Breaking The Demo (VIDEO)

[jmaj315]

Quote: Originally Posted by majesty95

Verticals | Nickel - Under Storm Verticals My first objective was to test out one of the most exploitable pass plays the past few years, Veritcals. More specifically, Four Verticals. I ran with Ohio State and played from the shotgun the entire game. I mostly ran Four Verticals but also mixed in some standard Verticals. Every offensive play was either Verticals or Four Verticals. Analysis AI Adaptation - While I was still able to complete a higher percentage of these passes than I would have liked, I did notice the CPU taking away routes that I was favoring. I would have liked to seem them recognize I was running this play over and over and take it away more consistently but overall I was glad to see the AI react more intelligently. Nickel - Under Storm On defense, I ran Nickel - Under Storm on every play. I did not very this strategy at all. Every play, every formation and every scenario saw the same play from me. Analysis Good and Bad - I liked that the AI noticed I was blitzing a lot and called more HB Screens than I had seen before and I like that the QB primarily targeted his HB and TE. I also liked that I was beat for several big plays and that there was a risk/reward for running this play. However, I sacked the Michigan QB seven times and held them to 13 points in regulation. I gave up big play and produced big plays, which I liked. However, the CPU did not adjust to the fact that I was only calling this play and force me to change. Due to this I saw more success overall than I would have liked. Conclusion Had I been running a more diverse offense I think that I could have easily won this game playing Under Storm the entire game. I believe that's an issue. Sure, some leagues and players have "house rules" that will prevent this type of playcalling. However, things like this should and could be prevented by a little bit more adaptive AI. Recognizing that a player is doing the same thing over and over and how to exploit it should be easy to implement and is a basic football strategy. I certainly don't want the AI to be "psychic" but I would like it to adjust when I do the same thing over and over.

Most of the sacks the QB just didnt read the D properly and tried to scramble instead of throw to the check down. Im 2 minutes in and every sack so far he tucks the ball instead of throwing to HB or TE :/

[majesty95]

Read Option | Nickel - Cover 3



Read Option

I had noticed when playing the demo originally that the read option (or zone read if you want to refer to it correctly) seemed to be very successful. I wanted to test this out by running it over and over and see what my results were.

I ran the Read Option every play up until the last minute or so when I was down and went back to verticals. I would just pick Read Option from a certain formation and then no huddle and scroll to the next formation, pick it again. Wash, rinse, repeat.

Analysis

Unstoppable? - My fears were confirmed very early in this game. It is next to impossible for the CPU AI to stop the read option consistently on default All-American. I did not throw one pass until late in the 4th and still put up over 250 yards on 3 min quarters. While the defense stuffed me a couple of times, overall it was fairly easy to run the ball.

I think the read indicators made it that much easier and I would like to see them removed from AA and Heisman (I heard they were so maybe this is just in the demo). It makes it much too easy to identify the proper read and will therefore lead to too many players having too much success with it, IMO.

Nickel - Cover 3

I tested this as I had heard that playing a base Cover 3 all game in NCAA 13 was an effective strategy and wanted to test it myself in the demo. I was also warned to expect QBs to take off on you this year so I made a small adjustment of spying one of the hook (yellow) zones after the first few plays. I alternated using press and off coverage and didn't notice either being more overly effective than the other.

Analysis

I would say that running a base Cover 3 the entire game will yield reasonable results. I did give up 263 yards on 3 minutes quarters so I can't recommend this strategy. However, it certainly looks like the DBs cover better so its a much more effective strategy than I remember. The drawback is that it has holes and even the CPU QB will find them with consistency.

Conclusion

I was much more satisfied with the way A&M exploited the Cover 3 than I was with Michigan and the Under Storm. QB #2 seemed to find the holes routinely and they ran the ball with enough success that I wouldn't have wanted to stay in it for an entire game.

Offensively, I wasn't as pleased. The CPU does not recognize your playcalling patterns as well as it should when calling its plays and the CPU players do not recognize them as well either. I'm quite concerned at how overpowered this offense could become in both online play and in online dynasties. Again, I'm hoping the AI was tuned after putting together the demo but, if not, they need to look into that just a bit.

EDIT: Let me add that Oregon made some BIG hits. The most I've seen when playing so far. Their return game is also MUCH more effective than anyone else's that I have played with. There are definitely some interesting returns and big hits.

[jmaj315]

Quote: Originally Posted by Cardot

LOL @ 6:10 "And there isn't anyone who is going to catch this corner back". 2 Seconds later he is pushed out of bounds 60 yards away from a touchdown.

LOL yea that was funny
but how did that ball get caught at 6:00? wow

And regarding the thoughts on the AI DBs... Michigan doesn't exactly have the greatest secondary. Even in "good" years (in the past 4 years) they've been sketchy.

[BA2929]

Your title misled me. I came in here figuring to see a video of you doing some wacky things and having the demo hilariously glitch out on you.

[majesty95]

It occurred to me that I may encounter that. Just haven't figured out how to do it yet.

[khaliib]

It really does no good to do this testing through Play Now games as most of the AI logic programming and animations are geared to function through the Dynasty Mode.

Before any tester goes this route to say what is/isn't a glitch or is broke, they need to do an "Install" of the game and look through each Data file to get some type of understanding on how things work.

Play Now games do not use the entire AI programming/animations which is why many of these so called errors occur.

There are literally hundreds, if not thousands of animations that are pieced together with certain measurable's establish to trigger them in a Dynasty to give the perception of an AI variance.

I could go on, but it would require pages to break down the Raw Data in the Data files of the game.

If you find something through testing it in Dynasty Mode (which would be a huge task because it must be tested against so many the different variables [that are in the programming] to find out if an occurance is due to bad code/programming)

Simply saying something is in error without providing a method of rectifying within the current programming does little to nothing due to fact of how everything is linked together.

Simply adjusting a triggering value up/down can and/or will impact the next aspect of programming linked a little further down the chain.
(as to why Tuners seemed to always affect some other aspect of the game that didn't need adjusting)

When I looked through the entire 6+ GB Data Files, I finally understood it's not as easy as 1-2-3, change this/that or tweak and tune here/there on the current foundation the game is built on.

Just wanted to share this info so that you can see it in it's entire scope as you try to help improve the game through your testing.

[majesty95]

Interesting. Thanks for the insight.

Let me ask you this: would something like the CPU hurrying to the line in the 4th quarter and then killing the clock not be something that affects both play now and dynasty? How about the zig zag running by the CPU, the lack of effectiveness shutting down the read option and the overall AI in regards to recognizing your play calling tendencies?

I guess what I'm asking is why is the gameplay experience so much different for play now and dynasty and how much of a correlation is there?

Also, I'm assuming that testing sliders in play now for dynasty use is pointless for the most part? It's crazy that there is so much of a difference, especially with how big online user vs user play is.

[khaliib]

Quote: Originally Posted by majesty95

Interesting. Thanks for the insight. Let me ask you this: would something like the CPU hurrying to the line in the 4th quarter and then killing the clock not be something that affects both play now and dynasty? The Demo is about 1.8 GB while the entire game is 6+ GB. That's a huge chunk of AI programming/logic missing. Think of it this way, Fatigue is there, but we don't see it in the 3 min we are given because it is programmed to be seen later in coded logic that says (ex "A's 100% animation rate will decrease to 80% animation rate [based on this predefined measurable] and so on, so forth). There's logic for hurrying to the line and running this play in this circumstance (who knows how many variables they've coded just for 4th down). There's yardage, time, score etc... that the "hurry-to-the-line" logic must be presented against and with 1.8 GB, not all variables can be included in a Demo so we can see the affect (similar to Fatigue example). As far as Play Now vs Dynasty. I would say that the basics of the logic/programming is given to Play Now with the understanding that the Hum gamer will take control of the player at some point during a play, voiding AI logic. Play now is not geared towards User vs AI, but User vs User. So put in place the bare bones for even User controlled team AI with the anticipation that the User may/will take control at any given point. How about the zig zag running by the CPU, Game Speed + Player Speed rating [0-127] dictate how fast/slow animations are driven. Put Game Speed on Very Slow and it looks/feels odd because our eyes are used to Madden and NCAA rendering fast animations during gameplay. We are given 3 min qtrs in Play Now, so most of the programming/logic that trigger animations are not there because of the Demo size. Speed ratings over 70 simply push animations to render too fast over the ratio of the field size. (this is calmed greatly with Coord Cam, which is why players look to render/move in a more fluid maner) We don't have access to player's Speed rating, but because Player Speeds are always on the high end of the scale and the Game speed most like to play at, there is still some zig-zag rendering instead of a fluid running motion. the lack of effectiveness shutting down the read option and the overall AI in regards to recognizing your play calling tendencies? To give you and others some hope in this area, there are separate/additional Awareness fields that will improve the AI across the board, that we must use Xan's editor to get to. Don't know the affect for Online gaming as I don't play online. I guess what I'm asking is why is the gameplay experience so much different for play now and dynasty and how much of a correlation is there? The quick and dirty is in the first very short explanations Also, I'm assuming that testing sliders in play now for dynasty use is pointless for the most part? If you look at most of the YouTube Breakdown vid's that folks have done over the years, most arrived at their conclusions from testing in Play Now and Practice Modes when the greatest experience of what the AI can/will do, is through Dynasty Mode.

This is not to say whether someones testing or what they see is right or wrong, but that there are a lot of variables that are often times left out as to why "this issue" may be happening. (ie so we say there's a bug, a glitch, it's broke, they don't know what they're doing etc...)
Much of it has to do with that there is so much that must be linked together just for one given instance, as well as a manually input value that helps to trigger into the next "whatever"....

The last thing I want to do is come off like some grand wizard of game programming/developement (far from it) but all I can say is that there is a ton of animations that programming logic must link together for the many aspects of this game play out right.

Much of what we want is pretty much there, it's just a very large amount of data/animations that must come together with very little room for error or miscalculation.

I couldn't imagine what the entire game would look like uncompressed.
It's beyond a lot to bring together what we (me included) often think is such an easy thing to do.

Look into the files, it will give you a better understanding on how to accomplish what you want to happen during gameplay with what's there.