NHL 24 AI Feedback

---

@Zkibu wrote:

Isn't 55 Scheifele? So 13 (Vilardi) is the F1, 81 is the F2 etc. There is no D in the video. This is probably 1-2-2 passive.Tons of text which was quite nice to read thanks for that and I totally agree that EA has bo idea how the forecheck works irl, but in the end he managed to intercept the puck so it's kinda ridiculous to do so much effort to prove that he "failed" in his forecheck. Besides the basic rules of fc (at least in my country) 1. Puck pressure, 2. Support 3. Coverage. I see all the principles happening here.

---

@Zkibu I appreciate this response as it's well thought out and offers me more to build on now, thank you! I "go this far" to "prove a point" because we see the SAME EXACT THING happen to our community members who post videos about offense vs defense balance and we always see "well you could've done x, y, and z" while missing the "high-level" point of the video, so I am merely providing my knowledge of the sport at a deeper level than many ask for in the same manner to show why doing a FC or NZ concept 50% right really isn't "right" and that it doesn't mean there's proper hockey concepts programmed into this game which my point. So, how are should I go to express my opinion? I feel like I can't make my points without providing context, concepts, and then applying to the videos provided. Again, I see feedback here all of the time that videos are too short, no context to the play, you couldve moved .1 milimeter this way to correct, so I'm being a resource for the community to provide that knowledge of the real sport to further our discussions around AI improvement. I don't see how adding tons of context and information is "going too far" I actually see it being the contrary. If I simply said "AI isn't programmed to play hockey" while failing to provide evidence, I'd be merely a "complainer" who doesn't have any real proof. So it's a fine line. I find it odd you find adding context and information to be arrogant...I'm simply providing what I feel to be the necessary infromation to backup my claims rather than being another complaining voice with zero context. My apologies if I came across as rude.

Back to the scenario at hand - 55 being a forward is a great point - it's actually why I asked Showtime to provide me with additional context to this play as this four second clip is what we're all supposed to use as objective proof that the AI can play hockey which we all know is an extremely small sample size.

I also think we can all admit that results-driven analysis isn't indicative that "everything went right" correct? I mean, we see "proper" coverage get beat simply by an excellent 1v1 move being made, and we see "improper coverage" leading to turnovers all of the time in real hockey. We can't say that something "worked properly" due to the result of a FC or a trap just because the other team didn't execute. That doesn't "prove" the AI were programmed to play a certain FC or NZ correctly. Am I wrong?

Back to your point, I approached this scenario as a failed FC assuming where guys are off-screen as it's a very small picture with  5/10 guys missing for 2 of the 4 seconds, and 4/10 guys missing the back half. But at the end of the day, my point is still going to stand that F1 attacking the puck is not objective proof that the AI knows hockey nor is this video objective proof of proper coverage and I'm going to explain why. After all, that was the argument. Showtime argued that his 4 second clip showing 30% of the ice is hard evidence that the AI are programmed to play hockey and I disagreed with evidence to the contrary.

Moving on assuming this is a one-man deep scenario (with 55 and 81 being high right?) and we're actually seeing a "1-2-2 Blue" being executed - then we are now discussing essentially an aggressive NZ rather than a FC.

If we're executing a 1-2-2 NZ, there's two common concepts that are usually applied interchangeably throughout a hockey game due to what fits the scenario properly:

1. A "push" where F1 would be on the strong-side boards pushing the puck away from him covering the strong-side boards, so forcing a regroup and F2 is covering the center pass, and F3 is covering the far boards outlet.
2. A "split" where F1 is splitting the regrouping defenders while F2 covers the strong-side boards and F3 covers the middle lane.

If we go back to our video, let's apply these rules and see if we truly are getting "proper" coverage:

• Puck starts middle with defensive support and 0 puck pressure, we're going to want to "split" here if running a 1-2-2 blue to isolate the puck carrier
• We can see that 13 takes an angle directly at the puck which is a bad start to a split. It allows for the puck carrier to make an escape to the inside which kills our split in theory
• No matter if its a split or push, 55 should be on the boards to prevent an easy up. He's clearly in the middle covering nobody providing an easy outlet. Again, not user error, AI started showtime there
• No matter if its split or push, 81 should be in the middle. His rotation is late but it actually works in this play because 13's initial angle is improper
  • This is a perfect example of improper rotation and improper angles working out as it covered the left side outlet
• We clearly see that the puck carrier had an opportunity to kick it back to his partner very early on, and we see he has a wide open board outlet, and we see he has an escape route behind the net. If we're running a NZ trap which the 1-2-2 is in theory right? Then we want the puck to advance to the choke points provided by 81 and 55 to force a turnover. So this NOT a well-executed "trap" if you will.
• At the end of the clip we see a turnover due to the human panicking from the pressure due to their improper reads

So again, back to the original point of this entire discussion:

• Did the AI execute a proper FC/Trap? I think I've demonstrated that we can safely say "no" but to each their own I guess
• Is F1 sprinting at the puck undeniable proof that the AI is programmed to play these concepts correctly? Again, I would say "no" and that F2 and F3's movements/rotations would be the indicative marker as well as F1's initial angle at the puck
• Did it work out in the end? Of course! But again, a "yes" here doesn't make the previous two answers "yes" right? 

So in closing, how is all of this context "ridiculous?" I guess to answer your list:

1.  I do not see puck pressure until there was already multiple passing options for the puck carrier. The pressure was late

2. Support - we lack strong-side support here. Boards are wide open.

3. Coverage - Again, the D partner is open, boards are open, we have two guys essentially beat here with a strong-side boards outlet. IRL, this an easy chip up the wall two 2 forwards in the NZ and the weakside D IRL would read an easy up and easily jump the play here as we see 13 "beat" by design right? 55 beat by being flat-footed and not covering the wide open passing lane...and we also see 81 covering the C which means the weakside D is unaccounted for in the backcheck. This is how fast IRL hockey moves at the upper-level. Thinking about the play up the ice is why you need to be so disciplined when executing FC's or traps.

So again, if everytime a puck doesnt get poked perfectly we're going to say "well you couldve been 1/2 a foot left, you couldve feather LS a bit more, etc. when answering "why did the offense get rewarded" then  I think it's rather fair and less than "ridiculous" to say "this FC or this trap only worked due to a mistake by the offense" no? It's imperfect game. Its an imperfect sport. You dont have to execute things to perfection for success, but that doesnt mean things are executed to perfection just because you had success. These are the same conversations we're having about game mechanics in other threads, I'm just applying it to hockey concepts.