Philaflava.com ::

If there was a predictive scoring model out there that created a FICO like score for a player based off certain variables to determine if you should sit or start them a certain day/week/etc; would you use it/buy it?

For example, you have two QBג€™s for week 4 match-up and you can only start one, Tom Brady (@ PITT) or Tony Romo (vs CAR). Youג€™re not sure which to sit or start, so you go to a website and input the players name to generate this score (0-100).

This score would then be created on an empirical level using attributes that impact a player and teams performance:

Time of year/season
Weather forecast
Turf/natural grass
Team facing (run D rank, pass D rank, offensive ranks, etcג€¦)
Players past performance
Injuries effecting upcoming matchup
coaches
Etcג€¦

The score is then delivered within seconds and boomג€¦Brady scores a 77 while Romo scores an 84 for this weeks matchup.

If such a tool existed, that looked at all of these variables, weighted them, and created this performance scoreג€¦do you think something like this would be used in the fantasy world? Create an industry standard score that leverages statistical sports data that is updated daily through various data feeds setup with respective vendors.

I know predictions/rankings exist today, but is it done on such a level where statisticians and analytical modeling is used? To me, it seems as if most of this stuff is simply objective and decided by 10 guys being in a room debating who they think should start over who and what not.

Season long subscriptions, or transactional pricing could be available. Dream big and pump this score into ESPN, Yahoo!, CBS Sports, etc and your cakingג€¦

What do you think?

lets do it

we need to recruit...

1. Statisticians
2. SAS Modelers
3. Programmers/IT team
4. Operations Team
5. Market Planners
6. Sales Team

we also need capital...data contracts...super computers...office space

But I'm down.

if you build it, they will come

Dwight Strawberry wrote:we need to recruit...

1. Statisticians
2. SAS Modelers
3. Programmers/IT team
4. Operations Team
5. Market Planners
6. Sales Team

we also need capital...data contracts...super computers...office space

But I'm down.

nbd

even if you had a model that worked for simple start/sit decisions, i feel like it would need to do more to get people to open their pockets. there is already so much free fantasy advice content out there. i would also look into recruiting a couple designers for user research/interaction design/visual design.

this is the future and its always better to be make things as interactive and realistic as possible. i know guys who are much more committed to fantasy then our phila group.

i would actually invest time and energy into working on this. we're in the stone age of fantasy and things need to evolve.


not to derail thread, but our baseball scoring system is outdated. judging guys on average and rbi is archaic. I been playing in sabermetic leagues and its so much more fun, The batting mechanic is pretty straightforward. Using run values we basically have wOBA.

The pitching mechanic starts with FIP, but the spread on pitchers from #1 to #10. we change this to run values as well, adding a point for IP and it works well. I also included a fractional point to try to balance the value of RP with SP. I believe I have it balanced well with the top reliever ranking about #45 in all pitchers.


the phila auction league next year should be run value based if we really want to see who has the best players

even if you had a model that worked for simple start/sit decisions, i feel like it would need to do more to get people to open their pockets. there is already so much free fantasy advice content out there.

1 this
2 i think groups like cbs (where my league is right now) already have this service. they dont really tell you what theyre doing but they have a bunch of different numbers that are probably entirely useless. but ive only done a cursory look around
3 one future problem youre going to have is the game planning problem that ive noticed with cbs. two years ago, when stafford was out and shaun hill was throwing to tron, the lions faced the jets - and cbs (in their laudable recognition that the stuff they have requires narrative)....
A) predicted that tron would get like 13 pts or something but
B) hadnt really figured into that number that shaun hill was not going to throw in revis direction
C) didnt really factor that revis is a shut down guy who pretty much shut tron down that game despite what qb was doing anyway into their number (which is important because the narrative was like 'well its gonna be a tough match up and tron may not do that well - this despite estimating 13 pts for the guy)
D) insert myriad problems here

and another thing is that this past year cbs came out with some new stat bullshit but it didnt really factor in that stafford was throwing to tron in double and triple coverage, and so tron was routinely putting up 18+ pts when cbs would predict like 13, but anybody who was paying attention wouldve laughed at that number at some point because the gameplan was so obviously based on getting tron the ball at every possible opportunity

and therein lies your problem if youre trying to come up with something unique and worthwhile, in my opinion. youre going to want to develop some sort of system for estimating gameplanning and situational play in a variety of contexts and then relate that to player performance. and thats a lot of time and motivation.

however, the challenge is intriguing and i do think its possible to build something. and if its volunteer time, id be fairly interested in contributing with the limited time that i currently have.

just my 2 cents.

Thanks for the feedback guys. I really think I'm going to put together a business plan for something like this in my down time.

First thing I better do is try and secure the web domain and brand name.

I did verify that the company who makes all of the baseball, football, basketball, and hockey reference sites...Sports Reference LLC, does sell their data. http://www.sports-reference.com/data_use.shtml

This could be fun!

Dwight Strawberry wrote:Thanks for the feedback guys. I really think I'm going to put together a business plan for something like this in my down time.

First thing I better do is try and secure the web domain and brand name.

I did verify that the company who makes all of the baseball, football, basketball, and hockey reference sites...Sports Reference LLC, does sell their data. http://www.sports-reference.com/data_use.shtml

This could be fun!

I hope you do well if you do do it, but no matter what it is still going to end up being a bunch of guys guessing and arguing how good or bad something is. How else you going to say how much an injury is going to affect someone or how good a defense really is?

naturalborn103 wrote:

Dwight Strawberry wrote:Thanks for the feedback guys. I really think I'm going to put together a business plan for something like this in my down time.

First thing I better do is try and secure the web domain and brand name.

I did verify that the company who makes all of the baseball, football, basketball, and hockey reference sites...Sports Reference LLC, does sell their data. http://www.sports-reference.com/data_use.shtml

This could be fun!

I hope you do well if you do do it, but no matter what it is still going to end up being a bunch of guys guessing and arguing how good or bad something is. How else you going to say how much an injury is going to affect someone or how good a defense really is?

It will have to be based off numbers and historic data. For example, an equation that weights the number of yards rushing a certain D has allowed YTD and last season, plus you can incorporate some kind of 'match-up' equation that if your WR1 is facing Revis, then Revis becomes Questionable on Wednesday, the score will increase for the WR due to Revis's probability of not playing. Also, if Revis were to be called Doubtful or Out, then the 'match-up' equation can be trained to look at the next CB on the depth chart who will be starting in Revis's place.

Dwight Strawberry wrote:

naturalborn103 wrote:

Dwight Strawberry wrote:Thanks for the feedback guys. I really think I'm going to put together a business plan for something like this in my down time.

First thing I better do is try and secure the web domain and brand name.

I did verify that the company who makes all of the baseball, football, basketball, and hockey reference sites...Sports Reference LLC, does sell their data. http://www.sports-reference.com/data_use.shtml

This could be fun!

I hope you do well if you do do it, but no matter what it is still going to end up being a bunch of guys guessing and arguing how good or bad something is. How else you going to say how much an injury is going to affect someone or how good a defense really is?

It will have to be based off numbers and historic data. For example, an equation that weights the number of yards rushing a certain D has allowed YTD and last season, plus you can incorporate some kind of 'match-up' equation that if your WR1 is facing Revis, then Revis becomes Questionable on Wednesday, the score will increase for the WR due to Revis's probability of not playing. Also, if Revis were to be called Doubtful or Out, then the 'match-up' equation can be trained to look at the next CB on the depth chart who will be starting in Revis's place.

Yea, but there is so much more things that can't be computed. Like a lot of personal match ups where certain type of players give certain teams trouble. Like big WR murdered Pats the whole year, but fast ones did not. Or if a teams 2nd DB was horrible and you have teams 2nd WR. Or if a teams relies on one WR and they are facing Jets or they pass to everyone vs. Jets (which would matter for QB. Teams who force to one guy will do horrible vs. Jets due to Revis). This is also does not even mention different style of plays and how they match up. For relying how defense did prior to that, you can not do that without looking at who they played.. And of course who they played is ranking is based on who that team played.

This type of stuff will always be a lot of guessing involved and can't be done from a formula.

I am not trying to be an asshole, but just trying to think of any reason this could not work, which I always think is the best way to find out if a idea is good or not.

yahoo definitely does this, for a couple years now i've started whichever FA defense that was projected for the most points and have done well with it.

though i will say, with the proliferation of real $ fantasy sites (where you bet however much per week versus a random's team), i'd be surprised if people already haven't modeled this.