Why Balancing for "Realistic Stats" is a Bad Goal

[Aestis]

OK, so it's a 20% click-baity title, but I wanted to put this out there as it's something I've observed for a few years now.

All over the Madden forums here, I see "realistic stats" or "unrealistic stats" used as the consensus criterion for determining whether a slider set, or a game mode, or an in-game ability, or what-have-you aligns ballpark with the NFL and thus is good gameplay.

This is a fool's gold way to balance your gameplay for a variety of reasons.

"Stats" is a really broad term already. What do you mean by that? Score? Raw yardage? If you gave me the ability to modify game length, I could make it so NFL Blitz ends up with realistic end-game stats. While that's a silly extreme, Madden operates under the same exact concept. I'll see guys cite "realistic stats" because they threw for 265 yds, and that's only slightly above an avg NFL game... except they did it on 13 pass completions. Whether that's good or bad--it's certainly not the kind of thing you hardly ever see in the NFL. Actual gameplay in that scenario is likely not remotely realistic, even though the endgame yardage & scoring totals look nice in a limited box score.

Conversely, and this is in many ways worse--I'll see "unrealistic stats" cited as a reason something is bad. As an example, in a Madden Discord, once had someone who was furious about the high sack totals. Even more furious when I said sky-high sack totals are going to be a result of realistic gameplay, for many Madden players. In a typical NFL game, any shaky OL (and these days that describes half the league) is going to see the QB under pressure *instantly* a few times/game. Sometimes more. But the QB has anticipated and either *instantly* checking down, or he hasn't anticipated but sees it *instantly* (since these giant men are right in front of him), and finds a way to take a step back & chuck the ball into the dirt or out of bounds. These plays only result in a sack when the QB doesn't anticipate the rush and can't do anything about it.

But Madden players? I've watched legit hundreds of Madden sim CFM games now from probably 100 diff players across our league over the past few years across a huge skill spectrum. Madden players look downfield. They rarely pre-snap call for extra blockers vs a potential blitzer or hit their hot read in the face of instant pressure. MANY plays that in the NFL would be an effectively broken play where the QB gets the ball out instantly just to avoid a sack, Madden players are taking that near-instant sack. Which, of course, feels awful and feels unrealistic as hell in the moment. Sacks are just one example of many, but the way most Madden players play Madden--even the guys who self-identify as hardcore simulation players and refuse to entertain anything to the contrary--is often not at all in line with NFL QB decision-making, and thus certain outcomes (e.g. sacks) should be happening far more often because players are making certain decisions far more often (e.g. waiting for deeper downfield routes even in the face of a vicious pass rush) than an NFL QB would.

In the above scenario, if you tried to balance the game so that sack #s were realistic... what you'd actually be doing is reducing pass rush efficacy enough that players could stick in pocket way longer than an NFL QB could, thus throwing downfield way more often & more successfully than an NFL QB could. Man coverage becomes far less effective. Blitzing becomes more necessary, leading to more huge RAC gains. There's a whole domino-effect that comes from balancing for a "realistic stat" instead of holistic gameplay risk/reward or success/outcome %s.


tldr; if you balance for the most NFL-like gameplay, you won't get NFL stats. The most realistic gameplay might have some stats that are way out of whack. Sometimes that's legitimately a Madden issue. Much of the time it's a human player issue.

[KingKojoO]

Amen. Totally agree.

[PunkMonk]

Very good points.

Do you have any recommended solutions, settings, sliders, etc?

[Matt10]

Well said. This is why my approach has always been to get animations right and remain contextual to those animations. If the stats balance out, so be it, but I take it with a grain of salt as it can heavily be skewed. One of the main reasons I rarely post box scores.

I do pay attention to stats that are relative to the franchise I'm playing - not the real life NFL. Facing a 50 OVR QB who has 90% comp all year? Do you expect him to all of a sudden fall to a 58 comp % just because he's facing you? Not likely. However, the common approach to stats is a logical confirmation of ratings. The 50 OVR QB should be at 40%, or lower, comp because his ratings are low. I think this is limited thinking.

This is why we run into the issue of true progression. If a player is a 50 OVR QB, and he should play like that each time, then how would he progress at all? Is he allowed to reach a different potential outside of his ratings at all? Not if stats chasers have anything to do with it.

[Aestis]

Quote: Originally Posted by PunkMonk

Very good points. Do you have any recommended solutions, settings, sliders, etc?

For potential solutions, anyone who follows my User v User thread knows my sliders are extremely data-driven, which is why it may seem strange that I'm very anti "stats." We can be smarter about how data informs sliders. I focus on risk v reward and efficiency/outcomes.

What I mean by that is, say you're evaluating the passing game. Beginning with pass yardage gets you nowhere, per my OP here. I try to align Risk v Reward as best I can to the NFL, so that whatever decisions a player makes will at least be made under realistic-ish circumstances.

Example: pass rush should be fast enough that QBs have to get the ball out quick, accuracy low enough that offenses aren't overpowered, coverage strong enough that WRs aren't wide open all game, and WRs need to still be able to sometimes catch the ball in these tighter coverage situations.

All these things work in concert with the passing game, but you can still use a general passing efficiency metric to determine if the overall game is functioning. I use Net Pass Yds per Dropback. This accounts for sack %, negative 'pass' yardage from sacks, comp %, and depth of passing all in one pretty good pass efficiency metric. It basically represents the overall Risk v Reward of passing. You can RISK sitting in the pocket & taking a lot more sacks in order to gain the REWARD of completing more deep passes, though you'll complete a lot lower comp %... but your Net Pass Yds per Dropback should roughly still align with NFL. So NFL QBs generally don't do all that at the rate of a Madden player, but if they did... the above is what it would look like.

Thus the statistics (passing yardage, comp %, etc.) are just the output. You're NOT balancing your games to force a 62.5% completion rate onto human player decisions that absolutely should not result in a 62.5% completion rate.

Where art meets science is that the data will tell you when the risk/reward is unbalanced, and it may give you an idea as to why (perhaps comp% is above NFL averages while your downfield air yards are way above NFL levels... that should not happen, aggressive downfield passing should never result in a higher comp%). But data won't always tell you what to do about it. That's where you do need to be watching the gameplay to align what you see with what the data is telling you.


Other guys will start with animations or more broadly player interactions. In theory if you set everything up perfectly, the right gameplay risk/reward and outcomes will follow. It's tricky and I still think those who ignore any measurement entirely are sticking their head in the sand almost as much as those who balance toward one or two raw stats.

Not everyone is going to have access to many games' worth of data, so in that case you may be at the mercy of what you can realistically do. Mostly I just wanted to bring up the idea that 'realistic stats,' even coming from the extremely data-driven perspective that I have, is a terrible thing to balance for.

[Jagsfan24]

That was a well written post, I have gotten better in the previous years madden and certainly this years when it comes to pre snap reading and all that. I still complain some, but in the end it’s a game and can’t get all what you want.

[tommycoa]

Have to disagree with this. Yes, getting 700 yards of total offense is not what you want, but getting the stats to even up to 700 yards in 16 games is what you want. A wide delta of stats and outcomes. Not worrying about stats is a defeatist mentality.

You can’t get real averages, so you just quit trying. Which i understand when it comes to hum/cpu action, that’s hum/cpu is impossible to get sim stuff. Only thing that matters is getting the cpu to a competitive level. That’s why cpu/cpu is the only way you can get animations and sim stats. And with the animations, you see way more in cpu/cpu mode. I see people talk about seeing an animation on here and being mindblown, but with cpu/cpu, we see awesome animation every play it seems.

[cdcool]

This is why I always say it's in the realm of NFL possibilities.