+- International Simulation Football League (https://forums.sim-football.com)
+-- Forum: Community (https://forums.sim-football.com/forumdisplay.php?fid=5)
+--- Forum: Media (https://forums.sim-football.com/forumdisplay.php?fid=37)
+---- Forum: Graded Articles (https://forums.sim-football.com/forumdisplay.php?fid=38)
+---- Thread: *Proposal for Changes to Player Regression (/showthread.php?tid=44083)
Pages:
1
2
*Proposal for Changes to Player Regression - aeonsjenni - 03-09-2023
Hi everyone, today I'd like to take a brief dive into the current regression system, make some comments about it, and propose a couple alternatives.
Defining Our Max Earner
I'm going to looking at a lot of this from the perspective of a max-earning player, partially because that's how I intend to earn, and partially because mathematically, the amount of TPE a player has is much easier to project if they max-earn. For the sake of simplicity, I'm going to assume that a max-earning player earns up to a peak of 1450 TPE, which I'm aware is far from the highest players have gotten to, but I believe this is a safe estimate. Additionally I'm going to assume a max-earning player earns 177 TPE per season, which does not account for one-off point tasks and fluctuations in prediction success. If anyone feels that different per-year or peak numbers would be more accurate, I would appreciate hearing it, as I intentionally did not spend too much time calculating those figures—they are good enough, though.
Current Regression System
Just for reference, the current regression system works as follows:
After 7th season 20% TPE loss
After 8th season 25% TPE loss
After 9th season 30% TPE loss
After 10th season 40% TPE loss
After 11th season 50% TPE loss
After 12th season, and thereafter, 75% TPE loss
15 season maximum, forced retirement if regressed below 150 TPE
One important thing to note is that while the percentage rate of regression increases quickly, the raw amount of TPE lost to regression will be lower if you have less TPE before regression. If we imagine our hypothetical player goes to the end of their career, they would lose 290 TPE after the 7th season, then 334, 354, 401, 389, 425, 239 TPE after each subsequent season until forced retirement. In this case, our player would actually be at 142 TPE after their 12th season, and would not be able to play in their 13th season. You'll notice that the raw regression numbers generally stay comfortably in the 300-400 range. I imagine this is by design, to cause max-earning players to gradually and consistently regress by ~350 TPE per year until they are forced to retire. My understanding is that many users outright retire shortly after regressing, and I can see why. Players will lose a hefty chunk of TPE, gain about half of it back, and immediately lose another 350 TPE again. It's a little depressing, but to a certain extent, it should be! Football is a demanding sport, and players can only keep up with it for so long. I understand that. However, my feeling is that the current model is not reflective of the general arc of a successful NFL career, nor is it particularly fun, so I see no reason not to look to replace it.
Based on minimal research, the average successful long-term NFL career (barring major injury) approximately follows the following arc:
- Steady improvement as a young player
- 2-4 year peak
- Sharp drop-off immediately after peak
- Steady decline during the twilight of career
- Sharp decline as the player's body fails them
- Retirement
Looking at our hypothetical player's TPE as they regress, we get this chart:
![[Image: Career_Regression_TPE.png]](https://cdn.discordapp.com/attachments/1072083825837998090/1083274218885632090/Career_Regression_TPE.png)
You can see that this regression is all too steady, as I discussed earlier. So in my opinion the current regression system accurately simulates stages 1 & 2 of a career arc, but I feel as though the immediate post-peak decline could be a bit sharper, and the decline afterwards could be a bit gentler. The other issue of course is that while the rules allow for a 15 season career, it is not possible to actually achieve this. Only 13 seasons are mathematically allowed by these rules, and even that requires going a bit ahead of how this player is earning.
Two Proposals
I want to submit two proposals, the first of which will be more conservative, but allow for a 15-year maximum career, and another which will be more radical, allowing for a 17-year maximum career. The first proposed regression model would look like this:
After 7th season 20% TPE loss
After 8th season 25% TPE loss
After 9th season 35% TPE loss
After 10th season 35% TPE loss
After 11th season 35% TPE loss
After 12th season 35% TPE loss
After 13th season 50% TPE loss
After 14th season 60% TPE loss
After 15th season 75% TPE loss (irrelevant, retirement is forced regardless of TPE)
The idea with this regression is that the decline in their 3rd season is just a bit stronger, which I hope would emulate the sharper decline I am looking for after the peak, while the decline afterwards is much shallower, remaining at 35% until the player is more or less forced out of a starting position in their last years. Thus a player could enjoy their peak, quickly drop down to low-end starter TPE (depending on position but that is for another article), and slowly decline but ultimately still contribute for some time before retiring. I'm a big fan of this model, and I would love for it, or some mild variant, to be considered as a reasonable replacement.
The second proposal, which is a fair bit more radical, looks like this:
After 7th season 20% TPE loss
After 8th season 30% TPE loss
After 9th season 35% TPE loss
After 10th season 40% TPE loss
After 11th season 30% TPE loss
After 12th season 30% TPE loss
After 13th season 35% TPE loss
After 14th season 40% TPE loss
After 15th season 50% TPE loss
After 16th season 60% TPE loss
After 17th season 75% TPE loss (irrelevant, retirement is forced regardless of TPE)
With this regression the immediate drop-off is even steeper, but then quickly mellows out, letting our player hang around 400-600 TPE for a few seasons. This is my attempt at extending a career as long as 17 seasons, but frankly I am not particularly happy with it. I don't believe that it would be very fun to lose and regain the same TPE 4 times over, but I wanted to look at this model just as a starting point for what kinds of things we could look at if we wanted to allow for exceptionally long careers.
Between the three models I've gone over, the TPE of a max-earner in each model would look roughly like this:
![[Image: Career_Regression_TPE_1.png]](https://cdn.discordapp.com/attachments/1072083825837998090/1083282048233504808/Career_Regression_TPE_1.png)
And the raw regression would look like this:
![[Image: Raw_Regression.png]](https://cdn.discordapp.com/attachments/1072083825837998090/1083282576376074301/Raw_Regression.png)
For reference, the three models side-by-side look like this:
![[Image: image.png]](https://cdn.discordapp.com/attachments/1072083825837998090/1083286338763239454/image.png)
Why Switch?
To close things out I want to investigate why I think switching to a new system might be worth it, and why it might not. When I first joined the league a little over a month ago, my goal was more or less to ride out regression until my player was forced to retire. I have a great fondness for players that stick around even after they've declined, and I wanted to do that for my player as well. I was a little disappointed to find out that the longest a player could stick around would be only 13 seasons. That just doesn't seem like a very long time to me! A potential downside to a change is that it would mess with long-standing career records. This, to me, is not so significant, probably because I am a new user and I don't hold any career records. However, I'd like to imagine a different perspective on the matter: this is an opportunity to introduce a sacred and time-honored sports tradition to the league, which is having a perpetual argument over which records are the "real" records based on whether they were set before or after a rule change. From sportswriters whinging over Babe Ruth's single season home run record being the most valid because it was done in only 154 games, to endless discussion over which 1980's QB would be the true GOAT if they had played in the QB-friendly league of the 21st century, these kind of arguments are the backbone of any true sports fandom. Now we can truly validate the ISFL by introducing such arguments as how Jay Cue Jr. has the real career rushing yards record because he did it under more punishing regression rules, or something. These kinds of changes should be embraced, I think, and I don't believe that some players losing a treasured career record is a strong enough reason to keep the league from evolving as needed.
A Silly Proposal
Just as something to tack onto the end of this article, I think it could be fun to introduce a mechanic whereby a user who sticks with their player throughout the entire length of a career can opt to play a final season at the DSFL level, just as a send-off tour. This would likely cause problems for DSFL balance, which would have to be considered, and many users would wonder why they would put their old player in the DSFL when they could just re-create and begin their new player also at that same level, so this idea is incredibly fraught, but I still wanted to mention it because it could be fun for users who might have sentimental attachment to their player.
Okay. That's all I have for now. Please let me know what you think of these proposals, I would love to hear about what veterans who have gone through regression think of it. Thanks so much for reading!
~Jenni
RE: Proposal for Changes to Player Regression - qWest - 03-09-2023
(03-09-2023, 02:40 AM)aeonsjenni Wrote: a user who sticks with their player throughout the entire length of a career can opt to play a final season at the DSFL level
Oh shit new Amadan multi just dropped
RE: Proposal for Changes to Player Regression - br0_0ker - 03-09-2023
13 seasons is actually pretty long compared to the NFL average, e.g. there's about 30-50 active NFL players with more than 13 seasons played, across 1500+ active players.
RE: Proposal for Changes to Player Regression - DREAMSLOTH - 03-09-2023
I'm a big fan of model #1:
After 7th season 20% TPE loss
After 8th season 25% TPE loss
After 9th season 35% TPE loss
After 10th season 35% TPE loss
After 11th season 35% TPE loss
After 12th season 35% TPE loss
After 13th season 50% TPE loss
After 14th season 60% TPE loss
After 15th season 75% TPE loss (irrelevant, retirement is forced regardless of TPE)
The raw regression chart is the one that sells it for me. On top of what you already mentioned, I think there are two key benefits for employing a model like this:
1. Players who are active and enjoy their current player and/or LR's have an opportunity to stick around longer.
2. I think we offer a necessary solution to two complex problems: a. drastic variation in quality of some draft classes, and b. smaller draft class sizes.
For #1, I like how we can continue to create a positive experience for a given user by extending the length of their current experience. I'm especially fond of the sentimental benefit for first-time creates like myself who are near-max earners and enjoy being on their current team.
For #2, when we run into situations like the past couple seasons where talent has been a much more shallow pool, having long-time players continue to be meaningful contributors offers teams a way to mitigate the damage done by regression. I'm not sure whether or not it would make the league more competitive, though at the very least I don't see any downside to any of this.
Additionally, I think having a regression system with a longer tail will more or less force teams to improve the quality of their draft scouting. Quality has varied by team and GM for what seems to be the entire history of the league; if there's additional incentive to draft players that could max earn (or at least appear like they'd stay active over time), I'd hypothesize that teams will have to step up scouting efforts. That then creates more competitive teams and a better experience for users during the draft process.
In summary, I think you should genuinely push for the adoption of that first model. Also, your analyses are amazing and I hope you continue to do this kind of work.
RE: Proposal for Changes to Player Regression - tstols - 03-10-2023
nice proposal and enjoyed reading all the data behind it, but I also have a quick idea on the side:
what if regression affected different positions differently, like it does in the NFL?
QB's generally have longer shelf lives if they avoid serious injuries, while players who live in "the trenches" have shorter ones. It would probably suck for OL enjoyers like myself if a system like this got implemented, but could be an interesting way to see how different players regress at different rates. If the rates were sped up at big skill positions, could also see more volatility at those positions too
that being said, it would likely be too complicated to really set regression up in multiple ways like that
RE: Proposal for Changes to Player Regression - aeonsjenni - 03-10-2023
(03-09-2023, 11:07 AM)br0_0ker Wrote: 13 seasons is actually pretty long compared to the NFL average, e.g. there's about 30-50 active NFL players with more than 13 seasons played, across 1500+ active players.
I understand why a lot of people think 13 seasons is long enough for a sim player, but this is really making the opposite point that you're trying to make. There are more than 30 active NFL players who have played more than 13 seasons, plus hundreds who retired after 13 seasons, and zero sim players. There will never be any sim players to play more than 13 seasons, barring a rule change. 13 seasons is a long time, of course, but it doesn't make any sense as an absolute upper limit.
RE: Proposal for Changes to Player Regression - aeonsjenni - 03-10-2023
(03-09-2023, 03:52 PM)dreamSloth Wrote: I'm a big fan of model #1:
After 7th season 20% TPE loss
After 8th season 25% TPE loss
After 9th season 35% TPE loss
After 10th season 35% TPE loss
After 11th season 35% TPE loss
After 12th season 35% TPE loss
After 13th season 50% TPE loss
After 14th season 60% TPE loss
After 15th season 75% TPE loss (irrelevant, retirement is forced regardless of TPE)
The raw regression chart is the one that sells it for me. On top of what you already mentioned, I think there are two key benefits for employing a model like this:
1. Players who are active and enjoy their current player and/or LR's have an opportunity to stick around longer.
2. I think we offer a necessary solution to two complex problems: a. drastic variation in quality of some draft classes, and b. smaller draft class sizes.
For #1, I like how we can continue to create a positive experience for a given user by extending the length of their current experience. I'm especially fond of the sentimental benefit for first-time creates like myself who are near-max earners and enjoy being on their current team.
For #2, when we run into situations like the past couple seasons where talent has been a much more shallow pool, having long-time players continue to be meaningful contributors offers teams a way to mitigate the damage done by regression. I'm not sure whether or not it would make the league more competitive, though at the very least I don't see any downside to any of this.
Additionally, I think having a regression system with a longer tail will more or less force teams to improve the quality of their draft scouting. Quality has varied by team and GM for what seems to be the entire history of the league; if there's additional incentive to draft players that could max earn (or at least appear like they'd stay active over time), I'd hypothesize that teams will have to step up scouting efforts. That then creates more competitive teams and a better experience for users during the draft process.
In summary, I think you should genuinely push for the adoption of that first model. Also, your analyses are amazing and I hope you continue to do this kind of work.
I'm a fan of model #1 as well, though I'd like to see some adjustments made to it, perhaps. I slapped it together in a couple hours and I think there are improvements to be made on the exact numbers. A part of me feels that the percentages should be abandoned in favor of fractions, but that's another thing. I definitely agree about your first point. I like my player! I'm already attached, and I genuinely enjoy writing little point tasks and interviews for her. I'm sure I'll feel a different way a couple years down the line but I really feel that since going through regression is entirely optional it's totally feasible to make it a little more fun! I really can't comment so much on your second point as I'm a new player and have never been involved in the drafting process, though I feel like from a competitive balance standpoint there will definitely be upsides and downsides, which could definitely be looked into.
All-in-all I'm very glad you think the model is worth genuinely worth pushing for. I really don't know what the process would be for pushing for this but I'd appreciate any support.
RE: Proposal for Changes to Player Regression - aeonsjenni - 03-10-2023
(03-10-2023, 12:06 AM)tstols Wrote: nice proposal and enjoyed reading all the data behind it, but I also have a quick idea on the side:
what if regression affected different positions differently, like it does in the NFL?
QB's generally have longer shelf lives if they avoid serious injuries, while players who live in "the trenches" have shorter ones. It would probably suck for OL enjoyers like myself if a system like this got implemented, but could be an interesting way to see how different players regress at different rates. If the rates were sped up at big skill positions, could also see more volatility at those positions too
that being said, it would likely be too complicated to really set regression up in multiple ways like that
In general I believe it's best to maintain simplicity where we already have it. Introducing complexity to the regression system demands that we account for a number of things such as the possibility of a position switch. If we wanted to find a way to extend QB careers (at least allow them to be successful for longer), then it may be in our best interest to find a way to allow QB's to need less TPE to be successful in the league, which would likely involve an alteration to the update scale. Either way I'm glad you enjoyed reading my proposal!
RE: Proposal for Changes to Player Regression - infinitempg - 03-10-2023
i think rich users should be allowed to bribe away the regression man 13 seasons is honestly pretty long already. ignoring the realism concerns - 13 seasons is over 2 IRL years. like, my last player Busch Goose was an S28 create - the creation date was December 21, 2020. Wolfie Jr (S41 create) was created on January 19, 2023. Goose played 12 seasons in the ISFL and 1 in the DSFL, for just barely over 2 years. I could have pushed him one more season too to probably get to March 31, 2023, putting Goose at 2 years and 3 months - that's a long time!
at a certain point, you could argue a user needs to be almost forced to start over and make a new player. it creates a new type of engagement for lots of users and allows them to explore the league again - or it could force users to take a short break between players and restart fresh.
maybe it's just because i've been here for a while, but i've found that careers have gotten pretty stale after a while and recreating allowed me to join new teams without feeling disloyal to the team that drafted me.
RE: Proposal for Changes to Player Regression - aeonsjenni - 03-10-2023
(03-10-2023, 11:27 AM)infinitempg Wrote: i think rich users should be allowed to bribe away the regression man 13 seasons is honestly pretty long already. ignoring the realism concerns - 13 seasons is over 2 IRL years. like, my last player Busch Goose was an S28 create - the creation date was December 21, 2020. Wolfie Jr (S41 create) was created on January 19, 2023. Goose played 12 seasons in the ISFL and 1 in the DSFL, for just barely over 2 years. I could have pushed him one more season too to probably get to March 31, 2023, putting Goose at 2 years and 3 months - that's a long time!
at a certain point, you could argue a user needs to be almost forced to start over and make a new player. it creates a new type of engagement for lots of users and allows them to explore the league again - or it could force users to take a short break between players and restart fresh.
maybe it's just because i've been here for a while, but i've found that careers have gotten pretty stale after a while and recreating allowed me to join new teams without feeling disloyal to the team that drafted me.
Thanks for the input, I am glad to see it! How do you feel about a change to the regression rates? What strikes me in particular is that the raw regression goes up over time, taking more from less in the later years. Do you think a 13 year cap with a softer regression rate would be beneficial? Do you think there might be ways to encourage re-creation while still letting users stick with their players a little longer if they want?
Edit: If this feels like I'm grilling you or anything I'm sorry. I'm just curious what people think.