General input/fluidity improvements
1.) Stance should not swap when locked on via target movement unless you are slowed by a well timed Dash or Evade (when slowing effect occurs). The player should only change its stance relative to a locked on target when done so manually or by using an attack that changes their stance. This makes it so player intention is not removed by simple movement or a poorly timed evade/dash by the enemy. It also allows feinted attacks to ignore re-orienting the player stance based on orientation to their opponent, and swapping to a different combo/repeating the same attack unintentionally. Making feints less wonky when used to counter Evade and Dash.
This wouldn't be much of an issue if lag didn't occur, but more often than not your character will swap stance a split second before input registration leading to using an attack you did not want to use. Even when on your screen the opponent hasn't actually moved to the location yet. Any amount of rubberbanding or jitter introduced by latency can make this even more inconsistent. Hence why it would be best to just remove the mechanic of auto swap when locked on, and work the disorientation into a consistent mechanic (well timed evade causing loss of tracking and switching the dodged players new stance; Functioning as if they lose lock on for a short period of time before re-locking once the slowdown ends).
2.) Whilst blocking an attack (during blockstun), stance should not reset to side previous to hit. Instead, it should only delay the stance swap by stopping it whilst the player is actively blocking an attack. Any fast string can lock a blocking player into one stance. Furthermore, due to the delay before defensive ability startup, attempting to Evade/Parry/Absorb/Dash will lower block prior and thus make the defendant take a hit. Even when their input was the correct counter to the enemy's attack. Giving this freedom to swap stance opens up more oppurtunity for counterplay and mindgames. While, at the same time, making the act of pressuring your opponent more dynamic (by increasing the risk to reward [proper reads lead to continuing pressure and getting a few hits in, failure to properly read causes pressure to reverse or disappear]).
3.) Hitstun should cancel any buffered inputs, and only allow input buffereing at the end of the hitstun animation. Because there are input delays on defensive abilities, and hitstun registers inconsistently when latency is present, it is easy for your attack/denfensive ability to come out right after you get hit even when the input occurred before being hit. This can cause unnecessary frustration when the player attempts to defend/attack, gets hit, and then when they notice they got hit and wish to attempt a different approach, the character actively does what they ment to do prior.
4.) All built in input delays should be removed and replaced with startup times. Evade/Parry/Absorb/Dash all have built in input delays before the actual attack occurs. The reason I know this is because you can input any of these a split second before getting hit, and they will buffer during hitstun and trigger afterwards. This change in conjunction with the previous would eliminate the issue of unintentional input entirely, whilst keeping the same timing associated with the defensive abilities.
Take into account that all these issues only take effect within a short span of time and are hard to notice if you don't actively search for them (The only exception being when latency is so high it compounds the problems associated with input). However, they still effect player enjoyment, and limit variety the of viable tactics with certain matchups. Even if they are hard to notice for most people.
This wouldn't be much of an issue if lag didn't occur, but more often than not your character will swap stance a split second before input registration leading to using an attack you did not want to use. Even when on your screen the opponent hasn't actually moved to the location yet. Any amount of rubberbanding or jitter introduced by latency can make this even more inconsistent. Hence why it would be best to just remove the mechanic of auto swap when locked on, and work the disorientation into a consistent mechanic (well timed evade causing loss of tracking and switching the dodged players new stance; Functioning as if they lose lock on for a short period of time before re-locking once the slowdown ends).
2.) Whilst blocking an attack (during blockstun), stance should not reset to side previous to hit. Instead, it should only delay the stance swap by stopping it whilst the player is actively blocking an attack. Any fast string can lock a blocking player into one stance. Furthermore, due to the delay before defensive ability startup, attempting to Evade/Parry/Absorb/Dash will lower block prior and thus make the defendant take a hit. Even when their input was the correct counter to the enemy's attack. Giving this freedom to swap stance opens up more oppurtunity for counterplay and mindgames. While, at the same time, making the act of pressuring your opponent more dynamic (by increasing the risk to reward [proper reads lead to continuing pressure and getting a few hits in, failure to properly read causes pressure to reverse or disappear]).
3.) Hitstun should cancel any buffered inputs, and only allow input buffereing at the end of the hitstun animation. Because there are input delays on defensive abilities, and hitstun registers inconsistently when latency is present, it is easy for your attack/denfensive ability to come out right after you get hit even when the input occurred before being hit. This can cause unnecessary frustration when the player attempts to defend/attack, gets hit, and then when they notice they got hit and wish to attempt a different approach, the character actively does what they ment to do prior.
4.) All built in input delays should be removed and replaced with startup times. Evade/Parry/Absorb/Dash all have built in input delays before the actual attack occurs. The reason I know this is because you can input any of these a split second before getting hit, and they will buffer during hitstun and trigger afterwards. This change in conjunction with the previous would eliminate the issue of unintentional input entirely, whilst keeping the same timing associated with the defensive abilities.
Take into account that all these issues only take effect within a short span of time and are hard to notice if you don't actively search for them (The only exception being when latency is so high it compounds the problems associated with input). However, they still effect player enjoyment, and limit variety the of viable tactics with certain matchups. Even if they are hard to notice for most people.