Forum Discussion

Limp_KidzKit's avatar
Limp_KidzKit
2 years ago

Dzone Coverage Issues

Hello all,

Back with another AI breakdown to show that in-zone play is fundamentally flawed which doesn't allow for a "sim" experience to be achieved regardless of sliders. For context, the user is in the steel blue, CPU in white. The team making egregious coverage mistakes will be in white.

Screenshot 1 - The faceoff

Notes:

  • FO alignment is just silly due to the boardside defensive wing lining up with the offensive wing (see assignments below).
    • Automatically isolates your boardside wing.
    • Causes assignment confusion/wide open option if puck is won to offensive wing as the defensive wing will need to pick between puck carrier or actual assignment
  • Proper Assignments should be the following (Defense -> Offense)
    • Center -> Center
    • Strong-side wing -> Strong-side D
    • Weakside wing -> Weakside D
    • Strong-side D -> Strong-side Wing
    • Weakside D -> Weakside Wing

First moments after puck drop:

Bare with me here as there's a lot to unpack just moments after the puck drop and one D-to-D pass:

  • Boxes
    • I drew three boxes just to highlight how egregious the coverage is 2-4 seconds after puck drop.
      • Red Box - the boardside issue talked about earlier shows immediately as the defensive RW is isolated in a 1-on-2
      • Blue Box - The Defensive left wing decides to tie-up with offensive right wing (77 blue) which actually isolates himself in another 1-on-2
      • Green Box - Due to a combo of bad alignment + bad decision in blue box, we have a 3-on-1 low while simultaneously not covering the "1" low due to coverage confusion
  • Orange Arrows:
    • These are what SHOULD be the assignments. It's messy, it's not a "pathing" ask, I'm simply drawing lines to what should be their assigned mark
    • Notice how the white team's RD (captain's patch on the hash) is rushing to "cover" the now open offensive RW (not his assigned mark) as the left wing (blue box) is now leaving 77 to go to the point like he should've been doing all along. Meanwhile his assigned mark is about to cut across the middle of the slot uncontested.
    • This is due to the fact that the white team LD (78) started wondering to find a new player to cover as his own LW took what should've been his assigned mark (77 blue).
    • And because both defensemen are out of position, the center (12, lower right of green box) is now leaving his mark to go stand and cover some arbitrary spot the game thinks the RD should be standing in.
  • The Results:
    • We actually have 0/5 offensive players covered in this screenshot after 2 seconds of gameplay. ZERO OF FIVE 
      • 1/5 due to bad alignment
      • 1/5 due to bad decision making
      • 3/5 due to domino effect of 1 bad decision - this quite simply is unacceptable and should not happen. 1 mistake should not cause 3 other players to wildly chase new assignments. This should result in 1 player open (the D) not musical chairs happening below the dots for the defensive team

Above is the proof. Look at all that space!! In the middle of the slot! Due to ONE D-to-D pass. One....one pass leads to the entire slot being open with multiple options to get the puck through a "seam" (not much of a seam when it's a wide open field due to a complete lack of defense whatsoever)...This is because:

  • Bad alignment confuses RD which confuses LD...they are hilariously seen here swapping sides of the rink 2 seconds in despite their OG assignments being on their original sides of the ice
  • Bad decision by 43 white means he's late to close down the D who has the puck now. This combined with a bad angle out = cross-seam pass wide open
  • 76 white is obviously late in his attempt to cover what wasn't his assignment which gives the offensive player tons of time and space
  • 12 white is covering the crease apparently? his mark is standing with a free stick right next to him...although not sure why 27 isn't screening, doing literally nothing where he is.

Remediation and Benefits of Remediation:

  • Quite simple solution actually. Make defense man-to-man in 5v5 settings. As soon as someone has "taken" a mark, it should not change. Is this the most realistic? Depends on the system, but in many cases, yes actually! Is more realistic than the current coverage assignments? 100000% yes!
    • Can someone with more programming experience comment on the "ease" of this change? Clearly they're programmed to play a zone, to identify threats, to even cover for blown assignments, I can't imagine asking them to simply identify one mark and stick to that one mark indefinitely would be any harder of a challenge than whatever they had to do to get the AI to do what it does now, right? Man-to-man sounds a lot easier in theory to me, am I wrong?
  • If we make assignments clear, the AI will be much more effective at providing a fun experience for ALL (offline and online - read WIN/WIN FOR EVERYONE) users.
  • On this note, it would make competitive 1v1 play more fun too. Blown coverages should be able to be easily diagnosed due to overcommits and the AI should be able to handle off-puck assignments much more consistently assuming the human hasn't pulled them
  • This should also allow us to reinstitute AI that actually pressure their assignments in the game. I know this was removed for offline and online users in 21 thanks to the HUT community and their "skill zoning" complaints which was a common tactic by anyone with an ounce of skill as it was painfully obvious that the AI couldn't be trusted to execute the most routine defensive zone coverages in any consistent manner. In reality, this tactic was the SYMPTOM, not the PROBLEM. This competitive community has proven over and over and over and over and over and over and over and over and over again that they can't properly diagnose problems, they merely chase symptoms away which in-turn introduces new symptoms because the underlying problem is quite literally unaddressed due to their inability to identify it. Players would not sit and "skill zone" essentially taking no mark, if their AI players could properly cover their marks consistently. Why would they allow the puck carrier free time, space, ability to isolate other AI players if they had confidence in their defensive AI teammates? The answer is simple, they wouldn't! It would be a losing strategy! 
  • Dynamic Scoring - We should be able to actually allow for first shot goals on goalies that aren't glitches or flukes. I'm talking just good shots from the "house" because we now have AI that play tight, positionally sound off-puck marks. So, if you allow time and sapce to the wrong player, first shot goals shouldn't require some magic pattern or a backdoor tap-in. Just let slot shots go in! Punish users for losing gaps! 
  • Reducing puck carrier speeds to something human - Awesome! Now that hockey is actually being played off puck, we don't need 270 pivots at full speed to be something you can do because now you can actually pass the puck, set up give and gos, and get rewarded for just good "normal" shots. Sweet! Dynamic gameplay that actually resembles the sport!
  • Reducing puck speeds - we also don't need 600 mph passes and shots because again, with hockey being played off puck, just planning your attack a bit with some touch passes and movement will naturally create time and space! 
  • Increasing incidental contact - This should help players on both sides of the puck who use it! Setting up picks, using your body to contain all things we should be wanting to see more now that coverage is tight. This will lead to the zone starting to imitate what I just watched on tv for the last 6 hours tonight. Lots of movement, passing, "screens" and picks, and rolls, that's what hockey looks like in the zone. Cycling is not a full sprint from corner to blueline then whip a 400 mph backhand pass low and rinse/repeat circles lol...that's not hoclkey..iots not even close.

Thanks for reading my rambling. Fix defensive coverage. That's it. It's a foundational piece that needs to be fixed and once it is, it will make tuning the game much easier to satisfy the sweats and the people that actually watch/know hockey with a middle ground. 

2 Replies

  • NeonSkyline21's avatar
    NeonSkyline21
    Seasoned Ace
    2 years ago
    @Limp_KidzKit First off, let me thank you for doing this. As someone who strictly plays online versus, teammate AI is this game's number 1 glaring issue that needs to be remedied for the game to become more fun and satisfying IMO.

    Now, as far as the faceoff alignment goes, the white team (which you say is CPU controlled?) seems to be using the "aggressive" faceoff alignment setting and therefore has three defenders all bunched up on the left side of the dot. I hate that alignment for defensive zone faceoffs and I don't see any benefit in using it. The standard alignment would have the RD directly behind the C as normal. That said, I don't see anything wrong with the white team's RW alignment. They are supposed to be there so they can attempt to rush to the point depending on where the puck goes.

    It's crazy how incorrectly the defending team's AI played the 2 seconds after the puck drop. Why would the LW go to the guy in the slot when the puck has been won to the point? That is supposed to be the LD's responsibility. As you said, it must be that the AI players are "forced" to play zone coverage and the defensive settings determine those zones. In this case, the CPU team likely has their setting set to "collapsing" so the defending LW refused to go to the point but in such a case, I'd rather see the strong side winger go to the point because the puck is in one of the opposing d-men's possession. You see this in the NHL all the time: a team will be playing a collapsing style D when the puck is low in the zone which allows the offensive team to bank the puck up the boards to the point easily but THEN (and this is key) the defending strong side winger attacks the point quickly while the weak side winger stays a little bit higher than he previously was but not super high.

    As far as programming goes, AI pathing in general is very difficult... especially in such a dynamic game like hockey. But as you said, a lot of the work is already done and right now it's more of a philosophy thing rather than a possibility thing. I'm sure they've tested out man-to-man coverage and I'd be curious to hear why they don't allow for us to have it as a coverage option. I think the main problem right now is that the AI gets too confused too easily and just chooses to hang out in "spots" rather than actively engage their mark more often than not. Take cross-crease defense, for example. If you don't take control of the weak side defender and actively do something about the guy on the backdoor, your AI defender will rarely disrupt the play and keep it from connecting. That shouldn't be the case, IMO.
  • Limp_KidzKit's avatar
    Limp_KidzKit
    2 years ago
    @NeonSkyline21 wrote:
    @Limp_KidzKitFirst off, let me thank you for doing this. As someone who strictly plays online versus, teammate AI is this game's number 1 glaring issue that needs to be remedied for the game to become more fun and satisfying IMO.

    Now, as far as the faceoff alignment goes, the white team (which you say is CPU controlled?) seems to be using the "aggressive" faceoff alignment setting and therefore has three defenders all bunched up on the left side of the dot. I hate that alignment for defensive zone faceoffs and I don't see any benefit in using it. The standard alignment would have the RD directly behind the C as normal. That said, I don't see anything wrong with the white team's RW alignment. They are supposed to be there so they can attempt to rush to the point depending on where the puck goes.

    It's crazy how incorrectly the defending team's AI played the 2 seconds after the puck drop. Why would the LW go to the guy in the slot when the puck has been won to the point? That is supposed to be the LD's responsibility. As you said, it must be that the AI players are "forced" to play zone coverage and the defensive settings determine those zones. In this case, the CPU team likely has their setting set to "collapsing" so the defending LW refused to go to the point but in such a case, I'd rather see the strong side winger go to the point because the puck is in one of the opposing d-men's possession. You see this in the NHL all the time: a team will be playing a collapsing style D when the puck is low in the zone which allows the offensive team to bank the puck up the boards to the point easily but THEN (and this is key) the defending strong side winger attacks the point quickly while the weak side winger stays a little bit higher than he previously was but not super high.

    As far as programming goes, AI pathing in general is very difficult... especially in such a dynamic game like hockey. But as you said, a lot of the work is already done and right now it's more of a philosophy thing rather than a possibility thing. I'm sure they've tested out man-to-man coverage and I'd be curious to hear why they don't allow for us to have it as a coverage option. I think the main problem right now is that the AI gets too confused too easily and just chooses to hang out in "spots" rather than actively engage their mark more often than not. Take cross-crease defense, for example. If you don't take control of the weak side defender and actively do something about the guy on the backdoor, your AI defender will rarely disrupt the play and keep it from connecting. That shouldn't be the case, IMO.

    No problem at all, glad I finally got a response that made this worth it lol.

    Faceoff alignment, I get it's a personal preference thing. I know that wing on boards is something seen depending on the coach, but I believe if you're playing man-to-man it's going to cause issues if the board-side offensive wing gets the puck. Who? tracks to the middle? If the wing is supposed to, it leaves the D wide open and poses a 2nd question of "now who goes out?" as the strong-side D has an lengthy gap between both the strong-side offensive D and the Strong-side wing who just cut the slot. 

    I much prefer the alignment below to keep coverage swapping to a minimum and have more straightforward rules/responsibilities on the draw

    The main issue being if your Center gets walked you're outnumbered low, but you also should hopefully trust your center's enough to not get beat super clean off of a faceoff haha but I understand the risk.

    As to the in-zone strategy conversation:

    I really appreciate your thoughts, glad it's not just me. I think after 30 years it's about time we retire collapsing, staggered, and tight point along with overload, crash the net, and behind the net haha. I mean seriously, I think these options have been available since NHL 95...maybe 98 at the latest. Hockey has evolved a million times over since the 90's and a thousand times over since even 2010 haha.

    I especially liked your 2nd paragraph because it outlines a very "universal" approach to defense in modern hockey. All teams "collapse" and dare I say run "tight point" at the same time depending on the side of the ice they're on relative to the puck. I believe none of the options in game offer a truly "man-to-man" approach to defense and this is a conclusion I've come to after 100's of hours testing them out. We don't need these options in the game anymore, we need one universal approach to defensive strategy and then either a slider or option for gap integrity or whatever you want to call it that controls how close you want guys tracking their marks. I know IRL we see true M2M and sort of a "2-1-2" hybrid model where D play zone, center is your "help" and the two forwards high play that yin-yang of pressure strong/collapse weak, but I just don't think we'd see a much different result when implementing that in a game compared to what we currently have. I think M2M is the optimal way to go as it simplifies hockey for the casual audience EA loves and caters their competitive defaults too (lol), it quite literally is an IRL approach to defense, it will make playing the game offline more engaging, and will allow us to have AI pressuring their marks again which is going to make the game much more dynamic, engaging, and easier to tune as the game will flow much more like the real sport.

    To this point, I'd say most teams run a fairly universal offense too with most teams running some variant of two guys low and a high forward + 2 defenseman trying to create an almost "umbrella" shape once established. I actually created a "common cycle" thread a long time ago on here that showed an example of what a common cycle could look like and how ti would lead to more in-zone off-puck movement as well as more consistent net-front coverage which would give users a much wider range of ways to attack the defense then we currently see with our 8 AI players standing still more often than not watching the two humans (or one 1 human and 1 on-puck AI D) play ISO basketball against one another. I believe this should be the future of this series. Get rid of these fictional approaches to both sides of the puck, take the time to program a fluid offensive concept, take the time to program M2M defense, and see how drastically improved the flow of in-zone play is compared to our current offerings. It would be night and day and would allow players of all backgrounds to have a path to success (again, WIN/WIN for ACCESSIBILITY WOOOHOO - cause you know, that's what we're about here!) while providing even more options for experienced users who have 1v1 skills and vision to truly create a skill gap in this game beyond twtich skills and mastering 1-2 moves.

    I also liked how you pointed out that the "zones" the AI are asked to cover and actually what's messing their decisions up and causing both non-commits and over-commits/coverage. I 100% agree with this. The more they are asked to "think" the more the game breaks. Hockey is an impossibly complex sport to program coverage rules for. Same reason why Madden struggles with zone defense. It's because "zone" IRL applications are much more of a "calculated man" and it works because humans have the capacity to override their "zones" when necessary which the AI in all sports video games don't seem to possess.

    Great response, Neon. Glad to have engagement on this. I won't speak for you, but I know that my advocacy for these AI updates is not coming from merely a "sim" perspective, but from a general gameplay enjoyment perspective. Regardless of whether or not a user truly understands hockey, having the ability to play/approach the game from a legitimate competitive standpoint in a variety of different ways will make this that much better of a video game in general, hockey sense aside. But an added plus would be that somsone who wants to learn more about the sport could actually see and feel something similiar to what they watch on tv right now. That's simply something they cant do with the current state of the game. It's a win-win from a video game stand point, its a win-win from a hockey standpoint, its a win-win from a casual standpoint, its a win-win from a competitive standpoint. Hockey is an extremely dynamic sport, its a beautiful art when flowing correctly, thats why many love it so much, its time the video game starts being a great asset for the sport and allowing people to experience the flow of the sport at their fingertips because once they do, they won't put it down. That the path we were on from 09-14, that's the path we need to get back to.

About NHL 24 General Discussion

Discuss the latest news and talk with us about your experiences in NHL 24.4,284 PostsLatest Activity: 19 hours ago