Player Floors / Ceilings (1 Viewer)

[Ryan99]

So, I enjoy doing stats stuff, and I had compiled player projections from 5 different sites to compare. And it occurred to me that the spread of the projections gives an indication of how confident the expert community is as a whole about a player's upcoming performance. And if you think about the players performance as being drawn from a normal distribution (for those who may not know, this is the name of an actual probability distribution), and the projections as also being drawn from that distribution, than the average of the projections is the average of the player's output and, more importantly for this, the spread of the projections is the spread of the player's possible output (assuming no injury).

So, given an average and a standard deviation, you can calculate confidence levels for a player's fantasy points. You can calculate something like a player's floor by looking at average - 2*standard dev, and a player's ceiling by looking at average +2*standard dev. Sorting by these gives the players with highest floors and highest ceilings. The reason I chose 2 standard devs is that it corresponds with 97.5%. So, 97.5% chance to better is my definition of 'floor', and 97.5% to do worse is my definition of 'ceiling'.

Generally, you want to draft high floor starters and high ceiling bench players. With that in mind, here are the players that have the highest floors (for starters) and ceilings (for bench) compared to their average (which correlates with ADP). Remember, these are players with high floor / ceiling relative to where they're being drafted (obviously Rodgers and Calvin Johnson have the highest floors and ceilings, but they're already being drafted first so it doesn't matter). Also, this doesn't take into account injury, so if a guy would be affected by injury (either negatively or, with someone else's injury, positively) that's not included. Oh yeah, starter means anyone you'd start in 1 QB, 2 RB, 2 WR, 1 TE, RB/TE/WR flex, so top 12 QBs, 36 RBs, 36 WRs, 12 TEs (most people wont flex a TE).

High Floor Starters

QB: Peyton Manning

RB: Michael Bush (maybe not a starter), Jamaal Charles, Shonn Greene, Steven Jackson, Willis McGahee, Mark Ingram

TE: Jermichael Finley, Tony Gonzalez

WR: Hakeem Nicks, Jeremy Maclin, Torrey Smith, Reggie Wayne, Miles Austin

High Ceiling Backups

QB: Jake Locker, Carson Palmer

RB: Rashard Mendenhall, Rashad Jennings, Kevin Smith, Mike Tolbert, David Wilson, Jahvid Best, Mike Goodson, Evan Royster

TE: Jared Cook, Dallas Clark, Marcedes Lewis, Greg Olsen, Anthony Fasano

WR: Robert Meachem (maybe a starter but really high ceiling), Denarius Moore, Michael Floyd, Sidney Rice, Greg Little, Darrius Heyward-Bey

Did my explanation make sense, and to these guys make sense as high floor or high ceiling guys? Any names that stand out as ridiculous or interesting?

 

[gheemony]

Really like the concept and implementation. When I see a large standard deviation between expert projections for a player, I flag them for greater study. Is it because the experts differ on expected usage (e.g., some experts may expect Smith, Jennings, and Royster to start, while others do not)? If so, what do I think the usage will be? Is it because a young player is expected to break out (Locker, Little, DHB) or an old player is expected to decline? Is a player coming of an injury which makes it difficult to project their performance?

In short, unless you have the time to study each player in detail, calculating standard deviation helps you identify the players to study in great detail (don't waste time on Rodgers, Calvin). Then decide what you're comfortable with and draft low risk players early and high upside players later.

The problem with this is that the "expert" projections are often their "mean" projection? That is, they probably have a range of possible outcomes for each player, but they can only express an average of those outcomes. So averaging these mean projections don't give you a complete picture, but it's the best that you can do with the info.

In an ideal world, a set of projections would give you probabilities for ranges. For example, Rodgers may be 85% top 6 QB (above average starter), 5% top 12 QB (starting QB), 5% top 24 QB (bench), 0% below 24 QB (replacement player). This would be more meaningful for a player like Jacquizz Rodgers.

 

[Knowlt]

Playing devil's advocate here....you may just be compounding the error in all of the projections, and miss on everyone

 

[Zeff]

So, I enjoy doing stats stuff, and I had compiled player projections from 5 different sites to compare. And it occurred to me that the spread of the projections gives an indication of how confident the expert community is as a whole about a player's upcoming performance. And if you think about the players performance as being drawn from a normal distribution (for those who may not know, this is the name of an actual probability distribution), and the projections as also being drawn from that distribution, than the average of the projections is the average of the player's output and, more importantly for this, the spread of the projections is the spread of the player's possible output (assuming no injury).

So, given an average and a standard deviation, you can calculate confidence levels for a player's fantasy points. You can calculate something like a player's floor by looking at average - 2*standard dev, and a player's ceiling by looking at average +2*standard dev. Sorting by these gives the players with highest floors and highest ceilings. The reason I chose 2 standard devs is that it corresponds with 97.5%. So, 97.5% chance to better is my definition of 'floor', and 97.5% to do worse is my definition of 'ceiling'.

Generally, you want to draft high floor starters and high ceiling bench players. With that in mind, here are the players that have the highest floors (for starters) and ceilings (for bench) compared to their average (which correlates with ADP). Remember, these are players with high floor / ceiling relative to where they're being drafted (obviously Rodgers and Calvin Johnson have the highest floors and ceilings, but they're already being drafted first so it doesn't matter). Also, this doesn't take into account injury, so if a guy would be affected by injury (either negatively or, with someone else's injury, positively) that's not included. Oh yeah, starter means anyone you'd start in 1 QB, 2 RB, 2 WR, 1 TE, RB/TE/WR flex, so top 12 QBs, 36 RBs, 36 WRs, 12 TEs (most people wont flex a TE).

High Floor Starters

QB: Peyton Manning

RB: Michael Bush (maybe not a starter), Jamaal Charles, Shonn Greene, Steven Jackson, Willis McGahee, Mark Ingram

TE: Jermichael Finley, Tony Gonzalez

WR: Hakeem Nicks, Jeremy Maclin, Torrey Smith, Reggie Wayne, Miles Austin

High Ceiling Backups

QB: Jake Locker, Carson Palmer

RB: Rashard Mendenhall, Rashad Jennings, Kevin Smith, Mike Tolbert, David Wilson, Jahvid Best, Mike Goodson, Evan Royster

TE: Jared Cook, Dallas Clark, Marcedes Lewis, Greg Olsen, Anthony Fasano

WR: Robert Meachem (maybe a starter but really high ceiling), Denarius Moore, Michael Floyd, Sidney Rice, Greg Little, Darrius Heyward-Bey

Did my explanation make sense, and to these guys make sense as high floor or high ceiling guys? Any names that stand out as ridiculous or interesting?

 

[FDC]

injury risk allstars = high floor?

hmmmm

 

[Da Guru]

Playing devil's advocate here....you may just be compounding the error in all of the projections, and miss on everyone

Projections really mean nothing after the opening kickoff. It is fun to speculate but every year after one week everything changes.

 

[FUBAR]

injury risk allstars = high floor? hmmmm

seems like it. Sure, if Peyton stays healthy he has a pretty high floor but his injury risk is the key here.

 

[Ryan99]

injury risk allstars = high floor? hmmmm

Yes, that does seem to be the trend. I assume this is because when people do projections they don't adjust for injury unless a guy is super injury prone (Vick). So these guys are gambles due to injury, but not gambles to just not play well. I'd rather have a guy who gets injured than a guy who just stinks because its much easier to make the decision to cut and replace the injured guy, or at least not play him.

 

[moleculo]

Really like the concept and implementation. When I see a large standard deviation between expert projections for a player, I flag them for greater study. Is it because the experts differ on expected usage (e.g., some experts may expect Smith, Jennings, and Royster to start, while others do not)? If so, what do I think the usage will be? Is it because a young player is expected to break out (Locker, Little, DHB) or an old player is expected to decline? Is a player coming of an injury which makes it difficult to project their performance?In short, unless you have the time to study each player in detail, calculating standard deviation helps you identify the players to study in great detail (don't waste time on Rodgers, Calvin). Then decide what you're comfortable with and draft low risk players early and high upside players later.

this is my take as well. I also average projections from as many sources as I can find and have stdev calculated from them. I think there is value in knowing that, but haven't been able to find anything useful for stdev outside of flagging guys who should be considered more closely.

 

[FDC]

injury risk allstars = high floor? hmmmm

Yes, that does seem to be the trend. I assume this is because when people do projections they don't adjust for injury unless a guy is super injury prone (Vick). So these guys are gambles due to injury, but not gambles to just not play well. I'd rather have a guy who gets injured than a guy who just stinks because its much easier to make the decision to cut and replace the injured guy, or at least not play him.

Perhaps your logic made some sense during the execution of your plan but the conclusion doesn't make any sense. I don't think anyone would argue a team full of injury risks has a high floor.

 

[Ryan99]

Perhaps your logic made some sense during the execution of your plan but the conclusion doesn't make any sense. I don't think anyone would argue a team full of injury risks has a high floor.

Indeed, although not all the high floor guys are injury risks. Shonn Greene, Steven Jackson, Willis McGahee, Gonzo, and Reggie Wayne should all be very consistent starters for their ADP.