.0.0, which we have minimal good tournament data for yet. (She was thus "inelgible" for changes in 8.0.0) She is likely to be buffed again if she continues underperforming.
  • :ultdoc: was buffed significantly in 7.0.0, which we have minimal good tournament data for yet. (He was thus "inelgible" for changes in 8.0.0) He is likely to be buffed again if he continues underperforming.
  • :ultlittlemac: is actually in the upper half of the roster in tournament usage, and still has >50% tournament win-rate. He also has better performance in Japan than the US specifically, and is uniquely crippled by the specific stage procedure/lists currently used by tournament. Still, Little Mac's overall performance is right on the edge of the group that was buffed last patch; it is somewhat likely he will be one of the next characters looked at.
  • :ultganondorf: is actually in the top 25% of the roster in tournament usage, and still has >50% tournament win-rate. Ganondorf's mix of very solid usage+performance at 99.9th percentile play yet complete lack of results at 99.99th percentile play makes him a complex outlier that is hard to make predictions about.

tl;dr - They are a group of ex-tournament players balancing around high-level tournament play.
 
Last edited:

Thinkaman

Moderator
Moderator
Joined
Aug 26, 2007
Messages
6,230
Location
Madison, WI
NNID
Thinkaman
3DS FC
1504-5749-3616
#20
When I said significant, I meant in the objective sense that it was 8 deliberate, non-trivial changes across 5 different moves, relevant to the criteria mentioned. (That they don't deliberately alter the balance of a character in consecutive patches.)

Whether or not one thinks the shoto-tier d-tilt confirms are subjectively significant is, well, apparently subjective.
 

Artmastercorey

Smash Journeyman
Joined
Oct 31, 2005
Messages
394
#21
There's a lot of underbaked takes on this thread.

To answer the op's question:
  • Smash 64 and Melee were balanced exclusively by Sakurai
  • Brawl was balanced exclusively by a group of the 4 best players at Monolith, who would identify issues and submit multiple solution proposals to Sakurai to pick one.
  • Sakurai has stated he believed the system was sound but insufficient in size. Smash 4 was the same system at Namco but with 12 members.
  • Ultimate is a continuation of the same, with the insight that we believe top Japanese players Amsa, Earth, 9B, and Ranai are/were in the group. Sakurai became so satisfied with the results of this team that he no longer requires approval for the changes they propose.
We have no official word from the balance team's metholodgy. However, we can guess:
  • They DO watch supermajors. Sakurai has commented on this.
  • They seem to balance the game with a primary target of a very high level, slightly above that of your average tournament player.
  • They probably have a slight bias towards data from Japanese tournments, which they are more likely to watch due to both language and time zones.
  • They tend to make conservative changes.
  • They do not balance around win-rates. (Which would be terrible)
  • They (almost) never make meaningful balance changes to the same character in consecutive patches.
  • They seem to be very patient when it comes to making DLC balance changes.
  • They do not appear seriously interested in balancing the Miis to the rest of the roster--only vaguely to each other.
We are pretty sure online data is not significant to them except maybe to the first patch after launch. None of their changes correlate with any of the online data we have.

What DOES correlate is this:

View attachment 278773

This is a chart indicating the 8.0.0 changes on a comparison of OrionRank 3 (pre-COVID 2020 weighted tourney results) and smash.gg lifetime tournament usage data. If we factor in various known conditions, this trendline would have predicted the 16 purely-buffed characters in 8.0.0 with 91% accuracy. (It would have predicted Lucario and Bowser Jr. instead of WFT.)

Specific character commentary that comes up a lot:
  • :ultjigglypuff: was buffed significantly in 7.0.0, which we have minimal good tournament data for yet. (She was thus "inelgible" for changes in 8.0.0) She is likely to be buffed again if she continues underperforming.
  • :ultdoc: was buffed significantly in 7.0.0, which we have minimal good tournament data for yet. (He was thus "inelgible" for changes in 8.0.0) He is likely to be buffed again if he continues underperforming.
  • :ultlittlemac: is actually in the upper half of the roster in tournament usage, and still has >50% tournament win-rate. He also has better performance in Japan than the US specifically, and is uniquely crippled by the specific stage procedure/lists currently used by tournament. Still, Little Mac's overall performance is right on the edge of the group that was buffed last patch; it is somewhat likely he will be one of the next characters looked at.
  • :ultganondorf: is actually in the top 25% of the roster in tournament usage, and still has >50% tournament win-rate. Ganondorf's mix of very solid usage+performance at 99.9th percentile play yet complete lack of results at 99.99th percentile play makes him a complex outlier that is hard to make predictions about.

tl;dr - They are a group of ex-tournament players balancing around high-level tournament play.
Great post, alot of new information I had no idea about. Especially looking at that chart theres a huge amount of character that fall behind the very best. I Also wonder if the expert smash players that helped balanced this game knew every character well enough. In the perfect world I would think you would take the best players of those characters and get their input on what to change or leave as is.
 

Thinkaman

Moderator
Moderator
Joined
Aug 26, 2007
Messages
6,230
Location
Madison, WI
NNID
Thinkaman
3DS FC
1504-5749-3616
#22
Especially looking at that chart theres a huge amount of character that fall behind the very best.
Yes and no--it's really important to understand the nature of this data and the trends that feed into it.

All results data tends to be driven by a handful of top players. Even in wide-scope weighted results like OrionRank, in the most extreme cases an upper-tier character can have as much as a third of their results driven by a single player. In other words, a large amount of the character positions at the high end are reflecting the performance of 1-2 top players, a relationship impossible to control for objectively.

But what about characters like Wolf, Snake, ROB, or Palutena, that dominate the charts through dozens of regional players? This is not the work of a single player, but it is a reflection of a single metagame being advanced forward. Every rising competitive Lucina player is a search away from two dozen top-level VODs of Lucina vs. any opponent in the game.

For these reasons, I would normally plot OrionRank logarithmically, which would put less emphasis on the "noise" of top level characters. However, I wanted a linear correlation for that graphic (to make those colored polygons), so I presented the data raw here.



I Also wonder if the expert smash players that helped balanced this game knew every character well enough. In the perfect world I would think you would take the best players of those characters and get their input on what to change or leave as is.
This is a natural concern--it is seemingly this bias that led to MK in Brawl. (Only 4 top level balance testers, MK happened to be the one they all sucked with.)

However, too many cooks in the kitchen leads to a hot mess. Design by comittee is so bad; this cannot be overstated. At some point in your life you will find yourself in a situation where you learn this first hand, and it's unlikely anything I say will spoil the impact of that experience.

Aiming for top level play is like aiming for the center of the bullseye--a good idea regardless of your specific goals. However, aiming for the whims of individual specific top players is asking for trouble. This is an eccentric bunch with strong opinions that are biased by their (outlier) set of skills and the specific dynamics with their equally eccentric peers. They, like most players, tend to have weird biases about the characters they play, yet these biases aren't even consistent and vary wildly in magnitude based on personality.


Whether it is top players or the most casual players, people don't know what they want. This is true in most fields. In fighting game balance, this usually manifests in the form of a gradient towards homogonization--a slippery slope towards making characters more similar. It's natural, isn't it? If character A works but character B doesn't, why not just make cahracter B more like A? It's the fastest and most reliable solution, but it's throwing the baby out with the bathwater.

People will endlessly ask for Mac to be better in the air, for Doc to have a better recovery, for Sheik to have better kill moves, for Ganon to have better frame data. All of these are pretty poor ideas. People wanted Falcon's quirky mobility "fixed", but instead they made Raptor Boost reward consistent--they made Falcon more Falcon, not less.

This is a common trend with the approach of all balance changes this team has done. It's not quite accurate to say that they "increase a character's strengths"; most of the focus has been on removing flaws, obstacles, or situational barriers in the way of a character's strengths. Great examples:
  • :ultdoc: having the best power-per-startup-frame in the entire roster, and being given a d-tilt that links into his combo tree and/or kill move.
  • :ultincineroar: Revenge being a solid answer to anyone trying to outcamp his poor mobility with projectiles, and being given superior range+timing on success so that the fastest characters like Pikachu have a smaller range where dash grab coverage exists.
  • :ultisabelle: having great defensive option behind her mine, and being given better timing + consistent activation so that fast enemies cannot just run over it or rush her down.
  • :ultfalco: vertical aerial combos being really potent, and being given a late hitbox on side-b that sets those combos up on people who threaten to zone Falco out.
  • :ultzelda: Phantom; pretty much everything about it, really.
 

Artmastercorey

Smash Journeyman
Joined
Oct 31, 2005
Messages
394
#23
However, too many cooks in the kitchen leads to a hot mess. Design by comittee is so bad; this cannot be overstated.
You make some interesting points and I agree having too much chefs in the kitchen can be alot. But once the game is released and you start to see the flaws of a character arise, it wouldnt be a bad idea to have top players of that specific character give input to create a more complete analysis of how to balance the character fairly. Does that make sense? Like the alternative is to not get expert opinions at all, and get minor upgrades that dont really address major flaws in characters (by people who dont really know the character well enough) like we sometimes get. Keep the identity of a character true, I agree, but lets discuss if we are overstating, or understating aspects of their toolkit compared to the characters who seem to be the most successful.
 
Last edited:
Joined
Apr 16, 2015
Messages
1,026
Location
Somewhere in this Universe
NNID
MrGameguycolor
Switch FC
7681-9716-5789
#24
When I said significant, I meant in the objective sense that it was 8 deliberate, non-trivial changes across 5 different moves, relevant to the criteria mentioned. (That they don't deliberately alter the balance of a character in consecutive patches.)

Whether or not one thinks the shoto-tier d-tilt confirms are subjectively significant is, well, apparently subjective.
I hope you're speaking under the logic the same logic the balance team is under, because the buffs for Doc weren't significant.

What good is a few slightly stronger moves and new kill confirm if you can't even get in range to land it.
 
Top Bottom