Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

[DeuceDouglas]

How Players Regress in Madden 17

There was another post over at DeepDiveGaming that was a continuation of his look into player development but this time took a look at regression and how it will effect your players based on position and development traits. It has tables for, I believe, every position and development trait showing you which attributes will regress and by how much over time. Very interesting stuff.

[Hooe]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

Great find!

To provide some high-level bullet points on Madden 17 player regression from the article:

- all positions start regressing by age 29, regardless of development trait
- development trait affects regression; "Normal" and "Slow" players can start regressing at 28 and generally speaking lose more total points in a given regression year than players with "Quick" or "Superstar" development
- a single attribute will never regress more than -2 in the offseason regression step (-2 is by far the most common penalty)
- a single attribute will rarely regress -1 in the offseason regression step

The biggest takeaway I have from the article is that, for the purposes of predicting an individual player's career arc in Madden NFL 17 Franchise, the development trait is probably the single most important attribute a player has.

In the future, it might be cool if Development and Regression were separate characteristics; maybe a player has Superstar development but "Slow" regression (re-labeled as "Bust"), so he peaks quickly but also falls out of the league quickly once his peak is over? I also think that, if I could personally tweak what currently exists, I'd make the first year of SPD regression -1 for every position and development level rather than -2, but that's not really a major pain point for me personally.

[BleedGreen710]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

In the future, it might be cool if Development and Regression were separate characteristics;

yes, I would love this! or, if they had a regression slider just like the xp slider.

[Tomgun]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

Great find!

To provide some high-level bullet points on Madden 17 player regression from the article:

- all positions start regressing by age 29, regardless of development trait
- development trait affects regression; "Normal" and "Slow" players can start regressing at 28 and generally speaking lose more total points in a given regression year than players with "Quick" or "Superstar" development
- a single attribute will never regress more than -2 in the offseason regression step (-2 is by far the most common penalty)
- a single attribute will rarely regress -1 in the offseason regression step

The biggest takeaway I have from the article is that, for the purposes of predicting an individual player's career arc in Madden NFL 17 Franchise, the development trait is probably the single most important attribute a player has.

In the future, it might be cool if Development and Regression were separate characteristics; maybe a player has Superstar development but "Slow" regression (re-labeled as "Bust"), so he peaks quickly but also falls out of the league quickly once his peak is over? I also think that, if I could personally tweak what currently exists, I'd make the first year of SPD regression -1 for every position and development level rather than -2, but that's not really a major pain point for me personally.

If that was the case and said player had "slow regression" this would essentially keep him at his peak the most and not being a bust but a star.. having "fast regression" would essentially make them regress at a fast pace and therefore being labeled a bust due the fast decline.

[mlb61]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

Great find!

To provide some high-level bullet points on Madden 17 player regression from the article:

- all positions start regressing by age 29, regardless of development trait
- development trait affects regression; "Normal" and "Slow" players can start regressing at 28 and generally speaking lose more total points in a given regression year than players with "Quick" or "Superstar" development
- a single attribute will never regress more than -2 in the offseason regression step (-2 is by far the most common penalty)
- a single attribute will rarely regress -1 in the offseason regression step

The biggest takeaway I have from the article is that, for the purposes of predicting an individual player's career arc in Madden NFL 17 Franchise, the development trait is probably the single most important attribute a player has.

In the future, it might be cool if Development and Regression were separate characteristics; maybe a player has Superstar development but "Slow" regression (re-labeled as "Bust"), so he peaks quickly but also falls out of the league quickly once his peak is over? I also think that, if I could personally tweak what currently exists, I'd make the first year of SPD regression -1 for every position and development level rather than -2, but that's not really a major pain point for me personally.

In the future, it might be cool if Development and Regression were based on some kind of real world performance data.

[Mattanite]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

Was this article done pre or post patch? Noticed in the patch notes they tuned QB regression in some way...

[Ueauvan]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

i still dislike player targets for non starters, 75 catches for wr#5 and #6??? 40 tackles for mlb#4, 10 sacks for lolb#3, 24 tds for qb#3.
they need to be dynamic and responsive. otherwise evertone regresses even though they haven't had a chance to achieve level 3/4/5

[timhere1970]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

i still dislike player targets for non starters, 75 catches for wr#5 and #6??? 40 tackles for mlb#4, 10 sacks for lolb#3, 24 tds for qb#3.
they need to be dynamic and responsive. otherwise evertone regresses even though they haven't had a chance to achieve level 3/4/5

I am with you. Targets should be realistic based on depth chart. Level two should be a realistic expectation.

[Essex Blue]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

i still dislike player targets for non starters, 75 catches for wr#5 and #6??? 40 tackles for mlb#4, 10 sacks for lolb#3, 24 tds for qb#3.
they need to be dynamic and responsive. otherwise evertone regresses even though they haven't had a chance to achieve level 3/4/5

Yeah this.... also am I right in thinking that the don't take into account match length? So if you're not playing 15 minute quarters your players will regress more quickly?

My other main gripe with this is that QBs shouldn't regress like this so early.A receiver peaking at 27/28 makes sense. A QB's physical attributes may drop off in their early 30s but most are still improving their accuracy and decision making up until 35/36/37

[timhere1970]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

This is why I draft physical freaks, they have to start out that way to be usable in their thirties. I am at this point trading my 1st, 2nd and third round picks to go after physical players to develop. This will also free up salary cap space to resign players and prevent my rookies from getting high overalls so quick that they are never resignable. When I counted 9 99 overall players on my team and have let 10 99 overalls go to free agency in the last few seasons. It finally dawned on me to make it a challenge. Back to regression, a hb that is super fast can be upgraded in strength and truck to be effective into their thirties. An 88 speed rb becomes useless in their thirties. Same for just about every position. I like to hold onto my star players, I want them to retire a steelers and go into the hall of fame.

[Ueauvan]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

yep mins per qtr doesnt change target. in reality my mlb 2 doesnt have a target this year he was injured until week 2 and he has zero incentive to tackle in the short term
in my opinion back ups should have generic team targets; win, beat rival etc unless they play more than a few snaps and they get real generated target for that game not season unless injury makes them a starter. thats my sort of logic

[Hooe]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

My other main gripe with this is that QBs shouldn't regress like this so early.A receiver peaking at 27/28 makes sense. A QB's physical attributes may drop off in their early 30s but most are still improving their accuracy and decision making up until 35/36/37

Quarterback regression in the NFL is weird.

Advanced Football Analytics did a study on it and found that, on average, quarterbacks start regressing at age 29, which would point to the regression setup in Madden NFL being basically correct.

However, AFA also found that if the final season of quarterbacks' careers were not considered (i.e. to evaluate how Peyton Manning regressed, we'd ignore his dismal 2015 season), quarterbacks don't appear to regress at all, statistically speaking. Basically, what they see is that quarterbacks basically have "it" until they suddenly wake up one morning and don't have "it" anymore. Once they don't have "it", they basically fall entirely out of the league and don't play much ever again. Look at Matt Schaub, for example; his career imploded in Houston in 2013, and he hasn't been a full-time starter since.

The question falling out of this, then - would we franchise types prefer that Madden continue to emulate the more gradual decline for quarterbacks (that gradual decline Tiburon appears to have nailed dead-on the money based on the information from this article), or would we prefer that quarterbacks just regress very sharply later in their careers (I'm talking like 10+ points OVR decrease if the Level 4 season goal is missed)?

[sebennayx]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

Quarterback regression in the NFL is weird.

Advanced Football Analytics did a study on it and found that, on average, quarterbacks start regressing at age 29, which would point to the regression setup in Madden NFL being basically correct.

However, AFA also found that if the final season of quarterbacks' careers were not considered (i.e. to evaluate how Peyton Manning regressed, we'd ignore his dismal 2015 season), quarterbacks don't appear to regress at all, statistically speaking. Basically, what they see is that quarterbacks basically have "it" until they suddenly wake up one morning and don't have "it" anymore. Once they don't have "it", they basically fall entirely out of the league and don't play much ever again. Look at Matt Schaub, for example; his career imploded in Houston in 2013, and he hasn't been a full-time starter since.

The question falling out of this, then - would we franchise types prefer that Madden continue to emulate the more gradual decline for quarterbacks (that gradual decline Tiburon appears to have nailed dead-on the money based on the information from this article), or would we prefer that quarterbacks just regress very sharply later in their careers (I'm talking like 10+ points OVR decrease if the Level 4 season goal is missed)?

Both, mix it up...maybe even from year to year with a random progression/regression based on their age and maybe even snap count:

Under 26 could have a high likelihood of a general progression, with a small chance of a major boost (higher chance with Quick/SS Dev).

Players aged 29+ could have a higher likelyhood of regression, but still a chance at progression - and even a minor chance at a major progression/regression.

Significantly older players, significantly higher likelihood of a major dropoff during the offseason.

Roll the dice when advancing into the offseason so those who want Brady to play into his 60s can restart from a save before the offseason and try again if he retires/regresses.

I'd prefer no matter how they do it, either way that they hide the regression, and their ratings in general, entirely. If the Broncos knew Peyton dropped 10pts before last year, would they have kept him, let alone started him on their way to the SB?

There's too much visibility as it is when it comes to ratings -- when drafting a player, or during the season, or with regression, etc...it's embarrassing to think how far from real life player management is in Madden.

[Hooe]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

There's too much visibility as it is when it comes to ratings -- when drafting a player, or during the season, or with regression, etc...it's embarrassing to think how far from real life player management is in Madden.

Slightly off-topic, but: real-life NFL front offices absolutely do quantify player measurables, skills, and intangibles with numeric grades in their scouting reports. They do this for both for incoming rookies and rostered players already in the league. By measuring the change in a player's preseason scouting report grades year-over-year, a front office's scouting department can determine if a player is progressing or regressing.

That pro teams routinely perform this quantification of players - they even give players overall grades - was the entire basis of the now-defunct FBGratings project, which used quantified primary source data from a real-life professional football scouting department to determine player ratings.

Surfacing player ratings and how much they change year-over-year is not unrealistic whatsoever.

[sebennayx]

Re: Madden 17 Deep Dive - How Players Regress (DeepDiveGaming)

Slightly off-topic, but: real-life NFL front offices absolutely do quantify player measurables, skills, and intangibles with numeric grades in their scouting reports. They do this for both for incoming rookies and rostered players already in the league. By measuring the change in a player's preseason scouting report grades year-over-year, a front office's scouting department can determine if a player is progressing or regressing.

That pro teams routinely perform this quantification of players - they even give players overall grades - was the entire basis of the now-defunct FBGratings project, which used quantified primary source data from a real-life professional football scouting department to determine player ratings.

Surfacing player ratings and how much they change year-over-year is not unrealistic whatsoever.

Fair point...I don't mean to continue taking it off-topic, I'll try to steer it back to regression.

I think the big difference is IRL the quantification (especially on intangibles) is subjective, with TONS of room for error. In Madden, they are static - the rating/trait is what it tells you it is, and it's fully visible.

I wouldn't mind seeing ratings/traits so much if they weren't as definite or clearly visible as they exist in Madden. As it relates to the original topic of regression -- the expectation IRL is that Steve Smith would have regressed this year what with injury and age; if regression was hidden and ratings were "blurred", I would expect that his ratings would drop somewhere. If you remove that unknown, and the Coach/Owner was able to see that Steve Smith's hasn't lost a step before he even put on pads (in Madden terms, regression did not apply to him during the offseason), it doesn't provide the opportunity for that difficult decision of keeping/releasing him.