2020 RULES: Playoff Simulation Efficiency and Automation

[Goodell]

While I was simming the playoffs this year (well every year) thoughts go through my head about how can I make this complicated task easier. And it's not really a matter of just having more help. It's just a very detailed process that involves a high level of attention to not make mistakes during the most critical games. Because of that, it also makes it harder for me to dive into playoff simulations knowing I need a big chunk of uninterrupted time I can be completely focused on it. I think I may have a better way, so long as most are okay with the change.

POLL SUMMARY: So my suggestions for league input are:

- Change how we do multi-game averages to not be based upon complicated individual game tracking, but change to playoff round averages that apply to all in that round using an average.

- Change using real playoff game update options so that if a player's real playoff game is going to be used that it's used across all leagues.

These changes are recommended for efficiency as well as possibly taking our game simulation to enhanced levels of fair results. Here's why:

PLAYOFF SIMULATION NOW
First, here's how we do it now -- and I really like our multi-game average setup. If a QB, for example, averages .49 interceptions per game, we don't round down to 0 and have him throw 0 picks across 4 playoff games. Realistically, we give him 0 picks the first playoff game, but his multi-game average would give him an interception in the next game because over two games he'd be expected to have interception statistically. You can see those multi-game averages on the player profile page.

But that's where it gets complicated with teams able to use real performances also. Say a team uses their QB's real playoff game wildcard weekend and averages for all others. If they advance and use all averages next round, I then use 2nd game multi-game averages for most of their team except the QB who would get his first game average. Once we get deeper into the playoffs, and teams have different real game updates each round, you might have all sorts of players on the team on different multi-game averages. So I have to carefully keep track of that and do each player individually checking each one for how many averages used so far. Across all the leagues, a player might have different average updates from league to league.

DIFFERENT WAY: Multi-Game Average Adjustment
One way to really simplify that would be forget keeping track of individual players and how many playoff averages they've used. Right now all players have 4 multi-game averages showing on their profile page. Instead of it being 1st game, 2nd, 3rd and 4th games, we could just say all those 1st game averages are now wild card round averages. All 2nd multi-game averages are divisional round averages. All 3rd multi-game averages for conference championships. All 4th multi-game averages for super bowl average update.

So if your playoff team had a bye, when you played your first divisional round game you wouldn't be using 1st game averages any more. You'd be using the 2nd round divisional averages. That would apply to all teams across all leagues. Brady doesn't have a 1st game average in one league, a 2nd game average in another league from a wildcard team, etc. Your opponent who won a playoff game isn't into their 2nd game multi-game averages while you're on your first. Both teams are using 2nd game multi-game averages for the 2nd round.

If you used a real game update for your star QB in the wildcard round and then averages for all the next round, that QB isn't on his 1st game average like now but everybody including him would use the 2nd game 2nd round average.

It makes it less complicated to track, not having to double-check each individual for their personal averages used so far, and easier to sim. It also creates an opportunity to possibly automate simulating games if the update for player X is the same across all leagues for that week and not something that has to be manually managed individually one player at a time.

DIFFERENT WAY: Real Playoff Game Options

But there is one additional complication keeping all players in each playoff round from having the same update: real playoff game options.

Currently I give teams the option to say if they want to use a real playoff game update instead of the multi-game average. I like that also. It allows real football stories to impact us also. If a player becomes the biggest story of the NFL playoffs carrying his team to a championship, we want to bring that reality in here also. It also makes the playoffs a little less predictable than just averages alone where results might be predetermined without real update impacts. I also like that our sim teams benefit from having their players making big impact in winning real meaningful games. It also makes it complicated to track when a player's real game is used in one league but not another, and impossible to automate when such manual changes needed.

So how do we keep that real playoff game impact and excitement but make simming the games more efficient or even automated where we need the same update across all leagues for a player?

One way might be still having GMs speak up and nominate their player should have a real playoff game used, but having the league making the stat comparison and applying the use of either an average or real game update to that player across all leagues. So Derrick Henry wouldn't get a 200-yard real playoff game update in one league but not another where the GM didn't post and would get an averaged game now. The league would say that player's real game stats this round are obviously better than the averaged game this round, so we'll use his real playoff game update for all teams this round.

There are pros and cons to all of these suggestions, but the end result if implemented would be a standard set of playoff data to use in game simulations that's the same across all leagues each round, that isn't complicated to track or maintain, and allows us to quickly sim the results and possibly automate the process.

AUTOMATION GOALS:

One of my goals is automating game simulation, so I don't have to manually click a button for each game (16 games a week most weeks x 5 leagues = 80 games each week simmed). I don't do as much manually checking and tracking of course during the regular season as is required in the playoffs, but still takes time for both regular season (because so many) and post-season (because so detailed player tracking).

If I could get things setup where I load the game data and then tell the system it's ready to go, it could push those simulation buttons for me if I can get it setup right (including standard playoff data). Less work for me. Less human involvement in simming games for transparency. Less life impact on delays. More reliable timing of game results posting. And possibly better quality results. It hasn't been popular when asked before, but if I got to where the system was simming games instead of me manually, we could have it deliver more fair results where your playoff fate for a superior team didn't come down to an oddity upset that might have happened 1 out of 10 games. In theory if the system is simming games, we could have it sim a matchup multiple times behind the scenes (say 2 out of 3 to still keep some upset possibilities) and better reward the best teams. You'd know that the result was less fluky. It wasn't a one-in-a-million oddity, but something that when the matchup simulation happened the result is what should happen most of the time. I know most like keeping more unpredictability and upsets with odd results sometimes on a one roll of the dice result, but I have an interest in having the game simulator deliver the best quality results with as little flukes as possible.

[larry linke]

The one thing I have wondered why does a GM get the choice of the real life average or the playoff game. I would like it better if the had to declare prior to the playoff game. As for above whatever is easiest for you

Larry
Minnesota AFFL

[Goodell]

The one thing I have wondered why does a GM get the choice of the real life average or the playoff game. I would like it better if the had to declare prior to the playoff game. As for above whatever is easiest for you

Larry
Minnesota AFFL

As to the reason, it's intended to be an ADVANTAGE for you to build a team with players who are excelling as winners in reality leading their teams to championships. If we did it before the results, it would be more like a guessing game than advantage having winning players.

There is a lot of injury risk having players on your sim team who are still in the playoffs, but those teams with winners should be rewarded not punished. We don't want it to be an advantage for teams that have a lot of players on crappy real teams sitting on the sofa come playoff team on losing teams. Those players don't face extra injury risk. If we did it differently, it would be to your advantage to avoid winning players on winning teams who'd have extra playoff injury risk and just build with players from losing teams that don't accomplish much but will be 100% healthy during sim playoffs at home.

Instead, we know having lots of playoff winners on your roster is an extra risk for injury, so we give you an extra advantage also to building a team with winners performing well in the most important games.

[JonC]

I think the average vs. real makes a lot of sense. From a transparency standpoint, if we were able to (maybe we already are??) see what the average stat line looks like for each round, making an across all leagues call for real vs. average becomes a whole lot easier.

From an automation standpoint, this data is old, but it's interesting. http://www.wagerminds.com/blog/nfl-hand ... nges-2496/

Essentially, I think upsets are a good thing when teams are closely matched. I don't like to see them in games that are 10+ point spreads. I wouldn't want to see an 0-11 team beat an 11-0 team in our sim outside of the very historically rare or injury defined time.

[kevinl]

Need to change- if a player goes AWOL and madden doesn’t give said player a 0 then you should step in! If this is real life as possible, Antonio Brown should’ve made NO impact in the playoffs.

[Goodell]

Need to change- if a player goes AWOL and madden doesn’t give said player a 0 then you should step in! If this is real life as possible, Antonio Brown should’ve made NO impact in the playoffs.

That's definitely something to think more about. Probably should be it's own thread (maybe I'll try to move it or think about ideas for another topic/poll on the subject). His impact last season, though, is over-rated I think. At least during the regular season when he had 0 game update stats for most weeks.

I know the results are the game box scores and people see players doing things there that can make it seem different, but the game simulator doesn't know anything about Antonio Brown. To it, he's just a 6.5 graded player (with grade penalty for being out of the league like all not on a team after week 4) with no stats during most regular season games. Name doesn't mean anything to the game simulator, just stats and grade. The passing game in the simulator is massively driven primarily by the QB, not the dependent position WRs. WRs in my opinion help most through grade by boosting QB grade-adjusted averages, but once the simulator starts spitting out plays and deciding complete or incomplete and how long, etc. it's mostly all about the QB stats.

If a QB had 5 TDs for example in his NFL update, but none of the receivers on the sim team had TD receptions in theirs, it's not going to have that QB not throw any TDs once he's in the red zone. It's a TD pass. Who catches it if nobody has any receiver TD catches is mostly random (or somewhat tilted to better graded receivers with more stats to get that day). It may look like a big play by the WR catching a TD pass and getting the credit, but that's 100% just because of the QB stats driving that and needing that predetermined score to be tagged to somebody when nobody had the reception scores to "earn" it -- but the QB did and he's going to get it and some receiver luckily has their name snagged.

His impact in the playoffs is a different deal, though. He would be more impactful there because he did have a season average. Based on very few games with very good results, so it's a good average to work with.

It's not something I want the league to do to subjective make decisions that can totally help or hurt my or other teams not based upon actual rules and treating players differently than all others. Like arbitrarily saying we're treating this one player differently than everybody else. Only they are subject to not having their season average count. Only they are subject to a different grade penalty than every other player in their situation out of the league. We can't have that. Has to be a universal rule to apply to all equally and without subjectivity. There were stories I believe in the real NFL about teams giving AB tryouts before the playoffs. He wasn't on any suspended lists or anything I know. So it's a difficult situation with lots of gray area. It's too bad a team didn't sign him toward the end of the season, because we'd know for sure if he'd have actually played or the NFL would have stepped in and suspended him or something. If so, we'd have our definitive answer and would have been able to mirror that.

It's something, along with a few other situations, to possibly look at and develop some new rules toward, but I'm not really for arbitrary fly by the seat of the moment decisions within a season that can drastically impact my team or other teams outside of the rules we'd established for strictly following actual NFL suspended or other lists only and not making up our own sim suspensions.

[kevinl]

I meant mostly for playoffs. I understand he didn’t have an update in regular season besides one game. That one game gave good averages for playoffs.

[Goodell]

I meant mostly for playoffs. I understand he didn’t have an update in regular season besides one game. That one game gave good averages for playoffs.

Yeah. I think one of the only kind of standard changes we could make and universally apply to that situation would be to change how playoff averages done for everybody.

I forget if we used to do it this way for a bit years ago or just discussed these two options, but I think at one point early on we did playoff averages based upon impact across 16 games for everybody. So it wouldn't be total stats divided by total games, but instead total stats divided by 16 games for everybody.

So a one big game season wouldn't turn into a nice playoff average per actual game played, but instead would just be a tiny average divided by small impact across 16 total games.

But that's a big change effecting all other players if mostly for the rare odd situation. It also means a fantastic player who missed part of the season but is generally spectacular when healthy would also take a huge penalty and be only a fraction of their usual healthy self here if everybody's average is over 16 games.

Say Mahomes missed 8 games next season early on. He comes back healthy and is lighting up the league as usual with 300-yard, 5-TD games everywhere. Come playoff time for averages and his typical game suddenly gets cut in half (he's suddenly a bad 150 yard passer by 16 game average) because he missed half the season. Suddenly he's way less effective here than he really is.

So either way, you're going to have some cases that create less desirable results. Our way now, you'll get some small sample players that have high season averages from few games and bigger playoff impact than maybe they should. The other way, you'll probably get more players who missed some time be much less effective than they arguably deserve to be when healthy.

I suppose another alternative might be throwing out any season averages based upon small samples. If a player played less than 4 games, for example, maybe their entire season stats are ignored because of the small sample size. That would limit those AB cases, but also maybe hurt other cases where a backup RB suddenly got his staring opportunity over the last 3 games of the season and ran 100+ yards in each game emerging as a star only to have his impact completely wiped out by the small sample rule where he'd have no average in the playoffs here.

No matter which way we go with averages, there's probably always going to be some odd case that ends up not as we'd like. I think we just have to go with the best way that works best for most situations most of the time and consistently apply that average method to all.

[kevinl]

Yeah that’s tough! I just think there needs to be something in place to eliminate a situation like that. Players that are getting arrested or doing drugs and don’t have a team should be eliminated from the sim all together IMO! Also, tough to explain Trent Williams situation as well but that ended up working itself out

[Goodell]

It's probably not popular because I think some might feel that way too about some of the unusual situations, but I don't think I ever want it setup where I decide which players are treated differently than all others subjectively. In part because I play the games and some of these might be my own players (AB is on my roster for example) that I'd make subjective calls on that could be seen as unfair advantage, but mostly because I just want a nice cut-and-dry, 100% clean exact line rule that's not subject to any gray areas or debate if a decision will go one way or another.

In the very early days of the league I think there might have been more subjectivity, to where even teams could petition me to examine cases where maybe a player went onto IR in reality but there were some comments in the press that if his team was winning more maybe he wouldn't have been put on IR after all so I could decide subjectively which players were really on IR and which didn't have to be here. Same maybe with grades people thought were bad and should be reviewed. We ended all that quickly and not something I want to get involved with examining individual cases and making subjective calls in gray areas for or against teams including my own. For fairness, objectivity, clarity and workload.

Much better, at least in my opinion, with zero subjectivity with hard and fast unbreakable conditions applied to all equally without exception.

All players are either suspended or not based upon being on an actual NFL suspension list, not our own subjective call on their availability.

For holdouts, I get some of that too, but by design our sim contracts are not the same as whatever NFL contract situation a player finds himself in. Our sim players largely play for different organizations, often with completely different contracts than reality. Whatever contract issue they have in reality is not applicable IMO to different sim contract. He may hate his old NFL deal with a bad organization and coach he hates in reality, but in the AFFL for example maybe he just signed the largest deal ever for a player at his position with a new organization. I don't think we'd say he wants to hold out from his new massive sim league contract situation too. Maybe a top draft pick holds out, but here our sim team has him under contract already. Do we invalidate that player's sim contract signed months ago keeping him off the field because he hasn't signed his NFL deal yet? For contracts, we have our own sim contracts that aren't dependent upon NFL contracts or NFL contract status. A player could be a free agent in reality but under contract playing here or the reverse. We have our own signings and terms different from real situations. Whether they are happy with NFL contract or not arguably shouldn't have anything to do with how they'd feel about a completely different sim contract in different simulated situation.

With the new CBA changes I was reading through briefly today, I think we may also see less holdouts with some changes there.