Soluna

Improvement Thursday #2

49 posts in this topic

Just now, alienufo said:

giphy.gif

 

So I actually got Manning (retired due to injury) confused with Thomas Johnson from the year before, who was the jogger murderer. But yeah, Johnson had some pretty severe mental issues that went unnoticed until, well, he went missing for a few days, then later hacked a random guy to death with a machete.

Share this post


Link to post
Share on other sites

I was brain storming about this last week and I thought that we could use a pipeline state type of system in order to not make the instate bonus so tough to overcome.

 

First we would start by saying that each state is worth a certain amount of Pipeline points (Which could vary from year to year). So Texas and Florida would be like 20 points and states like Vermont or Wyoming could be as little as 1 point. 

 

Then each year every team gets a certain amount of pipeline points (Based on team success) to add different states to their pipeline. 

 

Then you could get a small bonus on each recruit that is in your pipeline state. 

 

This would allow teams that are in states with a smaller # of recruits to get guys that are out of state.

caesari likes this

Share this post


Link to post
Share on other sites
3 hours ago, Soluna said:

What I would consider is this: a scholarship offer, in addition to providing points, cancels all automatic state points on that player

 

I think that might work if an instate school didnt put anything into recruiting the player after like week 6-8 or something. By then the recruiting process has begun to take shape and players would begin looking at OOS options if they still hadn't received interest from in-state schools. Otherwise, there would be no built-in benefit to recruiting in-state over better OOS prospects 

Share this post


Link to post
Share on other sites

Is there a way to change the high schools players are coming out of? I've noticed Oregon has a lot of players coming out of 4A (Oregon uses a 6 division system with 6A being the biggest schools) and lower schools, even 2A/1A which play 8 man and never have D1 croots. It's not a necessary thing but it could add to the realism of the site.

Share this post


Link to post
Share on other sites
4 hours ago, HAFFnHAFF said:
  • Denote ties.  Right now it is listed in reverse alphabetical order so I can't tell if I am leading or tied on guys I only put a scholarship on.

Please make this happen

 

4 hours ago, Soluna said:

What I would consider is this: a scholarship offer, in addition to providing points, cancels all automatic state points on that player

I think this should only apply if no in-state schools offer a scholarship to a player. The moment they do, the state points should also count. 

 

Let's take inspiral's example. A player is from Arkansas, but he got an offer from LA Tech. Instead of walking on at Arkansas, he chooses to accept the scholarship and go out of state, which makes a lot of sense. However, if Arkansas and LA Tech both offer a scholarship to this player, in real life he would pretty much always choose the in-state school, therefore the state bonus should apply if a home-state school offers a player a scholarship. This way nobody could recruit a class containing dozens of 3.0 or 3.5 potential players even if their home state is loaded with them.

Share this post


Link to post
Share on other sites

I would have a small percentage of players (maybe 7.5-15%) that instead of in-state bonuses, had in-state penalties, to include border states as teams penalized.  Some players just want to leave the area, not stay home.

 

This would, in effect, nationalize recruiting for a portion of the recruit population, and add countless new strategy options to the already dynamic recruiting system that was put in place last season.  In addition to adding a number of players that are wide-open to anyone who wants to spend on them (creating some really fun battles), it could also reduce the in-state advantages that some schools have, because they may be spending more points outside of their borders on these new wide-open recruits instead of just harvesting all the in-state recruits with all of their points, potentially allowing more people to cross state lines and reach in to steal players away.

Share this post


Link to post
Share on other sites

I'm glad there is talk about the pipeline system. I pipelined Deleware last season (their supply of recruits surprisingly lined up somewhat well with my needs), and feel like that was a 100% wasted move. I started so far behind Maryland and Navy, that I decided it wasn't worth it to keep pursuing any of those targets, and switched to CO, NE, KS prospects where I already had in-state or neighboring state points and try to salvage a decent class.

 

While I think it would be a lot of work, I'd really like to see a system where recruits will favor someplace where they can play sooner, rather than be buried in the depth chart.

 

I'm about to start my sixth season at Air Force, and I have never had a pocket passer play, it's been 100% scramblers. I feel like pocket passers wouldn't even want to come here, while scramblers know they'd get a good shot (minor bonus to scramblers, and a minor penalty to pocket passers). This bonus/penalty wouldn't be coach selectable, but be assigned by the sim based on history, including how long the coach has been there (a new coach may very well change everything up).

 

A system where the sim can assign a title of "linebacker U" for example, to a school that has a history of producing top talent at that position, leading to recruits of that position wanting to play there to increase their chances of getting drafted. Miami (FL) might be "Scrambler U" for example as they have a history of producing high quality scramblers. Would need something more than just rating to base it on I think. These titles could be gained and lost as seasons went by, and all possible titles wouldn't necessarily be 'in circulation' at the same time.

Share this post


Link to post
Share on other sites

Minor Adjustment to the Pipeline: Thieve the model from CBBHC.

 

Where you play a game is very important to the real programs. Entire conferences set themselves up to give teams access to playing in a specific state (ACC) and some teams go out of their way logistically to play more often in a state they recruit (Navy in the AAC West). This could be a total change to pipeline, giving 3 points for every game played in a state that year. This would give a team automatically 18 points in their home state if they play 6 home games. A large downside to this would be states like Texas or Ohio where some teams play a lot of in-state rivals, thus giving themselves 24-27 points a year.

Share this post


Link to post
Share on other sites
3 minutes ago, Rome said:

Minor Adjustment to the Pipeline: Thieve the model from CBBHC.

 

Where you play a game is very important to the real programs. Entire conferences set themselves up to give teams access to playing in a specific state (ACC) and some teams go out of their way logistically to play more often in a state they recruit (Navy in the AAC West). This could be a total change to pipeline, giving 3 points for every game played in a state that year. This would give a team automatically 18 points in their home state if they play 6 home games. A large downside to this would be states like Texas or Ohio where some teams play a lot of in-state rivals, thus giving themselves 24-27 points a year.

I like that a lot! Maybe limit the bonus to 18 points? So you have a full bonus in ur home state and then smaller bonus in other states?

Share this post


Link to post
Share on other sites
1 minute ago, HAFFnHAFF said:

I like that a lot! Maybe limit the bonus to 18 points? So you have a full bonus in ur home state and then smaller bonus in other states?

The issue with this though, is that it may make it a LOT harder for some teams to get OOC games scheduled.  People might not want to ever play @ Wyoming, for example.

Share this post


Link to post
Share on other sites
9 minutes ago, pumph said:

The issue with this though, is that it may make it a LOT harder for some teams to get OOC games scheduled.  People might not want to ever play @ Wyoming, for example.

 

Thats a fair point. 

Share this post


Link to post
Share on other sites
52 minutes ago, DescretoBurrito said:

I'm glad there is talk about the pipeline system. I pipelined Deleware last season (their supply of recruits surprisingly lined up somewhat well with my needs), and feel like that was a 100% wasted move. I started so far behind Maryland and Navy, that I decided it wasn't worth it to keep pursuing any of those targets, and switched to CO, NE, KS prospects where I already had in-state or neighboring state points and try to salvage a decent class.

 

I see the point about maybe making it easier to overcome, but a team in Colorado shouldnt be able to easily dip into Deleware and take players from in state schools. I feel like going out of the  immediate area should be reserved for one or two cant miss prospects.

 

That said I understand there are some places where the immediate area doesnt have much talent or has a lot of schools and that might be able to be addressed but I feel like if everyone had their way there would be little bonus to recruiting in state.

 

I think the one solution would be that if you pipeline a neighbor state, you get double the neighbor state starting points. So if I pipeline AL, I have 24 points on all Alabama players where Alabama schools have 25. That way I am still disadvantaged by 1 point and  their 20% bonus compared to my 10%. That would allow for more competitive regional recruiting but still not make it easy to go all the way across the country for the heart of your recruiting class. 

Share this post


Link to post
Share on other sites
30 minutes ago, cmcgill said:

 

 a team in Colorado shouldnt be able to easily dip into Deleware and take players from in state schools.

 

minor point but Deleware doesn't have any in-state schools in the sim.

cmcgill likes this

Share this post


Link to post
Share on other sites
2 hours ago, notoriousbigej said:

Is there a way to change the high schools players are coming out of? I've noticed Oregon has a lot of players coming out of 4A (Oregon uses a 6 division system with 6A being the biggest schools) and lower schools, even 2A/1A which play 8 man and never have D1 croots. It's not a necessary thing but it could add to the realism of the site.

 

its pretty easy to add schools to the list. for reference, heres all the schools currently in it: https://docs.google.com/spreadsheets/d/1eqzl75_gNVCBm9Ugc3gBp9ASNBPhwqjLWigm3SEypi4/edit?usp=sharing

notoriousbigej likes this

Share this post


Link to post
Share on other sites

I feel like scholarships aren't worth enough, but I'm not sure what a good number for them would be. IRL if I get a scholarship offer for a full ride from an out of state school, or can pay my way to an instate school, im probably taking thay full scholly.

 

I also am not sure about the instate bonus at all. IRL Wisconsin's spring roster has something like 52 of 103 people on it actually having gone to HS in Wisconsin. Meanwhile SIM Wisconsin hauled in 35 croots last year at the signing deadline, 33 of which were from Wisconsin. It just seems like an odd ratio, and while I don't think random shotgunning was perfect in the old system, i think out of state recruiting is discouraged a bit too much.

 

Perhaps we could make scholarships worth more, but you only get so many over a 4 year period. A player with a scholarship might be less likely to transfer later down the road, and you could award a guy who walked on and played great with one of your scholarships, rather than risk him transferring elsewhere where someone else might give him one. 

smckenz3 likes this

Share this post


Link to post
Share on other sites

Alright, I'm going to piggy back off Sleuth's point and kind of take it in a different direction and throw some data on y'all. Quick side-point: I don't know why Sleuth and I end up having similar opinions on everything on this site, I swear we don't do it on purpose nor even talk about any of this.

 

Scope of Data

To simulate the following assertions I'm going to be focusing in on just one kind of player in order to get the most accurate correlations possible;

  • Pocket Quarterbacks
  • From 2018-19 seasons (no injuries)
  • Must've played 12+ games in a season

While there's tons of other variables that go into quarterbacks on the personal, team, scheme, and conference levels, however the sample size will hopefully smooth some of this over. Additionally I feel like using QB data to measure personal production is best as they have constant stat sheet appearances and are arguably the most independent position on the field.

 

 

Assertion 1: I believe recruiting a prospect based on an average of their current and potential skill gives too much certainty on their production.

 

For this example I'm going to lay out the QBR of all these QBs and compare them to their current skill level, potential skill, and an average of those too (true skill level). Here's a chart throwing trendlines on all the data;

ydsBtI6.png

 

 

From here it's clear to see that "true skill" has a fairly tight correlation all things considered with an r-squared value of about 0.5. I don't personally think r-squared values are a great indication of correlation, so instead I put the data into a 95% confidence interval;

nh2LxMd.png

7G69bVv.png

 

What this data shows is that with 95% confidence, a players true skill can determine a certain range of QBR production. As said before this true skill is an average of current and potential skills, so like if you had a 3/5 QB his true skill would be 4. 

 

If you have a 5/5 or a 4.5/5 (5 - 4.75 true skill) then you can expect with 95% confidence (or 19/20 QBs) that he'll have a season QBR of between 104.3-113.1, and etc. As you go down the true skill ratings, it's apparent that these ranges hardly overlap thus showing that the true skill of a player can give a general and definite range of expected production. However, from here I wanted to see if just the potential skill level had a similar correlation.


 

Assertion 2: I believe recruiting a prospect based on solely their potential skill gives too much certainty on their production.

 

Same shit, different stat. Let's look through a 95% confidence interval with solely potential skill rating vs QBR.

DpJh2Pu.png

aOQNkVu.png

 

While there's still a little more overlap, these ranges are still very separated which shows to me that production of players can be reasonably determined by the skill values you recruit them by.

 

 

Solution : There should be more busts in CFBHC

 

This can be some weird attribute or statistic, this can be just more volatility in player's attributes, or it can be a difference in how we handle recruiting. I vote for the latter. I think the skill levels of players should just be "estimated" where there's a chance that 3/5 player you recruited ends up just being a 2/4.5, 3/4, or something. We could also do something where we scrap the skill ratings during the recruiting process in place of general letter grades that estimate a player's potential. Like an "A" would be 40% 4.5 potential, 40% 5.0 potential, 20% potential, etc. This will give some more variability and simulate "busts" without having to rework the inner workings of the simulation itself.

Dacder, Vxmonarkxv, Jumbo and 1 other like this

Share this post


Link to post
Share on other sites
44 minutes ago, AzulCaballero said:

What this data shows is that with 95% confidence, a players true skill can determine a certain range of QBR production. As said before this true skill is an average of current and potential skills, so like if you had a 3/5 QB his true skill would be 4. 

 

That isnt what a confidence interval is at all. It means that the probability of the true mean of each group being within that range is 95%, not that 95% of players in the group fall within that range

AzulCaballero likes this

Share this post


Link to post
Share on other sites

You're right, I definitely mis-interpreted that. But I think general point still stands, the production of players based off skill is a little too tightly correlated for all the hundreds of variables that I didn't take into account. You can recruit that 3/5 Pocket QB and generally expect a certain level of production when in reality that 5-star IRL QB may bust and play like a 3-star while some 3-star may develop and play like a 5-star.

cmcgill likes this

Share this post


Link to post
Share on other sites

Hold my beer, I'm gonna keep going.

 

 

So I decided to look into IRL college football QBs to see if the "bust-rate" is similar to our sim. For my initial post I used just 2 seasons (2018-19) of sim seasons as Pocket QB data in which it represents a spread of 6 recruitment classes - so when looking at IRL data I used 6 seasons of "Pro Style" QB recruitment. In order to get sufficient data I used the recruiting classes between 2007-2012 so that all IRL QBs have graduated and played sufficient time. Similarly to my initial post I tried my best to only represent 2 seasons of data (usually their latest and best seasons) unless they transferred then I did 3. Lastly, I only compared IRL 5-star recruits (essentially 5 potential) to the sim's 5 star potential players.

 

Here's all the IRL 5-star recruits, according to Rivals, and their data;

QU5vzLo.png

 

From there I used my initial posts' data to assign the expected "true skill" of each player. If a players' true skill is "5" then they're performing essentially at a 5/5, a "3" would be performing essentially at a "1/5" level, therefore performing at a 3 or less is essentially a bust because they're performing at or below the bare minimum of their potential. Here's how things played out:

sVkYyHV.png

n0OHtDj.png

 

 

Takeaways

  • There's too many 5-star potential QBs to be realistic in the sim (31 vs 16), especially considering I used 6 seasons of IRL CFB data vs 2 of the sim. If grouping in the 4.5 potentials, that number doubles to 62. In two seasons, 62 times an above-4 star player has played at least 12 games in our sim compared to 16 IRL.
  • Performing between a 3.25-5 true skill is fairly evenly distributed IRL while the sim heavily skews toward 5-potential pocket QBs to perform in the 5-4.75 skill range. That's performing at the level of a 4.5/5 or 5/5 QB.
  • The bust percentage of IRL Pocket QBs is 44% vs the Sim's 13%.
  • Though not confirmed, there's likely not a lot of the same thing, but backwards. Jameis Winston was a 4-star recruit that played at a 5 star level, etc.

 

constapatedape and bingo415 like this

Share this post


Link to post
Share on other sites

Azul, to your last point: Sean O'Reilly (MINN '19) was a 4.0 player who played all four years at 4.0/4.0. His first two years were truly at a 3.5 or so level, but his last two years (especially SR) he played a 5.0 level. His stats and 1st team All-BIG confirmed that. It does happen, but not perhaps as much as irl.

AzulCaballero likes this

Share this post


Link to post
Share on other sites

Soluna has noted that a 1/5 isn't really considered a 5* player...  a 3/4.5 would be considered superior to a 1/5 for instance. Not sure how you'd take that into account for a star system though

DangerZoneh likes this

Share this post


Link to post
Share on other sites
1 hour ago, bingo415 said:

Azul, to your last point: Sean O'Reilly (MINN '19) was a 4.0 player who played all four years at 4.0/4.0. His first two years were truly at a 3.5 or so level, but his last two years (especially SR) he played a 5.0 level. His stats and 1st team All-BIG confirmed that. It does happen, but not perhaps as much as irl.

 

True, and funny you say that because I just checked and saw that O'Reilly (18) and Raoul Gomez (18) were both 4/4's that played at a solid 5 level. It did go down steadily after those two though. 

 

1 hour ago, Jumbo said:

Soluna has noted that a 1/5 isn't really considered a 5* player...  a 3/4.5 would be considered superior to a 1/5 for instance. Not sure how you'd take that into account for a star system though

 

When I say "true skill" I mean the average of those two ratings. A 1/5 would be a 3 and that 3/4.5 would be a 3.75, for example. So when a five-star potential is playing at or below a "3 level" that means he's playing at a level that is the bare minimum or well below any possible 5-five star player's expectations because that would assume they'd be playing below a 1/5 level.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now