A Fantasy League Minus The Randomness?

Carson Cistulli, contributor-extraordinaire at my favorite baseball site Fangraphs, wrote me to ask an interesting question. He wanted to know how someone would set up a fantasy football league using commonly available scoring options that would remove most of the luck and leave mostly the results of skill. Personally I think the randomness of sports, including fantasy sports, is part of what makes them so compelling. But draining a lot of the randomness from a fantasy league would be an interesting experiment. Here's what I suggested:

"First, turnovers are extremely random. The correlate weakly from week to week and even less from season to season. Especially on the defensive side of the ball. I would exclude turnovers from your scoring (including TDs directly occurring on turnover plays).

Second, most leagues give points for both TDs and yards, but TDs are the end result of all those yards. A "scavenger-back" who comes in on goal line plays and gets an easy TD does not deserve 6 points. On the other hand, the yards near the goal line are the toughest in football. So I would either remove TDs from your scoring , or have a 1-pt bonus for them.

Next, special teams are very random. FG kickers get so few opportunities from the various attempt distances that it is almost statistically impossible to tell good kickers from bad kickers. Evaluating FG kickers after a full NFL season would be like evaluating MLB players' batting averages after 4 or 5 games into April. I'd exclude special teams in all its forms: FGs, return yds, etc. The only thing that is consistent is a kicker's kickoff depth, but that's not something you can score in most leagues.

Strictly speaking, the point value of a yard varies with the situation. A one yard gain on 3rd and 1 is huge, but a one yard gain on 2nd and 9 is almost worthless. But fantasy scoring systems don't take that into account, so we have to use a general estimate. Generally, one yard is equivalent to 0.06 "net expected point advantage" in the world of advanced football stats. That equates to about 1 point per 16 or 17 yards gained. If your system allows a bonus for 1st downs I'd consider adding a small point bonus for them.

Fantasy leagues usually give less weight to passing yards. It's typically 20 rushing yds for a point and 40 or 50 passing yards for a point (or something proportional, such as 10 yds rushing  and 20 yds passing). This is because if passing and running got the same scoring, QBs would become so important they'd dominate the outcomes in your league. But that's the reality of the NFL. Passing ability is critically important in winning games, and running ability is much less so. I'd consider using equal weights for running, passing, and receiving yards for your players.

Technically speaking, sacks cost about one half point of "net expected point advantage." If there is a way to make sacks count as .5 points, I'd use that as a penalty for the QB and as a bonus for the defense. Otherwise, go with one point.

For defenses in general, I'd stick with either yards allowed (-1 pt per 16 yds) or with the scoring scheme most leagues use for points allowed (10 pts for a shutout, 6 points for 3-7 pts allowed, etc.) Safeties are so rare (and so interesting to me) that I'd make them worth even more than 2 points. You get 2 pts for the score, plus you get the ball back for your team, which has value--usually about 1.5 expected points after the free kick. You could make the case they are really worth 3 or 4 pts. On the other hand, just like touchdowns, the plays that create the safeties are often just straws on the camel's back. The plays that put the opposing offense's back up against the goal line is just as important if not more important as the safety itself.

The bottom line is that turnovers, special teams and touchdowns are the most random things in the NFL. I'd stick with yardage as much as possible. But that might make for a boring league, with very few comebacks and upsets."

  • Spread The Love
  • Digg This Post
  • Tweet This Post
  • Stumble This Post
  • Submit This Post To Delicious
  • Submit This Post To Reddit
  • Submit This Post To Mixx

25 Responses to “A Fantasy League Minus The Randomness?”

  1. Dan Rosenberry says:

    One of the reasons that QBs and receivers might have a lower value per yard is to avoid double counting. This works if you're trying to make the model equal total value (or marginal value).

    This could be taken a step further and give a negative point value for attempts (running/passing) to make the baseline closer to replacement level (or average, or whatever desired), rather than 0. This does have the problem of tracking WR attempts as they're nonstandard.

  2. billsfan says:

    As the gentlemen at Football Outsiders say: "holding on to the ball is a skill; recovering it isn't."

    So you can still give points for Forced Fumbles to the defense, and subtract all fumbles (not just lost) from the offense.

    One of the commenters on FO a while back tried to reverse-engineer a fantasy scoring system that would match their advanced statistics. The gist of it was that for each passing or rushing attempt, you *lose* points that you would then gain back if the play were positive (i.e. certain yardage threshold, first down, or score.)

    For example, a running back would lose a point for each rushing attempt, but score a point, and thus break even, for gaining an average number of yards on that attempt, and actually score points if it were an above-average carry.

    The conclusion, I think, was that it would very boring to play like that. Randomness makes fantasy football fun. Part of that is watching your rival deal with Neil Rackers putting up 6 field goals.

  3. DSMok1 says:

    Brian, I ran a fantasy league last year using similar principles last year. I counted negative points per attempt and per target, then positive per yard. I counted touchdowns a few points to account for the nonlinear yard value near the goal line. Unfortunately, I couldn't count first downs properly (ESPN wouldn't do that).

    I counted negative points for fumbles on offense, positive for forced fumbles, but didn't include recovery as valuable.

    The league is viewable here: http://games.espn.go.com/ffl/leagueoffice?leagueId=332764

    Basically, I tried to adapt your research to a fantasy perspective. The difficultly was balancing/crediting WR/RB/QB; I don't think I did that quite properly.

  4. DSMok1 says:

    "The conclusion, I think, was that it would very boring to play like that. Randomness makes fantasy football fun. Part of that is watching your rival deal with Neil Rackers putting up 6 field goals."

    On the contrary: week-to-week variation was at least as great in my league as in a more standard league. Why? Well, I was counting yards above replacement or something close to that. Week-to-week, there is PLENTY of variation in that. So a player who got 80 yards and 1 TD on 15 carries one week and got 70 yards and 1 TD on 8 carries the next week scored quite a bit more in week 2.

    Here's a sample game with those settings, showing the tremendous variation: http://games.espn.go.com/ffl/boxscorefull?leagueId=332764&teamId=1&scoringPeriodId=15&seasonId=2009&view=scoringperiod&version=full

    Drew Brees went 29/45 for 298 yards, with a TD and an Int (I do count off for them), but with 2 fumbles. Those 2 fumbles and the low YPA (6.6YPA) were enough to push him just below 0(-6.9, to be exact) for the day. On the other hand, Andre Johnson's 196 yards and 9 catches on just 10 targets netted him a ridiculous 47 points.

  5. Anonymous says:

    What about something radical, like using WPA (or EPA?) and eschewing any other stat? I'm sure there are major problems, but has anyone else thought about it?

    Also, I've never liked that RB's are the most important position in fantasy football. Why should Addai get taken before Peyton?

  6. Anonymous says:

    I've tried to create a custom league based on these values in Yahoo. Here is what I setup for scoring. I also factored in -3.75 for an INT (60 yards, 16 yards = 1 point). 1 point TD bonus for any TD. No kickers.

    http://football.fantasysports.yahoo.com/league/blackswan

  7. Brian Burke says:

    billsfan-That's essentially the same model that Dave Berri uses (for football and basketball). The player is debited for each attempt (shot, pass, or run) and then credited for success. It's not a fantasy system, but it works the same way.

    You could always just do a WPA-based league too.

  8. DSMok1 says:

    @Anonymous: WPA is boring in baseball at least--there is much less of an element of balancing a team. If all players get WPA from the same thing--say, HRs, then in real life it wouldn't be as good of a team: you need people on base. It's more fun in general to manipulate multiple categories.

    In football, though, I don't think there are competing multiple things that need to be done like in basketball and baseball. In basketball, a team must have rebounders, not just scorers. In football: there are different positions but the goal is yards. So WPA would probably work. I guess you could compare TDs, Yards, and First Downs as competing objectives that all need to be met...

    @Anonymous: We can't view that league; it's not public for viewing.

  9. Florida Danny says:

    Brian...

    "The bottom line is that turnovers, special teams and touchdowns are the most random things in the NFL."

    I agree wholeheartedly about special teams. However, correct me if I'm wrong, but I've found that the number of TD passes by QBs are relatively consistent from year to year, i.e., not random. Also, isn't it pretty much received wisdom in the advanced stats world that, although, fumble recoveries are random, forced fumbles aren't? In fact, didn't you post something to that effect on here?

    I'd say keep TD passes. Also, instead of using fumble recoveries & fumbles lost, use defensive forced fumbles & offensive fumbles (both excluding fumbles on special teams, of course).

  10. Brian Burke says:

    I agree. TD passes, etc are not excessively random. Scoring for both TDs and yards double counts the yards that have been gained to get to the end zone.

    It's true that forced fumbles are less random than fumble recoveries, but not by much. I have more data now than I did when I did the original research on that. But the bottom line is that very few if any fantasy sites would have a scoring option for forced fumbles.

  11. Florida Danny says:

    Good points on all 3 counts. Thanks for clearing that up.

  12. Brian Burke says:

    By the way, I love your screen name--Florida Danny. I'm picturing a mob family with 2 Dannys in it.

    "Well just send Danny over to collect da money from dat no good bastard."

    "Which one, Short Danny?"

    "No, send Florida Danny."

  13. Florida Danny says:

    Does it add to the mental image if I tell you I am, in fact, of Sicilian heritage?

  14. Adam Davis says:

    It seems like your commentary about turnovers contradicts your commentary about safeties.

    "Safeties are so rare (and so interesting to me) that I'd make them worth even more than 2 points. You get 2 pts for the score, plus you get the ball back for your team, which has value--usually about 1.5 expected points after the free kick. You could make the case they are really worth 3 or 4 pts. "

    I don't think anyone denies that a safety can be a game changing event or that they are rare, but aren't they also essentially random? Certainly we can't forecast with any accuracy which defenses will score more safeties. And if we can't forecast it, then aren't we back to scoring items that are essentially random, just like fumbles?

  15. DSMok1 says:

    "But the bottom line is that very few if any fantasy sites would have a scoring option for forced fumbles. "

    ESPN does.

  16. Burt Likko says:

    I play in a scoring-only league. This significantly changes the dynamics from what's under discussion here. Quarterbacks become the most valuable players by a large degree. That doesn't make the rest of the picture irrelevant, but the other elements of a player's performance are important only to the degree that they help the player score.

    Thoughts?

  17. Anonymous says:

    1) RBs are more valuable than QBs in fantasy more because of the stupid rule to start 2 RBs and only 1 QB than because of the scoring.

    2) Seems like we are confusing two things: one is taking out randomness and the other is calibrating fantasy values to more accurately reflect the game. Safeties and turnovers are random (Brian did mention that safeties are very random), but hugely influential.

  18. Brian Burke says:

    Cool that ESPN does that.

    Nothing contradictory about rare and random. Solar eclipses are rare, but they're not random.

  19. Zach says:

    Another thing: Get rid of head-to-head matchups. YES, it makes it more exciting. But nothing is worse than losing your matchup when you score the second- or third-most pts that week, only to have faced off against the #1 scoring team that week, or missing the playoffs after scoring in the top three all season long.

  20. zlionsfan says:

    myfantasyleague.com also tracks fumbles forced (and first downs as well). It's pay-only, though. My league uses it because it has good support for auction leagues.

    Another suggestion I've seen with respect to removing luck from fantasy football is not having bench players; you get points for everyone on your roster each week. Of course that's talking about a different type of luck (from a fantasy perspective rather than an in-game football perspective).

  21. Adam Davis says:

    "Nothing contradictory about rare and random. Solar eclipses are rare, but they're not random."

    Yes, I understand that, but we can forecast an eclipse with absolute certainty. Can you forecast safeties with any accuracy? What is the correlation factor on safeties from week to week or season to season? I have not researched this, but I imagine that it is minuscule.

    The original challenge posited here was to create a scoring system that would remove most of the luck and leave mostly the result of skill. For fantasy purposes, "skill" would probably be defined as the ability to accurately forecast a player's/team's stats.

    Are you saying that it is reasonable to expect that someone can accurately forecast safeties?

  22. Brian Burke says:

    Adam-That's not what I wrote. Read it again. The goal is not to make a league that is entirely predictable. Why be so argumentative?

  23. Unknown says:

    "Another thing: Get rid of head-to-head matchups. YES, it makes it more exciting. But nothing is worse than losing your matchup when you score the second- or third-most pts that week, only to have faced off against the #1 scoring team that week, or missing the playoffs after scoring in the top three all season long."

    Zach,
    One way to combat this common complaint is to use a site that allows more than just one matchup each week. That way you can just schedule everyone to play everyone else each and every week. So in a 10-team league, each week each team is playing 9 games, one against every other team. No playoffs should be necessary - best overall record at the end of NFL season wins the league championship.

    Tom

  24. Marver says:

    "Adam-That's not what I wrote. Read it again. The goal is not to make a league that is entirely predictable. Why be so argumentative?"

    I think he's raising a valid point, not being argumentative. A safety occurs due to bad field position, likely as a result of offensive ineptitude (either by way of failing to convert on 4th and goal or punting from midfield) or a great special teams play. The safety itself is just a tackle for loss, or (an even luckier) fumble...essentially just as 'random' for the defense as a 1 yard RB TD, just in lower volume by nature of the number of chances afforded to defense at creating a safety.

  25. Dave says:

    I think a great stat to look into for fantasy football is how much does the previous season's success contribute to the following season in terms of both certain players and certain teams assuming all the players and coaches return?

Leave a Reply

Note: Only a member of this blog may post a comment.