EA Forums - Banner

NHL 20 Patch Details April 3rd

image
Check out our April 3rd patch details here.

Why does a.i. do this?

Replies

  • bryta47
    297 posts Member
    The A.I has taken a step backward for NHL 20 sadly. It's a mess.
  • bryta47 wrote: »
    The A.I has taken a step backward for NHL 20 sadly. It's a mess.

    The whole game is a mess
  • Loys33
    4 posts New member
    Hi,
    I don't know if alot of people are playing "Be a pro" mode or position lock, but there is a thing I find REALLY annoying. The thing is that while the players on the team I'm facing play almost perfectly, the AI teammates on my side are acting stupid. I understand that AI players may be "programmed" to make mistakes from time to time, but the following mistakes happen WAY TOO FREQUENTLY, in my opinion: 1- passes in front on our own net in the defensive zone, 2- back passes just before passing the blue line while entering the offensive zone, 3- passes to a player who has two or more adversaries covering him. And the higher the difficulty, the harder it gets to "compensate" for there stupidity.
    On top of that, it seems like AI players never want to shot on goal, they prefer making passes to the human player(s) AT ALL COST, which often ends up has a turn over. My point is that the AI, too often, makes mistakes that they teach you not to do in junior league hockey. I agree that those mistakes happen in the NHL, FROM TIME TO TIME, but not as often as they do in the game. It ends up to be really fustrating and you feel like you are "alone" on your team because the AI is too dumb to be allowed to do anything by itself. Could you fix your stupid AI please ?
    Thank you.
  • NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.
  • Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    I think we're getting into a gray area. I want the AI to make human errors. Their pinpoint passing and shot accuracy (even with Shot Acc at 20) is incredibly annoying. There is a fine line between AI simulated human error and AI poor logic. More times than I can remember, the AI makes such awful passing decisions and the more I see this, I think it was hard coded more than it was "AI choice".

    Example: AI skating out of their own zone, consistently passes BACK into their own zone even if there is an opposing player there in the lane or within interception range. I have posted a few videos of this illogical passing. A few times, this has resulted in the opposing team intercepting the pass and scoring. I've had my AI controlled players on a breakaway, clear path to the goal...and...they pass it back instead.

    The AI in this game is either pinpoint accurate (shooting at the side of the net goes in probably 95% of the time) or they make such illogical AI decisions.

    About two years back, EA marketed the whole idea of the AI touching up so they weren't offside, or holding the line to not go offsides until the puck went past the line. That logic obviously went out the window because I've had AI player keep skating about 5-10 feet in front of me, into the offensive zone and clearly offsides. No attempt to wait for the puck to cross the line first. Heck, I had a player skate to the offensive face off circle before the puck even crossed the line!
  • NHLDev
    1658 posts EA NHL Developer
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.
  • Loys33 wrote: »
    Hi,
    I don't know if alot of people are playing "Be a pro" mode or position lock, but there is a thing I find REALLY annoying. The thing is that while the players on the team I'm facing play almost perfectly, the AI teammates on my side are acting stupid. I understand that AI players may be "programmed" to make mistakes from time to time, but the following mistakes happen WAY TOO FREQUENTLY, in my opinion: 1- passes in front on our own net in the defensive zone, 2- back passes just before passing the blue line while entering the offensive zone, 3- passes to a player who has two or more adversaries covering him. And the higher the difficulty, the harder it gets to "compensate" for there stupidity.
    On top of that, it seems like AI players never want to shot on goal, they prefer making passes to the human player(s) AT ALL COST, which often ends up has a turn over. My point is that the AI, too often, makes mistakes that they teach you not to do in junior league hockey. I agree that those mistakes happen in the NHL, FROM TIME TO TIME, but not as often as they do in the game. It ends up to be really fustrating and you feel like you are "alone" on your team because the AI is too dumb to be allowed to do anything by itself. Could you fix your stupid AI please ?
    Thank you.

    This. When I'm position locked, my AI players are pretty much pointless. Even with all the sliders equal, they take a majority of the penalties, can't shoot or pass well, and constantly offsides. Something is happening that shouldn't be. Why can the CPU team pass and shoot with pinpoint accuracy or hardly go offsides? if my team is 99% AI controlled, wouldn't they act the same?
  • NHLDev wrote: »
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.

    Obviously I don’t think you’re just coding ”random error”. I’m trying to point out that when people here complain about the defensive AI it’s usually because the AI is making horrible misjudgements in situations it probably shouldn’t or like in the OP it’s just flat out freaking out in a fairly common situation and it’s doing so consistently. So I have to assume there is something wrong with the logic the AI makes it’s judgements in those situations.

    Then this doesn’t make sense to me either. Why the player’s mistake is affecting the AI’s judgement in the first place? If there’s a 2 on 2 rush, the weak side d man’s job is to cover the pass and not allow the weak side option to drive the net. If the player takes himself out and it becomes a 2 on 1 the AI’s job doesn’t change. It’s primiry job is still covering the pass. Then if the player wants to take control and do something more he is able to do so. I don’t want or need the AI doing miracles or be a hero. I just want it to do the simple things right with reasonable consistency. Like recognizing it’s assignment on defense, understanding what direction the play is going in and not get stuck behind the net.

    ”And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.”

    Then there’s this. Yes the better players are the better players no one is disputing that. The problem is how this game needs to be played in order to be a better player. As of now playing defense revolves around avoiding mistakes and knowing what the AI’s quirks are so you can minimize the damage it causes. It has caused most people in the 800-1000 range to play really passive defense. One example is that a lot of people never go behind their own net to chase the puck carrier because they are afraid that the AI in the slot stops covering the other side of the net or if you switch players to make sure of that, the player you brought behind the net will get stuck and lose time.

    There’s a plethora of minor things like this that you need to constantly be aware off in order to be succesful. If you aren’t aware of something weird the AI does you need to quickly notice it, switch players and bring the AI back to it’s place while you’re yelling at the screen ’no brodie your job is to cover the slot for now you can go chase in the corner later” Babysitting.

    When the ’intended’ gameplay is mostly about mastering the movement and about being in the right place at the right time to poke or hit or to take away a lane, everytime you need to micromanage your AI is time lost and affects the outcome of the game. You can say it’s part of the ’intended’ to learn the AI’s quirks and how to play with it but it makes for a really frustrating experience.
  • VeNOM2099
    2906 posts Member
    edited March 6
    NHLDev wrote: »
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.

    No... No, no, no, no, no, no, NO!

    The AI is ALWAYS out of position because it's what the "elite" asked for. They don't want the AI to be a factor in bailing out humans for their mistakes. It would work as intended if the AI didn't make more mistakes by itself than the human does. Either it allows the opponent to get behind them in their zone (allowing forced passes to get through for easy scoring chances) or following the USER around when the user is in good position just because it's programmed to chase the puck.

    They lack basic hockey awareness. What position they play. What area they are responsible for. If they would just pick up their man and stick to him and NOT allow him to get behind, human defense would improve by a significant amount. Instead, because you have to babysit the AI you're forced into this ballet of playing a mix of skillzoning with on puck defense and hoping to baby jeebus that the AI does what it's supposed to (which it doesn't).
  • bryta47
    297 posts Member
    NHLDev wrote: »
    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage.

    No, that's not what we're saying. We're saying the A.I always does this - even on a 100% CPU controlled team
  • NHLDev
    1658 posts EA NHL Developer
    bryta47 wrote: »
    NHLDev wrote: »
    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage.

    No, that's not what we're saying. We're saying the A.I always does this - even on a 100% CPU controlled team

    Yep, I get a few of you were making that point after the original callouts in the thread. That is what I was covering in my last paragraph as those two pieces would hopefully improve the main issues on both human and ai teams.

    However, the first piece that you and Venom quoted was just a response to Juppo about consistency in the ai since he was taking what I was saying before the wrong way.
  • NHLDev wrote: »
    bryta47 wrote: »
    NHLDev wrote: »
    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage.

    No, that's not what we're saying. We're saying the A.I always does this - even on a 100% CPU controlled team

    Yep, I get a few of you were making that point after the original callouts in the thread. That is what I was covering in my last paragraph as those two pieces would hopefully improve the main issues on both human and ai teams.

    However, the first piece that you and Venom quoted was just a response to Juppo about consistency in the ai since he was taking what I was saying before the wrong way.

    Imo, both camps (offline and online) are wanting to see CPU super abilities like puck pickups, Incidental stick contact while clipping through limbs dislodging pucks, and the world’s fastest reaction times nerfed and their general awareness improved. Doesn’t necessarily seem like separate issues, but I get that you were responding to two different sub-threads if you will.

    You talked about all of these factors that go into AI decision making, is there a priority list to those factors where they’re possibly really over-valuing the threats of certain factors which results in them totally stopping? For example, puck carrier is on their forehand plus has >x shooting attributes, therefore the AI will drop all other threats and immediately try to pressure the carrier?

    If so, I’d love to get a feel for that list so I can try to replicate those scenarios more so I can provide more targeted feedback. Obviously a big factor in the backside D issue is the angle they take once they’ve decided they’re going to pressure the carrier, but I think it’d be beneficial to kind of “prove” or get a good-feel for when the AI decides to abruptly stop as I think that’s the main issue here.

    On 2 on 2’s especially, if you take the carrier across and the drop the puck, I see an obvious lack of “communication” between the AI players on who should stick with who. This scenario seems to be the root cause of similar 2 on 1 and 3 on 2 situations, so again it’d be interesting to hear from you if there’s any sort of priority in the checklist of factors you’ve given the AI. It almost seems like both AI defensemen in my given scenario treat the carrier as a critical threat and they both are sprinting to cover the carrier which inevitably leads to the weakside Dman stopping and giving up his man/passing lane.

    As always, thanks for the discussion.
  • VeNOM2099 wrote: »
    NHLDev wrote: »
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.

    No... No, no, no, no, no, no, NO!

    The AI is ALWAYS out of position because it's what the "elite" asked for. They don't want the AI to be a factor in bailing out humans for their mistakes. It would work as intended if the AI didn't make more mistakes by itself than the human does. Either it allows the opponent to get behind them in their zone (allowing forced passes to get through for easy scoring chances) or following the USER around when the user is in good position just because it's programmed to chase the puck.

    They lack basic hockey awareness. What position they play. What area they are responsible for. If they would just pick up their man and stick to him and NOT allow him to get behind, human defense would improve by a significant amount. Instead, because you have to babysit the AI you're forced into this ballet of playing a mix of skillzoning with on puck defense and hoping to baby jeebus that the AI does what it's supposed to (which it doesn't).

    Agree 1000% ,19 was proof of this and certainly in 20 they continue to cater to the elite , while the rest of us are simply ignored and are told to adapt
  • bryta47
    297 posts Member
    edited March 6
    NHLDev wrote: »
    bryta47 wrote: »
    NHLDev wrote: »
    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage.

    No, that's not what we're saying. We're saying the A.I always does this - even on a 100% CPU controlled team

    Yep, I get a few of you were making that point after the original callouts in the thread. That is what I was covering in my last paragraph as those two pieces would hopefully improve the main issues on both human and ai teams.

    However, the first piece that you and Venom quoted was just a response to Juppo about consistency in the ai since he was taking what I was saying before the wrong way.

    Sorry for making the thread confusing! Not my intention. I really appreciate you being here btw.
  • VeNOM2099 wrote: »
    NHLDev wrote: »
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.

    No... No, no, no, no, no, no, NO!

    The AI is ALWAYS out of position because it's what the "elite" asked for. They don't want the AI to be a factor in bailing out humans for their mistakes. It would work as intended if the AI didn't make more mistakes by itself than the human does. Either it allows the opponent to get behind them in their zone (allowing forced passes to get through for easy scoring chances) or following the USER around when the user is in good position just because it's programmed to chase the puck.

    They lack basic hockey awareness. What position they play. What area they are responsible for. If they would just pick up their man and stick to him and NOT allow him to get behind, human defense would improve by a significant amount. Instead, because you have to babysit the AI you're forced into this ballet of playing a mix of skillzoning with on puck defense and hoping to baby jeebus that the AI does what it's supposed to (which it doesn't).

    Agree 1000% ,19 was proof of this and certainly in 20 they continue to cater to the elite , while the rest of us are simply ignored and are told to adapt

    I'm not trying to sound like a contrarian on purpose and I mostly agree with Venom as well but catering to some NHL 'elite' is starting to sound really conspiracy theorish. That's probably not the case.

    What I get from Ben's responses (even though he claimed that I'm misunderstanding him) is that the AI just has way too much on it's plate and it's making weird decisions because it's losing it's priorities.
    NHLDev wrote: »
    They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.).

    I just start to think after reading this that we have some pretty fundamental disagreements on what the purpose of the AI is and how it's behavior should relate to the players actions. I'm aware that there could be some language barrier going on and I'm actually just misunderstanding but what I'm reading between the lines is this:

    1. The AI is considering a lot of lower priority things that are affecting it's judgement and causing it to make odd desicions. Maybe I'm oversimplifying things but IMO in the defesive zone the AI has two jobs: recognize the man it's supposed to cover and stay between him and the net, I as the player will take care of the rest to best of my abilities.

    2. The AI's performance is heavily affected by the actions of the player. If the player is playing out of position, the AI will play worse and it's intended to be that way. If this is correct the obvious follow up question is how the AI wants me to play then? What is the perfect way to play to maximize the performance of the AI? There's probably as many 'correct' ways to play as there are NHL 20 players so wouldn't the easier path to decent defensive AI be that the AI wouldn't give a crap how the player plays and just do the job assigned to the individual player it currently operates? Like a real hockey player.

    Or am I misunderstanding something?
  • Juppo1996 wrote: »
    VeNOM2099 wrote: »
    NHLDev wrote: »
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.

    No... No, no, no, no, no, no, NO!

    The AI is ALWAYS out of position because it's what the "elite" asked for. They don't want the AI to be a factor in bailing out humans for their mistakes. It would work as intended if the AI didn't make more mistakes by itself than the human does. Either it allows the opponent to get behind them in their zone (allowing forced passes to get through for easy scoring chances) or following the USER around when the user is in good position just because it's programmed to chase the puck.

    They lack basic hockey awareness. What position they play. What area they are responsible for. If they would just pick up their man and stick to him and NOT allow him to get behind, human defense would improve by a significant amount. Instead, because you have to babysit the AI you're forced into this ballet of playing a mix of skillzoning with on puck defense and hoping to baby jeebus that the AI does what it's supposed to (which it doesn't).

    Agree 1000% ,19 was proof of this and certainly in 20 they continue to cater to the elite , while the rest of us are simply ignored and are told to adapt

    I'm not trying to sound like a contrarian on purpose and I mostly agree with Venom as well but catering to some NHL 'elite' is starting to sound really conspiracy theorish. That's probably not the case.

    What I get from Ben's responses (even though he claimed that I'm misunderstanding him) is that the AI just has way too much on it's plate and it's making weird decisions because it's losing it's priorities.
    NHLDev wrote: »
    They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.).

    I just start to think after reading this that we have some pretty fundamental disagreements on what the purpose of the AI is and how it's behavior should relate to the players actions. I'm aware that there could be some language barrier going on and I'm actually just misunderstanding but what I'm reading between the lines is this:

    1. The AI is considering a lot of lower priority things that are affecting it's judgement and causing it to make odd desicions. Maybe I'm oversimplifying things but IMO in the defesive zone the AI has two jobs: recognize the man it's supposed to cover and stay between him and the net, I as the player will take care of the rest to best of my abilities.

    2. The AI's performance is heavily affected by the actions of the player. If the player is playing out of position, the AI will play worse and it's intended to be that way. If this is correct the obvious follow up question is how the AI wants me to play then? What is the perfect way to play to maximize the performance of the AI? There's probably as many 'correct' ways to play as there are NHL 20 players so wouldn't the easier path to decent defensive AI be that the AI wouldn't give a crap how the player plays and just do the job assigned to the individual player it currently operates? Like a real hockey player.

    Or am I misunderstanding something?

    The AI from what I have seen just struggles with general decision making in its hockey IQ. It also seems to have too many "preset" ways of handling situations regardless of what is going on on the ice. Many times I see the AI step up and defend a player that I am already shadowing. Why does he do this? Does he not recognize I have it covered, or is it just a preset motion he is doing? Seems like a combination of the 2 imo.

    As for how to play better with your AI, I would suggest playing the position of the player you are controlling. For example, if you are in control of your RW, then your main focus should be to protect the right side point (or cover the opponents LD for example). If you are pulling every and any player you can gain control of to the high slot for example, then your AI has to scramble every other player and relocate them. Better to switch to the player that is in the vicinity of where you want to control rather than pulling any player you are controlling to that area. It isn't a perfect solution, but it will greatly help the AI do its job rather than focusing on moving everybody around to compensate for you pulling everyone out of position. Will also help you if you want to player switch to another area. It will be better if you could switch to a player that is already there rather than switching to a player your AI is trying to relocate into a better position.
  • Juppo1996 wrote: »
    VeNOM2099 wrote: »
    NHLDev wrote: »
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.

    No... No, no, no, no, no, no, NO!

    The AI is ALWAYS out of position because it's what the "elite" asked for. They don't want the AI to be a factor in bailing out humans for their mistakes. It would work as intended if the AI didn't make more mistakes by itself than the human does. Either it allows the opponent to get behind them in their zone (allowing forced passes to get through for easy scoring chances) or following the USER around when the user is in good position just because it's programmed to chase the puck.

    They lack basic hockey awareness. What position they play. What area they are responsible for. If they would just pick up their man and stick to him and NOT allow him to get behind, human defense would improve by a significant amount. Instead, because you have to babysit the AI you're forced into this ballet of playing a mix of skillzoning with on puck defense and hoping to baby jeebus that the AI does what it's supposed to (which it doesn't).

    Agree 1000% ,19 was proof of this and certainly in 20 they continue to cater to the elite , while the rest of us are simply ignored and are told to adapt

    I'm not trying to sound like a contrarian on purpose and I mostly agree with Venom as well but catering to some NHL 'elite' is starting to sound really conspiracy theorish. That's probably not the case.

    What I get from Ben's responses (even though he claimed that I'm misunderstanding him) is that the AI just has way too much on it's plate and it's making weird decisions because it's losing it's priorities.
    NHLDev wrote: »
    They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.).

    I just start to think after reading this that we have some pretty fundamental disagreements on what the purpose of the AI is and how it's behavior should relate to the players actions. I'm aware that there could be some language barrier going on and I'm actually just misunderstanding but what I'm reading between the lines is this:

    1. The AI is considering a lot of lower priority things that are affecting it's judgement and causing it to make odd desicions. Maybe I'm oversimplifying things but IMO in the defesive zone the AI has two jobs: recognize the man it's supposed to cover and stay between him and the net, I as the player will take care of the rest to best of my abilities.

    2. The AI's performance is heavily affected by the actions of the player. If the player is playing out of position, the AI will play worse and it's intended to be that way. If this is correct the obvious follow up question is how the AI wants me to play then? What is the perfect way to play to maximize the performance of the AI? There's probably as many 'correct' ways to play as there are NHL 20 players so wouldn't the easier path to decent defensive AI be that the AI wouldn't give a crap how the player plays and just do the job assigned to the individual player it currently operates? Like a real hockey player.

    Or am I misunderstanding something?

    As for how to play better with your AI, I would suggest playing the position of the player you are controlling. For example, if you are in control of your RW, then your main focus should be to protect the right side point (or cover the opponents LD for example). If you are pulling every and any player you can gain control of to the high slot for example, then your AI has to scramble every other player and relocate them. Better to switch to the player that is in the vicinity of where you want to control rather than pulling any player you are controlling to that area. It isn't a perfect solution, but it will greatly help the AI do its job rather than focusing on moving everybody around to compensate for you pulling everyone out of position. Will also help you if you want to player switch to another area. It will be better if you could switch to a player that is already there rather than switching to a player your AI is trying to relocate into a better position.

    Everything you said is correct and the problems usually arise when you're forced out of that into non ideal situations . Like in your example I were to bring the RW to the same side corner to help pressuring the puck carrier and then the AI would insist on bringing the center to replace the RW on the point even though it's not needed and the center's original position was a much more important one to take care of. I'm not sure if this particular example actually happens but it kinda shows what type of situations I mean.

    Or another example that I know that actually happens fairly often... The puck carrier is circling around between the right corner and behind the net (from the goalie's perspective) but you don't want to go chase or hit him for some reason with the RD. You just take away his chances of cutting to the middle. Then the AI controlled LD will come stand right next to you and leaves the backdoor open because I assume the AI wants you to go chase the puck carrier.
  • VeNOM2099
    2906 posts Member
    Juppo1996 wrote: »
    I'm not trying to sound like a contrarian on purpose and I mostly agree with Venom as well but catering to some NHL 'elite' is starting to sound really conspiracy theorish. That's probably not the case.

    I don't particularly like that notion myself. It's too easy of an "out" to blame the devs of only listening to the "elites" out there. However, many of the things they have/ at one time or another, complained about have been taken care of via an update.

    Remember back when they cried about the backup goaltenders in HUT? That option was removed soon after. A few years back, when they complained about the scoring being too difficult. Goalies got nerfed. When they complained the AI was too good and was bailing out "bad players" and "skillzoners", AI got hit.

    Let's say (hypothetically) someone like Nasher complaining alone can't sway EA's mind, but what if half of his 465K youtube subcribers decided to flud EA's facebook and/or twitter acc with said complaint. Would EA listen? I think so...

    Again, it's a little unfair to simply foist this off on their laps and blame them. But you can't underestimate the influence they have over thousands, sometimes hundreds of thousands who may or may not know any better than to parrot what they utter on stream or elsewhere. And whether it's directly or indirectly, they ARE responsible for some of the changes we've seen affect this game. Sometimes for the better, but sometimes... maybe even often times, for the worse.
  • JetsNY81
    4 posts New member
    The 'elite' deciding the direction of the game makes zero sense. I've watched top players complain and yell at their AI just like we do, they don't want the AI to be dumbfounded in common sense situations. Why would they want that?

    Even the year people complained about skill zoning (was it 18? can't recall the exact year), the AI was still as bad it is now and would make non-sensical plays. The AI skating away from the desired positioning has been a thing since current gen, it's not something new to 20.

    None of this is anything new. The AI has been bad in this series forever. I'm not asking for the AI to be overpowered, but they should not be randomly abandoning positioning. I've watched the AI literally skate away from passes the second I release the button, go offsides the very second I choose to cross the blue line, and shy away from the puck as if they were scared to death of it for some reason. There needs to be more common sense logic.
  • Juppo1996 wrote: »
    Juppo1996 wrote: »
    VeNOM2099 wrote: »
    NHLDev wrote: »
    Juppo1996 wrote: »
    NHLDev wrote: »

    If done right, the ai shouldn't be perfect, they should make decisions based on what they see in front of them and make organic realistic mistakes. In that programming, they can also make some unnatural and weird mistakes

    Is this really the creative direction you guys are going for? I’m not a video game coder but to my layman brain this seems like a terrible choice for a competitive game. Shouldn’t the AI in 1vs1 online gameplay rather be predictable and consistent so it leaves the max amount of the outcome to the players?

    Honestly the last thing I want as a player is the AI making simulated human errors. I just want the AI to be predictable so when I make a mistake I can be 100% sure where my other players are so I can cover up for my own mistake. And if I end up screwing up I can blame myself and not have this gray area of the AI being weird constantly.

    You are basically leaving intentional room for exploits and frustrating inconsistensies to happen with this aproach.

    No wonder the defense feels like AI babysitting.

    We aren't coding random error to the ai. They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.). And I am only saying that in response to your comment about how it should leave the max amount of outcome to the players as we see that consistently the same players are having success and do know how to consistently have success.

    We still want to improve the ai. Nothing said above takes away from how they can be improved. Their desire to worry more about pass lane coverage rather than stay goal side of the weak side player is a big one that contributes to some of the issues as well as their scoring for where the biggest threats are but a human player playing proper defense themselves definitely limits the errors with the ai as well.

    No... No, no, no, no, no, no, NO!

    The AI is ALWAYS out of position because it's what the "elite" asked for. They don't want the AI to be a factor in bailing out humans for their mistakes. It would work as intended if the AI didn't make more mistakes by itself than the human does. Either it allows the opponent to get behind them in their zone (allowing forced passes to get through for easy scoring chances) or following the USER around when the user is in good position just because it's programmed to chase the puck.

    They lack basic hockey awareness. What position they play. What area they are responsible for. If they would just pick up their man and stick to him and NOT allow him to get behind, human defense would improve by a significant amount. Instead, because you have to babysit the AI you're forced into this ballet of playing a mix of skillzoning with on puck defense and hoping to baby jeebus that the AI does what it's supposed to (which it doesn't).

    Agree 1000% ,19 was proof of this and certainly in 20 they continue to cater to the elite , while the rest of us are simply ignored and are told to adapt

    I'm not trying to sound like a contrarian on purpose and I mostly agree with Venom as well but catering to some NHL 'elite' is starting to sound really conspiracy theorish. That's probably not the case.

    What I get from Ben's responses (even though he claimed that I'm misunderstanding him) is that the AI just has way too much on it's plate and it's making weird decisions because it's losing it's priorities.
    NHLDev wrote: »
    They are consistent overall but there are more variables than you make out. There isn't just a puck carrier and a weak side pass option -- There are players on their forehand or backhand, gliding ready to pass/shoot vs skating to or away from the net, different velocities/future positions, more or less teammate support, etc.. so even consistency across those pieces when they all line up the same will have variance.

    What has been called out so far in this thread is actually that the ai is consistently blowing their coverage when the human blows their coverage. So you are correct that the ai could be in a better spot for you to then switch into and have a second chance to make up for your mistakes but it is still driven off a human error in a lot of cases in the first place (either directly through a missed play, or giving your opponent more time for your team to get out of place, etc.).

    I just start to think after reading this that we have some pretty fundamental disagreements on what the purpose of the AI is and how it's behavior should relate to the players actions. I'm aware that there could be some language barrier going on and I'm actually just misunderstanding but what I'm reading between the lines is this:

    1. The AI is considering a lot of lower priority things that are affecting it's judgement and causing it to make odd desicions. Maybe I'm oversimplifying things but IMO in the defesive zone the AI has two jobs: recognize the man it's supposed to cover and stay between him and the net, I as the player will take care of the rest to best of my abilities.

    2. The AI's performance is heavily affected by the actions of the player. If the player is playing out of position, the AI will play worse and it's intended to be that way. If this is correct the obvious follow up question is how the AI wants me to play then? What is the perfect way to play to maximize the performance of the AI? There's probably as many 'correct' ways to play as there are NHL 20 players so wouldn't the easier path to decent defensive AI be that the AI wouldn't give a crap how the player plays and just do the job assigned to the individual player it currently operates? Like a real hockey player.

    Or am I misunderstanding something?

    As for how to play better with your AI, I would suggest playing the position of the player you are controlling. For example, if you are in control of your RW, then your main focus should be to protect the right side point (or cover the opponents LD for example). If you are pulling every and any player you can gain control of to the high slot for example, then your AI has to scramble every other player and relocate them. Better to switch to the player that is in the vicinity of where you want to control rather than pulling any player you are controlling to that area. It isn't a perfect solution, but it will greatly help the AI do its job rather than focusing on moving everybody around to compensate for you pulling everyone out of position. Will also help you if you want to player switch to another area. It will be better if you could switch to a player that is already there rather than switching to a player your AI is trying to relocate into a better position.

    Everything you said is correct and the problems usually arise when you're forced out of that into non ideal situations . Like in your example I were to bring the RW to the same side corner to help pressuring the puck carrier and then the AI would insist on bringing the center to replace the RW on the point even though it's not needed and the center's original position was a much more important one to take care of. I'm not sure if this particular example actually happens but it kinda shows what type of situations I mean.

    Or another example that I know that actually happens fairly often... The puck carrier is circling around between the right corner and behind the net (from the goalie's perspective) but you don't want to go chase or hit him for some reason with the RD. You just take away his chances of cutting to the middle. Then the AI controlled LD will come stand right next to you and leaves the backdoor open because I assume the AI wants you to go chase the puck carrier.

    I know there are fundamentals flaws with the AI. I see it as well when I play 6s in EASHL being the LD and I see my RD partner quit and leave me with an AI partner.

    On every single rush, the AI RD will actually impose himself on me and break out all the way up the left side. He will also literally stand 1 foot away from me in the offensive zone for about 3 seconds before realizing either I got it, or he should be elsewhere. The only time he won't do it is if I break out very aggressively and practically go ahead of my wingers which is to say, the dumbest thing to do as a defender under most circumstances.

    Every.

    Single.

    Time.
Sign In or Register to comment.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!