Player skill sets: realistic?

[jcard]

One thing I am struggling with in trying to immerse myself in this game is the nagging feeling that the player attributes are each randomly assigned values, leading to players that have combinations of attribute ratings that are not logically structured nor make for a valuable use of a roster spot. For example, quarterbacks with ratings for the five distances that are like 52, 16, 74, 09, 42. Or defensive backs with single-digit man ratings and 90+ zone ratings. This feels like saying someone is good at algebra, poor at geometry, and excellent at trigonometry; in reality, related types of skills (passing, left-brain processes) share some common level of base competency around which you might observe variations. Similarly, a low-rated quarterback whose one strong attribute is Third-Down Passing, which the documentation describes as the ability to make successful passes when the defense is expecting a pass. If you show me a good third-down passer in the NFL, you’re showing me a good passer... period. One more is a RB with mediocre running attributes, great route running ability, and poor hands. If you are trying to win a roster spot, the combination of skills is more important than just the sum of those skills. A poor runner with good hands and good route running ability might have less overall talent than the former, but more value to a roster.
Basically, I feel like some form of archetype system should have been used in generating players. Matching up random skill combinations is fun from a tactical or puzzle-solving perspective, but it compromises the feeling of simulating professional football when you are doing this mixing and matching with players with skill sets for which you see no corresponding examples on the field every Sunday.

[Front Office Midget]

One thing to consider is that "0" is probably not as low as you think as it is. A QB with a very low rating in short passing will still complete most of their short passes. A CB with low m2m isn't just going to get burned nonstop. A LB with low run defense will still make plenty of tackles in the running game, etc. etc.

[Dawgfan19]

FOF does simulate NFL stats effectively (although from the 1990s). But if you attempt to correlate this level of detail to the NFL, you will be disappointed.

[tzach]

Originally Posted by jcard

One thing I am struggling with in trying to immerse myself in this game is the nagging feeling that the player attributes are each randomly assigned values, leading to players that have combinations of attribute ratings that are not logically structured nor make for a valuable use of a roster spot. For example, quarterbacks with ratings for the five distances that are like 52, 16, 74, 09, 42. Or defensive backs with single-digit man ratings and 90+ zone ratings. This feels like saying someone is good at algebra, poor at geometry, and excellent at trigonometry; in reality, related types of skills (passing, left-brain processes) share some common level of base competency around which you might observe variations. Similarly, a low-rated quarterback whose one strong attribute is Third-Down Passing, which the documentation describes as the ability to make successful passes when the defense is expecting a pass. If you show me a good third-down passer in the NFL, you’re showing me a good passer... period. One more is a RB with mediocre running attributes, great route running ability, and poor hands. If you are trying to win a roster spot, the combination of skills is more important than just the sum of those skills. A poor runner with good hands and good route running ability might have less overall talent than the former, but more value to a roster.
Basically, I feel like some form of archetype system should have been used in generating players. Matching up random skill combinations is fun from a tactical or puzzle-solving perspective, but it compromises the feeling of simulating professional football when you are doing this mixing and matching with players with skill sets for which you see no corresponding examples on the field every Sunday.

i agree with both responses above. remember that this is a simulation of real outcomes. some other points for you to consider.

- for real players we don't observe the individual ratings as we do in a simulation. we just observe play outcomes. in real life there's no easy way to reverse engineer most bars (some you can, like drops, route running, punishing hitter). so you shouldn't put much weight on whatever way you think the bars (skill set) should be distributed (no one knows how to do that!).

- "If you show me a good third-down passer in the NFL, you’re showing me a good passer... period" -> this is a good example of what i mean -- a similar case could be made for you other concerns. the game models a behavior that some qbs will play better than others in obvious pass situations (like 3rd down). this is all that you could take from that bar. it even doesn't mean that the qb will complete more 3rd down passes based on that bar alone (other skills would come into play as you would expect). and we observe that in the nfl -- not all good 3rd down passers are good passers. to give you a real life example -- few people would know who the #2 qb in 3rd down efficiency was in the 2020 nfl season (that's gardner minshew).

[MIJB#19]

Originally Posted by Front Office Midget

One thing to consider is that "0" is probably not as low as you think as it is. A QB with a very low rating in short passing will still complete most of their short passes. A CB with low m2m isn't just going to get burned nonstop. A LB with low run defense will still make plenty of tackles in the running game, etc. etc.

Just a minor correction to this, a rating of 0 is impossible to get a read on how bad a player actually is, but anywhere between 1 and 99 will be a guideline on how good the player is compared to other pro football players. And depending on how good or bad your staff members are in evaluating talent (scouting).

Originally Posted by Dawgfan19

FOF does simulate NFL stats effectively (although from the 1990s). But if you attempt to correlate this level of detail to the NFL, you will be disappointed.

FOF8 is quite on par with NFL stats of the late 2010s as far as I've seen, isn't it? But my perception could be wrong by lack of single player experience and being too much drawn into the multiplayer experience.

[Dawgfan19]

Originally Posted by MIJB#19

FOF8 is quite on par with NFL stats of the late 2010s as far as I've seen, isn't it? But my perception could be wrong by lack of single player experience and being too much drawn into the multiplayer experience.

I suppose it depends the stats/metric one considers. Team scoring per game is actually similar (within 1 point) until you reach the NFL 2018 season. However, the pass stats (specifically completion pct. and QBR) mirror more closely the 90s or early 2000s. And I am drawing on stats from my MP leagues. Of course, perhaps the GMs in my leagues do not deploy an effective pass offense. I have previously speculated many guys don't throw short often enough.

[Solecismic]

Originally Posted by Dawgfan19

I suppose it depends the stats/metric one considers. Team scoring per game is actually similar (within 1 point) until you reach the NFL 2018 season. However, the pass stats (specifically completion pct. and QBR) mirror more closely the 90s or early 2000s. And I am drawing on stats from my MP leagues. Of course, perhaps the GMs in my leagues do not deploy an effective pass offense. I have previously speculated many guys don't throw short often enough.

I'm glad you recognize that scheme drives the numbers in the NFL.

I've written about this in the past, and it gets lost or misquoted, so I like to have it in several places.

This is how I "calibrate" the engine, for lack of a better word. I take the current player file, which has a controlled distribution of rating initiators. I play it dozens of times for the current year.

I try to match a set of NFL averages as closely as I can, tweaking the engine here and there. My calibration "file" - it's really just archaic #defines - has at least 100 data points I use for tweaking (and no, it's not exposed to the user - maybe some day if I weren't so damned old). In reality, it should be tied to the player file for a given year. Ideally, it should "change" over the years, forcing players to change scheme with it.

The averages I use are five-year rolling averages (so, for 2021, I'd use 2016-2020). The reason I use five years is that an NFL season only contains 256 regular-season games (272 now). When it comes to many stats, particularly lower numbers like score or interceptions, 512 data points actually has quite a high variance. But schemes change considerably over the years, so I have to limit how much I include.

Each season is a trial over "real-life" unknown averages. The 2019 offensive explosion was part scheme and evolution and part "this must have been considerably over real-life expectations". My goal is not to write a replay sim. If so, I'd be a lot more into player files (I really hate doing them).

Then, and here's the part that I got in trouble for, and maybe rightfully so, I reduce some of the key offensive targets - particularly scoring (which is a combination of quite a few reductions). Something like completion percentage (which has steadily increased in real life over time) won't take much of a hit, but maybe a percentage point or so.

I did this because of MP leagues. They make up maybe 2-3% of my customers, but leagues form from discussion forums and other communications methods. There are fewer checks on roster limits and game plans in MP (it's the same engine - the engine doesn't know if the game is SP or MP, scoreboard with play-calling or no scoreboard, etc).

So, what happens when someone tries something unique? The engine can either shut it down or not. In real life, opponents would quickly adjust. I've read a ton about scouting and a lot of it is grunt work taking film and knowing what teams like to do (coaches still tell you to read Steve Belichick's - the Great One's dad - book on scouting from Navy in the 1950s if you're a coach and are expected to help with scouting). It takes a lot of practice time to train a team to run things differently.

That's not fun. Not a good game at all. So I assume scouts already do all that and that's where you get your "opponent is familiar" message. If you go all-long-pass, you'll get shut out because of that. But I can't make it too stringent, or there would be no value in detailed game-planning. So the MP game becomes learning the game's limits and finding what keeps working - especially against opponents who aren't watching to see what you're doing.

But the bottom line is that when people optimize against the engine, offense can go up a lot. And so my reduction in target scoring ends up greatly reducing the maximums from the most extreme seasons from the thousands of seasons generated from the 2020 engine. So it's not 1970s (those particular claims reflect a lack of knowledge about the history of the game) or 1990s or 2000s. It's something a bit different, reflecting current NFL schemes.

Why do I care? Because people like to post about their extreme experiences. Just a natural thing to do - nothing wrong with that. But others read it and make conclusions about how the game is just "Tecmo Bowl" or something. And I'm already dealing with the frustrated MP owners who blame me for every loss they have. That part I can't reduce. There are former owners who, years later, still go to Steam and downvote everything. Even a couple who "buy" the new game for a half-hour just to write a scathing review, then return the game.

I have a love/hate relationship with MP. Love, because this is a small community with which I have a lot in common. If things were different, you'd be real-life friends. Hate, because a small percentage of this small community got really, really angry with me and wanted me out of business. I've had people posting pictures of my home online. I still get nasty notes on social media from time to time. Ben had personal threats against him as well just for moderating. I'm not sure I would include MP if I had to do it again.

[redfox000]

Thanks for that explanation. I can appreciate the difficulty in writing a game that is based on real life. I enjoy the game immensely and only play SP. At the moment the only real thing missing would be the immersion one would get if they could see their players.

[tzach]

many thanks for the detailed post, jim, and for producing this game that 99% of us love.

[Dawgfan19]

Originally Posted by Solecismic

I'm glad you recognize that scheme drives the numbers in the NFL.

Jim,

First, a big thank you for designing a game that truly mirrors the NFL (albeit, slightly from the past, more on that below!!!). I have recently began reading a book written by Pat Kirwan, Take Your Eye Off the Ball 2.0. The book was published 5 years old, so I am late to the game. But the internal Xs and Os of NFL football described by Pat very much illustrate how FOF (especially V8) has been built. The book almost looks like a software design document for FOF. Even the defensive game planning elements, which the game has been criticized for, is appropriate. On defense, it’s more about how to contain an offense which is reflected in the game.

Anyway, to the point, I wanted to let FOF validate some points based on data. So, I created 100 years of history with no human interaction, all AI game plans and team management. The results were interesting.

Over those 100 seasons, the passing completion pct was 61.1% with an average yards per attempt of 6.47. I was too lazy to calculate the 100 year historical QBR. If you look at NFL stats, it was around 2004 when the NFL league-wide truly adopted the west coast style of viewing the short pass game as an extension of the run game. In 2004, the passing pct approached 60% with a yards/attempt of 7.1. Those stats have consistently increased since then. We can also look at individual statistics. In my test 100 year sim, I had 70 QBs with a QBR greater than 100 (.7 per season). In the past 5 plus seasons (2016 to 2021), the NFL has had 45 QBs with a 100+ QBR (7.5 per season). So, I would agree with you that the completion pct is appropriate, but the yards/attempt seems low – not sure if the lower average yards completely accounts of the lower QBRs. Part of that may be due to you listening to a vocal minority to nerf the passing game since some GMs were “gaming the game” with a 100% pass offense. Also, the 5 year average method contributes, implying the game will always be behind the curve when there is a significant increase in passing efficiency in the NFL.

Based on my statistical background, I understand one can look at data in a multitude of ways (there is the joke about you have your lies, and you have your damned lies, and then you have your statistics!!!). So, perhaps my statement that the game reflects the 1990s NFL was a bit aggressive. However, I would argue the late 90s or early 2000s looks appropriate. That seems to be an appropriate conclusion – or perhaps the AI can’t effectively generate game plans and that skewed my 100 year data!!!

I only wish you could have maintained FOF and we could have a game more reflective of the current NFL. But you obviously need to do what is right for your family and career. Wishing you the best and continued success. Have a Merry Christmas.

[Solecismic]

One thing I realized a couple of weeks ago. Since I'm out of the game now, I can talk more. And I love the game and have tried very hard to separate what makes it fun for fans from the 24/7 tedium of running a team or coaching or even studying and training as a player studies and trains.

I have always wanted to have a set of FOF players I could bounce ideas off of. People who could discuss it from an intellectual perspective, without the distraction of "am I saying something that could change the game" or something along those lines. A couple of the beta testers are definitely people I would trust, but I was always worried I was asking too much.

That's one of the great things about OOTP - it's a team. A large team, some volunteer, some semi-volunteer, and then the core programmers. But getting from solo to part of a team like that is quite difficult. In the end, the timing was wrong.

So I really appreciate your comments. You're finding some of the things I've done intentionally, or perhaps unintentionally and just liked the way they turned out. The aging algorithm, for one, mutes some of the extremes in how players might develop. I found that ends up helping the defense, since defense is accumulated competence while the offense can be tailored to stars more easily. I wanted league records to fall slowly, though not unattainable, since players last less than 20 seasons.

So when I tune, I do run a few leagues out 25-30 years, but the grunt work comes with the first-season sim. Again, that approach is perhaps not an ideal one.

The five-year thing - more that 2018 would be 2016-2020, and I figured it would be better to use last five for the most recent rather than three. It's true that if trends are rising rapidly, it's not the best approach, but do we know the true mean is always rising? Sometimes it falls, too. In the end, it still must be football. I have a nice spreadsheet somewhere than goes back to 1974 with about 20 different measures, mostly of the passing game. The five-year approach is sometimes optimal, and I've even found a smoother curve with seven - enough to consider that change if I were to do all this again.

Yes, Kirwan's book is often on my desk. If you see some of it in FOF8, I think I first read it during that development phase. For certain, the idea of separating first play of drive from other first-down plays came from Kirwan. It's a wonderful book. I had the first edition as well.

[Dawgfan19]

Originally Posted by Solecismic

The aging algorithm, for one, mutes some of the extremes in how players might develop. I found that ends up helping the defense, since defense is accumulated competence while the offense can be tailored to stars more easily.

This is a very interesting comment. An anecdotal observation in my two MP leagues is that the defense seemed to become more important recently. But I did not think of or have the time to test that theory.

[Red Zone]

"I'm glad you recognize that scheme drives the numbers in the NFL."

Which is what a lot of us have said since the patch.
Just because your old game plans don't work anymore, doesn't mean you cant make new ones that will. Its only 1990s football if you let it be.

"One thing to consider is that "0" is probably not as low as you think as it is. A QB with a very low rating in short passing will still complete most of their short passes. A CB with low m2m isn't just going to get burned nonstop. A LB with low run defense will still make plenty of tackles in the running game, etc. etc."

I also have always felt this was true
Even though the ceiling can be very high in some cases, The Floor is not nearly as low as some expect it to be, particularly on defense and in the kicking game.

Jim,
I am sorry that some folks took this "game" to such an extreme that you felt uncomfortable. They would be the minority and they should be ashamed of themselves.
While most of us have some part of the game that we wish was different, or better, I think most would agree that this is the only text sim Football Game that has been worth playing for the last 20 years or so.
I know I have personally enjoyed it for 17 years and still play in 7 MP leagues and commish an 8th.
A Big Thank You and Here's wishing you and your Family a Very Healthy and Happy Holiday Season.
I Hope you are doing well in your new Endeavors.

[tzach]

Originally Posted by Solecismic

This is how I "calibrate" the engine, for lack of a better word. I take the current player file, which has a controlled distribution of rating initiators. I play it dozens of times for the current year.

I try to match a set of NFL averages as closely as I can, tweaking the engine here and there. My calibration "file" - it's really just archaic #defines - has at least 100 data points I use for tweaking (and no, it's not exposed to the user - maybe some day if I weren't so damned old). In reality, it should be tied to the player file for a given year.

the engine you created is so amazing that when i run my very run heavy schemes, i get something similar to what NED got against BUF a couple weeks ago. and when i run my pass heavy scheme, i get a similar outcome as we saw in BUF vs KCY in the last playoffs (for those who don't remember, BUF passed the ball on almost every meaningful play).


to me, your last point above about tying the engine to the player file is one of the key points. in the extreme run or pass scenarios above, it makes sense that for a static engine i'll most often get 4 rushing TDs if I plug 4 Quenton Nelsons in my OL, or 5 TDs if I have 4 Calvin Johnsons.

[MIJB#19]

Originally Posted by Red Zone

Jim,
I am sorry that some folks took this "game" to such an extreme that you felt uncomfortable. They would be the minority and they should be ashamed of themselves.
While most of us have some part of the game that we wish was different, or better, I think most would agree that this is the only text sim Football Game that has been worth playing for the last 20 years or so.
I know I have personally enjoyed it for 17 years and still play in 7 MP leagues and commish an 8th.
A Big Thank You and Here's wishing you and your Family a Very Healthy and Happy Holiday Season.
I Hope you are doing well in your new Endeavors.

Well said, Red Zone. Well said.

[bgbob1967]

Originally Posted by Solecismic

I have a love/hate relationship with MP. Love, because this is a small community with which I have a lot in common. If things were different, you'd be real-life friends. Hate, because a small percentage of this small community got really, really angry with me and wanted me out of business. I've had people posting pictures of my home online. I still get nasty notes on social media from time to time. Ben had personal threats against him as well just for moderating. I'm not sure I would include MP if I had to do it again.

Been playing in MP for a couple years roughly and I can't believe idiots would threaten people over a game.

Some people need to get a life !!!!

Thank You for what you do to keep FOF going.

[SFL Cat]

Originally Posted by Solecismic

I have a love/hate relationship with MP. Love, because this is a small community with which I have a lot in common. If things were different, you'd be real-life friends. Hate, because a small percentage of this small community got really, really angry with me and wanted me out of business. I've had people posting pictures of my home online. I still get nasty notes on social media from time to time. Ben had personal threats against him as well just for moderating. I'm not sure I would include MP if I had to do it again.

Situations like that are where you tell people to get out of their basement and get an effin' life.

[jcard]

Originally Posted by Solecismic

I've had people posting pictures of my home online.

Silver lining- I bet you got some good offers in this housing market...

[Gallifrey]

For me one of the best things about FOF is how it can feel like the old NFL.

Other football games fall into the usual trap of having stat engines. Of course the NFL today is a stat engine, where a QB with any skill at all can have a 300 yard passing day; WRs can run uncovered because lord knows what can happen if you touch them.

Yikes, I started to rant. Enough of that.
Great game Jim. Too bad you are out of the game, you made the best ones.
And maybe it is best you never did TCY2. That would have made me a recluse.

[Kcarr716]

Regarding the initial question, we see a lot of corners in the NFL that may be elite at one coverage area but lacking at others, as is especially seen when you see guys leave one system for another. An obvious example is Nnamdi Asomugha who was elite as a man coverage guy on an island in Oakland but struggled a year later in Philly.

As for the qbs that is a more odd set but I tend to think of the distance ratings as how far above that qb's baseline as described by the rest of the bars he is to that area. An example that somewhat fits what you described, though likely higher on every bar, would be drew Brees in his peak in new Orleans. He had the ball handling ability to be elite in the screen game and the touch to be the best in the league on the deep ball but didn't maybe have the arm strength to have the zip on the ball to rise above his, granted very high, baseline on some long throws or into the tightest windows on 5 to 10 yard throws