"Learning Differential," and learning's % of OVR

[RougeRogue]

Having read this excellent post by Ebongreen sometime ago initially got me thinking about learning, and how unique a trait it is in this game -- after all, it seems to be the one trait that is, essentially, used and disposed of, as far as the player is concerned.

... Perhaps that needs some explanation: Once your starter masters your playbook (somewhat) thoroughly, and assuming that you're not adding dozens of plays each year, the learning stat seems to no longer have significance to that starting player -- or, as Ebongreen illustrated in his examples in the above, linked thread, there is a "Learning Differential" between, say, a 90 OVR WR with 99 learning, and a 90 OVR WR with 55 learning, as once both players have mastered your plays, which one would you rather have on your team?

Given that assumption, I've been working my teams with this philosophy on learning: Starters DON'T need as high learning (I shoot for 60-70 with my starters), while back-ups DO need high learning (I shoot for 85+ with my back-ups), so they might fill in as seamlessly as possible, picking up the playbook more quickly, until the starter returns. I do try to keep all my players with at least a 50 in their learning skills, as there is that in-game loading message, revealing how players with learning traits of less than 50 might never learn all of your plays....

But, all this raises the question: What % of the OVR is the learning trait? After all, knowing this would really quantify whether you'd be better off taking, say, that POT 94 with 96 learning, or whether you'd be even better served by taking the POT 88 with 60 learning (with other traits being close for each, but leaving you wondering which is truly the better candidate), even if the latter is picked at nearly the same time as the former, in your draft?

I think the best way to determine this % will be to compare players in one position that have nearly identical traits (except learning, obviously), and the best population to pull from will be the largest population in the game, Wide Receivers (there are 39 new WRs each year in the draft, just edging out the 36 new CBs each year), as the largest population will give the best chance to find close matches. Of course, matches may be used across years/draft classes, so long as the same grading philosophy is used (I'd be using "Balanced," for simplicity's sake).

Now, realizing that each position's (and, each grading philosophy's) OVR might have a different number of traits making up that OVR (and, thus, resulting in a slightly different % for learning's part of that OVR), can any of you see any short-comings to this test-method, before I really pursue it? Or, has this already been done, in case I'm re-inventing the wheel?

[ebongreen]

First, I haven't done the analysis. Despite a former life as a math major, I don't know enough about statistics to properly do a multi-variable regression test with curve-fitting. I'd love to collaborate with someone who does to reverse-engineer this and other mathematically-based aspects of this game.

Second, you'll need to be very careful what you're analyzing and what you're asking. OVR includes the production grade and LRN; I'm guessing that MaxPOT assumes 100 production and every developmental attribute (including LRN) maxed out. The Athleticism and Intangible grades may not be directly counted towards OVR - it may be attribute subsets.

It would be a good start to assume that OVR = f(ATH) + f(INT) + f(LRN) + f(Prod) and see what level of correlation you get. I wouldn't necessarily assume it's a linear fit - I know that the salary curves for different positions aren't.

Third, I'm not sure I agree with your understanding of Learning's importance in the game. Learning a play is one thing; mastering plays is something else, and retaining play knowledge is a third. Play mastery gives your players enhanced performance, and learning determines how quickly a player progresses in part because it affects performance.

My understanding of the feedback loop: learning -> play mastery -> better performance -> progression points (* percentage learning?) -> progression -> higher learning. Lower-learning players will have difficulty mastering all but the most frequently-run plays, won't retain the knowledge, won't perform as well, and won't improve their skills as quickly.

Special skills throw a wrinkle into this dynamic. There is a saying I've heard attributed to scouts and coaches: "You can't make chicken salad out of chicken$#!*." In HC09, you kinda can. Disciplined Routine, Off-Season Seminar, Regular Season Seminar, and Playbook Training all affect (and boost) a player's ability to learn and retain plays. What happens when your player with LRN 70 plays and acts like he has, oh, LRN 80? Good things - for you. Special skills overall make worse players perform like better players, so you can afford to draft and keep less-fabulous players for your squad.

What those skills may not do is increase his progression rate, if that progression is always fixed to his base learning score. If a LRN 70 player will only ever get 70% of the possible progress points he earns, it's very different from a (70+Disciplined Routine+Playbook Training)% of his earned progress, even if he performs exactly the same in-game.

tl;dr: I'm skeptical of how a set of starters with LRN 70 will perform both short- and long-term. Mid-80s (especially with special skills) should be good enough, but I think 70s is too low.

[RougeRogue]

Quote: Originally Posted by Dan_Bekins

Now for progression. .... There are a few things that affect how many progression points a player will receive for each action. Players with high learning abilities will receive more progression points. These guys will reach their potential very quickly. Players that report to a skilled coaching staff will receive more progression points. The point multiplier is actually compounded as you walk up the chain of coaches. For example, a running back reports to the RB coach, then the offensive coordinator, then the head coach. If all three coaches are highly skilled with running backs, he will progress extremely quickly. 2. Regression happens at the end of each season, and is based on a player's age, career phase, learning ability, and career health. Basically a player's ratings and max potentials will be lowered a bit once he gets toward the end of his career.

Ahhh, that is all good stuff to know -- I had initially thought (nay, hoped) that learning might affect progression, but discounted that possibility, given the progression I was seeing in even the less-learned on my team (thus, the basis for my assumptions). Good to see that I was wrong, and kudos should go to my coaching staff (built up over two seasons of making sure I hired the best coaches available), for pulling up even my most slow-witted players.

I've taken this post to heart, and will be switching my focus from developing 'physical development' to 'intangible development' in my coaching staff (at least, those that aren't already maxed in both), and will be interested to see if there is a noticeable boost, both in individual trait stats so associated, and in on-field performance. There's strong logic to the groupings, so I'd wager you're spot-on with them.

Finally, I'm always amused to see a number associated to a trait called "intangible."

[ebongreen]

As life would have it, I received the "Get Lots of Learning for Free" Gamechanger in my current career a week or two ago, so I'm able to quantify this a little bit better. In general, the game changer seems to add about 10% to LRN, up to the player's maximum.

It's interesting to get the game changer and realize that while the new skills show on the individual player's Progression chart, HC09 doesn't recalculate the OVR until another week's football has been played. This lag gave me the ability to tally before-and-after scores for LRN and (Balanced) OVR for players at multiple positions who hadn't already maxed out their LRN.

• QB (Jake Scott): Before 79 OVR, 82 LRN; after 81 OVR, 90 LRN. Diff +2/+8.
• HB (JR Myers): 83/86; 85/94. +2/+8.
• FB (Barone): 89/94; 90/96. +1/+2.
• FB: 84/86; 84/89. 0/+3.
• WR (Bellows): 91/97; 91/99. 0/+2.
• WR (DJ Crosby): 84/91; 85/95. +1/+4.
• WR: 74/63; 74/69. 0/+6.
• TE (Cousins): 84/89; 86/95. +2/+6.
• TE (Peterson): 60/32; 61/35. +1/+3.
• DT (Peko): 74/67; 75/73. +1/+6.
• OLB (Tarver): 88/91; 88/95. 0/+4.
• ILB (Bomani - moved): 87/86; 88/93. +1/+7.
• FS: 84/86; 86/92. +2/+6.
• SS: 80/85; 82/93. +2/+8.
• P: 82/85; 82/91. 0/+6.

I guarantee you that neither Jake Scott nor Domata Peko took any snaps in that week's game: both were inactive. That said, the results are, to put it mildly, all over the yard - but the direct correlation between LRN and OVR is pretty high. Generally, my guess is that (~4 points of LRN = ~1 point of OVR), and it may count differently for different positions (e.g. QB demands more learning than any other).

[Mike3207]

The learning development skills affect the speed at which your players reach their potential in learning. From what I've seen, the HC has the most influence on learning speed.

[RougeRogue]

Quote: Originally Posted by ebongreen

As life would have it, I received the "Get Lots of Learning for Free" Gamechanger in my current career a week or two ago, so I'm able to quantify this a little bit better. In general, the game changer seems to add about 10% to LRN, up to the player's maximum. It's interesting to get the game changer and realize that while the new skills show on the individual player's Progression chart, HC09 doesn't recalculate the OVR until another week's football has been played. This lag gave me the ability to tally before-and-after scores for LRN and (Balanced) OVR for players at multiple positions who hadn't already maxed out their LRN. QB (Jake Scott): Before 79 OVR, 82 LRN; after 81 OVR, 90 LRN. Diff +2/+8. HB (JR Myers): 83/86; 85/94. +2/+8. FB (Barone): 89/94; 90/96. +1/+2. FB: 84/86; 84/89. 0/+3. WR (Bellows): 91/97; 91/99. 0/+2. WR (DJ Crosby): 84/91; 85/95. +1/+4. WR: 74/63; 74/69. 0/+6. TE (Cousins): 84/89; 86/95. +2/+6. TE (Peterson): 60/32; 61/35. +1/+3. DT (Peko): 74/67; 75/73. +1/+6. OLB (Tarver): 88/91; 88/95. 0/+4. ILB (Bomani - moved): 87/86; 88/93. +1/+7. FS: 84/86; 86/92. +2/+6. SS: 80/85; 82/93. +2/+8. P: 82/85; 82/91. 0/+6. I guarantee you that neither Jake Scott nor Domata Peko took any snaps in that week's game: both were inactive. That said, the results are, to put it mildly, all over the yard - but the direct correlation between LRN and OVR is pretty high. Generally, my guess is that (~4 points of LRN = ~1 point of OVR), and it may count differently for different positions (e.g. QB demands more learning than any other).

Great analysis -- this is exactly what I had been waiting for in my franchise, as it had happened once before in a previous one, but I hadn't the test in mind to make use of it, at the time. It looks like you're spot-on with your approximations (~4 to ~1), with variances likely attributable to rounding. ANOVAs and regressions be damned, when we have the cheese-factor of a "game-changer!"

By the way, my GM has an old bag of used footballs laying around he'd be willing to trade for your TE Peterson. If he's not your ace-in-the-hole, blocking tight end, have your GM's office call my GM, if you're interested....

[ebongreen]

Quote: Originally Posted by RougeRogue

By the way, my GM has an old bag of used footballs laying around he'd be willing to trade for your TE Peterson. If he's not your ace-in-the-hole, blocking tight end, have your GM's office call my GM, if you're interested....

Ah, if only. I just finished the English 2011 draft and acquired Storm Stewart to party with Cousins and Drew Goldman, so Mike just got cut in the preseason. He was too fragile and I had other smarter better-blocking tight ends to play with, so I never did manage to develop him worth a darn. Now my TE group is ready to par-tay!