Simulation Adjustments for 2023

[Goodell]

I hope to find some time this off-season to dig into the game simulator a bit more and ensure everything's working as intended or better ahead. If there are any things you've noticed in simulations you'd like me to take a look at, respond here or send PM.

Toward the end of the regular season going toward the playoffs, I did clean some things up that I'd had on a to-do list from last year's notes. So hopefully things ran a bit smoother later in the year. I modernized some of the decision making a bit toward modern NFL where they are more likely to go for it on 4th and short in more situations, and modified some of the decision making in some cases I'd noted before that I thought should be adjusted. Hoping to do more with that ahead.

QB passing yards was one stat that seemed consistently lower than I'd expect usually outside of sim shoot-out games. Part of that is understandable if there's a big blowout and the sim team wants to run out the clock more than pile up passing stats. A win is a win no matter the stats. Receiver stats also play into keeping QB numbers down sometimes. While the QB stats matter most, receiver stats also play a part in completion yard calculations. We've adjusted the time per play a bit the past few seasons and will take another look at that in hopes that we get our sim players as much opportunities as real ones get for their stats. May look into some later game make-up plays to roster backups to try to get in more yardage for QBs without getting depth chart receiver stats too far out of whack. Also could make sense later in the game as attempts wind down for RBs WRs also to boost per plays even more if players were further behind their stats and chances to make that up lowering.

While we have to keep touchdowns realistically tied to being within scoring range, we could look at possibly extending performance per play even more in cases where players weren't getting enough redzone chances to take advantage of their autoscore updates. Probably some adjusting needed there too in play calling to where if it's a run or pass should perhaps consider if there are any autoscores in that situation and push the decisions more in those directions toward those stats instead of considering that if a certain type of play is called for the situation.

Punting and kicking are a little tricky to work within the the sim because we're dealing with small sample sizes and game situations where sometimes it's worse to have a longer punt if it goes for a touchback instead of a shorter one pinning the team down more. But heads up to teams who don't carry or replace hurt kickers/punters that we're taking a closer look at those in the programming ahead to ensure it's as it should best be. I think realistically that should mean lots more misses for teams with poor kicker situations/updates outside of short ones, and perhaps more makes for teams with real kickers with real game stats who didn't have bad misses to account for with some reduction in randomness odds there. But will have to look deeper into kicking/punting stats and how we're doing and see what adjustments make sense.

We'd previously discussed and set some provisions in place to prevent against small sample averages unrealistically being exploited in the playoffs that use averages. Like a backup RB getting a 1-game week 17 chance to play when all the starters resting and puts up big stats in one game, we'd said before I think we'll at least require 2 games minimum for playoff averages so it would at least cut any extreme small sample cases in half. Had some complaints about backup QBs with lesser stats or garbage time stat padding and their sim performances on smaller updates. May look at minimizing small samples with higher game requirements for the average calculation and looking more at how players perform past their update stats to ensure it's as expected.

I never get around to it as it's complicated, but also a goal to get things setup to have the system run games itself for me on a schedule instead of me manually doing it for each game across six leagues.

Also hoping to enhance the actual game displays a bit as I can as time allows. Maybe getting more into further emphasizing the grade advantages that are helping dictate the results beyond just the fantasy numbers and scores people are seeing on their TVs. I was always most interested in just getting reasonable end results to make the roster-building game have some way of determining winners/losers over a season of simulations, but will also look at cleaning up some of the output more for those more focused on that. Could also perhaps have it show scores in progress one quarter at a time to have the suspense build as the game plays out more, or even make it more like watching an NFL game on some of the nicer gameplay updating sites.

Let me know if there are things you'd like to have checked out or adjusted in the game simulator before next season.

[bpboguta1483]

Possibly look into the default stats for QB's, if I recall correctly the generic stats are 14-24-200-1-1. Had a week 1 game with Garoppolo went 9-22 that day, it just didn't seem very realistic (not confusing him for Montana trust me ),

[Goodell]

Possibly look into the default stats for QB's, if I recall correctly the generic stats are 14-24-200-1-1. Had a week 1 game with Garoppolo went 9-22 that day, it just didn't seem very realistic (not confusing him for Montana trust me ),

Yeah generally we should look at defaults. But where a default update starts gets impacted strongly then by grades after that for what the sim does of course. If a QB with default update is going against an elite defense with max disadvantage, it severely impacts their performance by design where grades of surrounding cast matter (down 15% completions, additional turnover). Or if a default QB has an elite offense going up against a truly bad defense, they'll get max advantage for +15% to their completion percentage and extra TD toss perhaps. I would guess grades had an impact there where we reward teams with excellent defenses or elite offensive supporting casts.

[Heffay72]

Just a few things I've noticed about the sim,
In playoffs for qbs, they'll have an INT in their stats but won't throw one in the game.
Players on IR getting credit for sacks on their bye weeks. I've also noticed sacks seem to almost always happen in the 1st quarter/half of games and rarely in the 2nd half.
For QB adjustments, if say a qb has a 50% completion percentage and rates give him a .07 bonus. Would his pass % for the game be 57%? Because that doesn't happen alot. I've even lost % when I had the advantage. I may be mistaken on how it works though.
If you have 3 rbs with good stats, but your starting qb gets hurt and you have to use default stats on backup qb, can there be a way where you run the ball more than relying on the pass with a qb using the defaults?

[Goodell]

Thanks, I'll look over those.

If you have 3 rbs with good stats, but your starting qb gets hurt and you have to use default stats on backup qb, can there be a way where you run the ball more than relying on the pass with a qb using the defaults?

In theory it should factor in passing vs rushing stats in neutral game situations and lean more toward which has the stronger stats to work toward. However, game situations sometimes dictate things (for better or worse). If a team was down 21 points in the second half, would be kind of odd to run the ball every play way down in the game. In reality, if a team is down big and time getting shorter, they're going to start passing more.

In playoffs for qbs, they'll have an INT in their stats but won't throw one in the game.

Something to look closer at, but it might be that the QB in that sim game had far FEWER attempts. The way it should work is that if the real QB had 1 interception in 50 passing attempts, it doesn't start super forcing interceptions if the sim QB has only thrown half as many attempts in terms of picks per attempt ratio. But if a QB was behind their INTs later in the game with chances running down, it should kick up the odds of a pick more. Just probably doesn't do auto-picks until the attempts are closer to the real attempts. After all, 1 pick in 50 throws isn't the same as forcing 1 pick if a guy only threw 15 passes. The ratios are way off there, so the closer the attempts get the more it pushes auto interceptions matching the real game. But like a lot of things, something to look closer at how things are working, intended, and perhaps to be adjusted a bit.

On the QB adjustments, it works toward the calculated number on each play and has some of those figures listed in the play-by-play read outs. However, if a QB goes above their attempts, it then starts working more on default values so wouldn't be working toward that any further. Another thing that can happen is that for 95% of the game the QB is right in line with their per play expectations (give or take a play), but perhaps had an incompletion or two that it would have made up for with balancing completions to work back up toward it's expectation but the clock or opportunities with the ball ran out to get those completions. But for the most part, it should be working toward that adjusted percentage. If it's right at it, the next play maybe rolls the dice on an incomplete to drop it below the expected average. Then being below it, the next pass should have higher odds of being complete to make up for it to keep it toward that expected figure. It goes a little up or down throughout the game where if it gets too low it should bounce back up and if gets too high should slide back down. All toward the expected end goal, within a range of a play or two of being right around that average most of the time going a little up and a little down as the game goes along. But sometimes the game runs out of time or chances to throw the ball when it's a little more up or a little more down before those balancing plays got a chance. But in those cases, a couple of throws ago it should have been right around being on the money. It doesn't start with the end result and work backwards. It starts at 0 and goes one play at a time toward goals but doesn't just output the exact same result every time (a 4 yard per carry average doesn't mean the back only runs 4 yards every time he gets a carry). If a player got a 7 yard rush and was above his average, then his next carry likely would be less to balance that out back toward the average again. But if the game ended before that balance achieved, it might be a little up or down, but shouldn't really be beyond being at that expected average just a play or two before.

Thanks!

[jerrydlux]

I would agree with the QB adjustment issue. I have had games with a max advantage that has not played out. I think this is something to address because those max advantages are things we have done in building our teams to make up for lesser QB play.

So if I dont have a QB who throws it 25+ times a game I am less likely to get the advantage and the goal of building your team should be solely around getting the best QB.

Now, I do believe QBs drive the NFL and if Josh Allen goes down or Mahomes goes down...those teams arent as great as they should be. But a team like the 49ers was built in a way that could handle their qb going down...and I am not ready to say Brock Purdy is a star.

But it is pretty telling that Mahomes, Brady and Allen were in the majority of our Super Bowl sims.

Is there a way to adjust the matchups and actually take into account each players rating? Not sure exactly how but to get Qbs and specialists mixed into some type of equation.

[Royce R]

I would agree with the QB adjustment issue. I have had games with a max advantage that has not played out. I think this is something to address because those max advantages are things we have done in building our teams to make up for lesser QB play.

So if I dont have a QB who throws it 25+ times a game I am less likely to get the advantage and the goal of building your team should be solely around getting the best QB.

Now, I do believe QBs drive the NFL and if Josh Allen goes down or Mahomes goes down...those teams arent as great as they should be. But a team like the 49ers was built in a way that could handle their qb going down...and I am not ready to say Brock Purdy is a star.

But it is pretty telling that Mahomes, Brady and Allen were in the majority of our Super Bowl sims.

Is there a way to adjust the matchups and actually take into account each players rating? Not sure exactly how but to get Qbs and specialists mixed into some type of equation.

I'm guessing the sim hasn't changed a ton since I last played. But I did win a championship with Kerry Collins surrounded by a solid team. So to me the grade advantages always seemed to work like they should.

[Goodell]

RE: max advantages
They apply whether a QB throws 1 sim pass or 60 sim passes. On every pass play it considers the QB's adjusted completion percentage which is based upon his stats + grade adjustments. So from pass 1 it's factoring in grade advantages/disadvantages in his expected completion percentage.

Lots of talk about backup QB performances and related things recently, and as with lots of things each off-season things we should always re-evaluate. Where I stand, we strive to be an overall roster building game. Unlike fantasy football, to me it should matter a ton how good you are at building an offensive line and defense on your sim roster. Sim games, to me, should not be decided 95% by QB only, otherwise we're wasting a lot of time with a roster building game and grades for offensive lineman and defensive players if they shouldn't matter much to the results. It could just be a QB draft game instead, but it's intended to be all about building a complete elite all-around team with high grades all over the field and tons of strong depth to survive a long season to reward the best roster building sim GMs. If the very best elite of elite sim team with the highest grades across the board had their star QB go down to injury, I think it would be logical to expect their offensive performance to go down without star QB but that if they had a competent (by solid grade or stats) backup QB that they logically should be able to still win games as top teams do in reality with backup QBs frequently (and even won super bowls with backup QBs on strong teams). I don't think it should be if your starting QB gets hurt, you loose almost all the rest of your games no matter how strong of a roster you built otherwise. To me, we're all about building the best overall rosters here. We can't ignore that QBs are super important in reality, and should be here too. But reality is that other positions also matter to team strength and good teams have won with both somewhat lesser and backup QBs. Have to ensure we have the right balance of QB vs supporting grade importance, and that teams can win both ways as happens in reality.

We'll take another look this off-season at default values for players who get into the game (such as backup QBs sometimes) who don't have NFL stats in a real game to use, and make sure it's setup as we best think it should be. More on that ahead.

[Knighty Knight]

We'll take another look this off-season at default values for players who get into the game (such as backup QBs sometimes) who don't have NFL stats in a real game to use, and make sure it's setup as we best think it should be. More on that ahead.

There should be a distinction in the simulation between a rostered backup QB and the generic "backup QB" used when you don't have a backup on your team.

Use the Kendall Hinton stats from 2021 a baseline for the generic backup QB adjustment:

https://www.espn.com/nfl/story/_/id/304 ... nishes-1-9

[Goodell]

For default values for QBs without NFL game stats in a given week, we could adjust our current programming slightly to increase the number of levels of performance differentiation. For example:

QB grade over 7 (89 madden):
16-for-25, 275 yards, 2 TD / 1 INT // 64%, 17.2 ypc

QB grade over 6 (80 madden):
14-for-23, 225 yards, 2 TD / 1 INT // 60.1%, 16.1 ypc

QB grade over 5 (70 madden):
12-for-22, 180 yards, 1 TD / 1 INT / 1 FUMBLE // 55%, 15 ypc

QB grade under 5 (69- madden):
10-for-20, 140 yards, 0 TD / 1 INT / 1 FUMBLE // 50%, 14 ypc

No QB on roster:
9-for-19, 125 yards, 0 TD / 2 INT / 1 FUMBLE // 47%, 13.9 ypc

There's some danger, though, if numbers get too extreme that it helps teams tank if they intentionally get a QB in their depth chart that matches a very low grade who isn't going to play most likely in reality and expect they'll go all year here throwing no sim TDs and hoping to intentionally lose games.

We'll do more, I think, in the programming for the sim coach to have more flexibility for QB changes even looking beyond the depth chart to guys on the roster to sub in at some point if they give him the best chance to win.

We have to be cautious about how such grades get multiplied over a full season and not having it be too unrealistically extreme. So would probably put in some programming provisions where WRs might be able to run the ball into score more and not be stopped from scoring by the script if 0 TD in the update in the extreme cases so that it's possible that team might score sometimes as it would be unrealistic if they never could otherwise all year in those situations. We could do a randomizer in the extreme cases also where it's 80% chance of 0 TD update and 20% chance of 1 TD so some games they'd have a 1 TD update out of multiple game opportunities.

Of course those default values, just like any real game stats, get adjusted further based upon supporting cast grades versus strength of opposing defense grades to be adjusted +/- 15% and possible additional TD or INT added if large grade differences in matchups.