New and Improved FBG Expert Rankings (1 Viewer)

[rolliehouliranson]

Example, Shaun Alexander,

rankings 3-4-4-4-5-5-6-6-8-8-9-10-10-16

Median is 6

So you have 3-4-4-4-5-5-6 and 6-8-8-9-10-10-16

Median of each of those is 4 and 9

Interquartile range is 5, 1.5 of that is 7.5, add & subtract from 9 and 4

Anything lower than -3.5, and higher than 16.5 would be considered an outlier. So, he has no outliers.

 

[Doug Drinen]

It seemed like he arbitrarily did it by giving us an option to throw out highs/lows, but what if those values are legitimate? Why not just employ an official and tested means to do so? I don't mind an option instead of automatically doing it, becuase some don't agree with throwing out outliers, but some would at least like to see it without having to manipulate the data in our own spreadsheets.

From my standpoint, this thread's purpose is to ferret out any bugs that might be lurking in the app. I don't want to turn it into something as as the mean/median thread of awhile back. So I'll just offer a very quick answer that you may or may not find satisfactory.No test is going to tell you whether a particular ranking is "legitimate" or not. The automatic deletion of data points that satisfy a particular mathematical criteria isn't a practice that I've ever heard recommended.

I included the high-and-low-thrown-out average simply because people frequently request it.

 

[phthalatemagic]

Can we get deviation too, pretty please!I like to nominate high standard deviation players early as they tend to go for the max amounts. Leaves me more $ to play with....

Alright, this is for the seriously 1. Add this to end of the URL: &stdev=12. Mouse over a player's name.

Cool!Question - why did you create a function to throw out highest and lowest instead of performing a proper test for outliers?

Sorry, what is an outlier and what kind of tests do you have for it?

An outlier is a data point that may be considered bad data. Q test is a very simple way identify outliers that may be bad data. I don't know how valid it is, considering the rankings lists aren't linear, but it would be a nice option, and if a "throw out high/low score" is done, certainly Q test can be done too. Maybe I'm asking for too much...

 

[phthalatemagic]

It seemed like he arbitrarily did it by giving us an option to throw out highs/lows, but what if those values are legitimate? Why not just employ an official and tested means to do so? I don't mind an option instead of automatically doing it, becuase some don't agree with throwing out outliers, but some would at least like to see it without having to manipulate the data in our own spreadsheets.

From my standpoint, this thread's purpose is to ferret out any bugs that might be lurking in the app. I don't want to turn it into something as as the mean/median thread of awhile back. So I'll just offer a very quick answer that you may or may not find satisfactory.No test is going to tell you whether a particular ranking is "legitimate" or not. The automatic deletion of data points that satisfy a particular mathematical criteria isn't a practice that I've ever heard recommended.

I included the high-and-low-thrown-out average simply because people frequently request it.

OK I put that last response in before I read yours. Hopefully I'm not turning this into a nerdy math discussion.

 

[Doug Drinen]

I put that last response in before I read yours. Hopefully I'm not turning this into a nerdy math discussion.

No problem. I certainly don't object to the discussion itself; I'd just prefer to keep it out of this thread if possible.

 

[3nOut]

Cool!Question - why did you create a function to throw out highest and lowest instead of performing a proper test for outliers?

I am actually the one who requested excludiong certain rankers. I wanted this because I feel that one person can really skew a persons rankings if it is ridiculaously low or high. I wanted a more general consensus. For example in the redraft rankings, Reggie bush is ranked 35th overall and 17th RB by John Norton. Including him in the rankings places him as RB12, while excluding John places Bush at RB9. To me, thats a huge jump.

 

[rolliehouliranson]

I put that last response in before I read yours. Hopefully I'm not turning this into a nerdy math discussion.

No problem. I certainly don't object to the discussion itself; I'd just prefer to keep it out of this thread if possible.

That's my fault. No more math from me in this thread. Great work with the cheatsheet.

 

[Carter_Can_Fly]

Thanks, Greatly.

 

[3nOut]

Now with the ability to exclude different rankers. Includes sortable columns for mean, median, and adjusted average (throwing out the low and high) against selected rankers.

Thanks for everyone that made these suggestions earlier. Thanks also goes to Doug Drinen who programmed these changes.

Redraft Rankings

Dynasty Rankings

Rookie Rankings

Not the only one but one of many who suggested this.

I think I'll join you in (not often I can do that on this site ). I made these suggestions a while ago, glad FBGs took them to heart. Now can you guys create a atmyselfontheback: smilie?

 

[rolliehouliranson]

The automatic deletion of data points that satisfy a particular mathematical criteria isn't a practice that I've ever heard recommended.

:( Wow.

http://en.wikipedia.org/wiki/Outlier

 

[Doug Drinen]

The automatic deletion of data points that satisfy a particular mathematical criteria isn't a practice that I've ever heard recommended.

:( Wow.

http://en.wikipedia.org/wiki/Outlier

I'm not getting your point.

 

[LHUCKS]

Does Colin really like McCardell that much or is that a typo?

 

[Power Monster]

Rudi Johnson and Chad Johnson in the top 12 no thanks.......

 

[ookook]

I will echo Doug's sentiment about exclusion of outliers...it is a slippery slope.

The best way is to deal with "outliers" (aka bad data points) is to...collect more data so such cases are less inluential.

In this case, I would suggest providing an applet for board members to input rankings and then allow a ***ntion of how much wieght to apply to the aggregate members' rankings relative to the staff.

 

[Fred Lane]

I like the changes, and would like to make a recommendation.

For the ranking movement, I think it is better that downward movement be reflected in red and upward movement in blue. I'm more used to red movement as being bad for the player....

 

[inca911]

By showing the SD, one can gauge the need to look more closely at the individual rankings using whatever statistical treatment is desired. The mouse-over method is nice, but it is still difficult to easily see those players who are more "controversial". I think it's worth adding as a dedicated column.