2011 Community Projections - August Update [Scoring Method Changed]
The season is now 5/6ths complete! Only 27 games to go in the regular season As a reminder, I will be donating a Nellie Cruz bobble head from the 6/20 promotion against the Astros to the overall winner of the projections (or second place if the LSB Average wins).
NOTE: The scoring method has changed!
Initially I used a basic RMSE method. I was taking each projected value and comparing it to the actual results and determining a percentage difference. After making all the comparisons I would average everything together to come up with a final score.
However, as the season progressed, it became obvious that Julio Borbon and Tommy Hunter were skewing the results. Everyone was off massively on those two players and it came down to who had the best projection on those players alone. If everyone is within 5% on Elvis but mix in the best Borbon as 250% off compared to 600% off, Borbon would dominate the average.
To compensate... each player projection now has an individual score associated with it. The person who makes the best projection on a player gets 100 points. The worst projection receives 0 points. You also receive 0 points for any player you did not make a projection for. This now allows for the leaderboards to include people who only made some projections but not all projections.
So while best gets 100 points, worst gets 0 points... everyone in between gets a weighted value based on how close to the best or worse you were. If you are within 98% of the best projection, you get 98 points. It is not linear... you could have 10 people all get 99 points if they all were that close to the best projection.
With this new method there is a possible 1500 total points. 1100 for the position players and 400 for the pitchers. I removed Tommy Hunter from the scoring entirely since he played so little... every projection was pretty much worthless.
Our new overall leader is seanathan with an amazing 1103.54 points out of 1500. I say amazing because he never made a Michael Young projection and received a 0 for him. So his 1103.54 is actually out of 1400 points based on the projections he made.
In the Google Spreadsheets linked below, commenters who made all projections have a score in green... if you failed to make at least one projection your score is in pink. The "Percent" next to your score is the percentage of possible points you received based on how many projections you actually made.
Position player projection data has been tabulated in this Google Spreadsheet - Batters.
Pitcher projection data has been tabulated in this Google Spreadsheet - Pitchers.
The overall leaderboard can be found in this Google Spreadsheet - Leaderboard.
Here are the top 3 and bottom 3 projections [bottom 3 only includes those who made all 15 projections] as well as the LSB Average, ZiPS & Marcel. Individual player ranks can be found in the respective spreadsheets.
| Overall Combined | All Position Players | All Pitchers | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
10 comments
|
2 recs |
Do you like this story?
Comments
Wait!
On the pitchers, did you not include Tommy Hunter!?!?! I should get some credit for thinking he’d be fairly worthless this year…
Go Rice Owls!
Would be a Matt Harrison fan, but I only like superstars
Tommy had a 2.93 ERA in 15 innings.
It all boils down to who picked the least amount of innings.
You projected 120 IP… everyone except for very few thought at least 100+. Seems worthless to include such a small sample of meaningless data.
"I'll text AJM a picture of my junk if he wants to use it" by ab03 on Jul 7, 2011 9:30 PM CDT
I'm just grouchy
I’m not a huge fan of this type of analysis. I can see the argument for removing the effects of a Hunter or Borbon, since if weighted the saem they’ll skew everything.
But at the same time, what about a guy like Andrus, who everyone seemed to be within 15% RMSE on? It seems silly that guessing within 15% on Andrus gets me 0 points, whereas roughly the same error it got me 98 points for Harrison… I mean, I’m proud of my Harry prediction and all, but that doesn’t make much sense to me.
What if you had a system where for each player you get max(100-RMSE, 0) points? So you can’t get negative points if you miss by more than 100%.
Go Rice Owls!
Would be a Matt Harrison fan, but I only like superstars
For a competition though...
I think it’s a matter of how you compare to your peers. Elvis is easy… everyone knew roughly what he’d do.
Harry was hard. Lots low points on Harry.
"I'll text AJM a picture of my junk if he wants to use it" by ab03 on Jul 7, 2011 9:30 PM CDT
Wait... I see more of what you're saying...
Since Elvis is easy… the range of 100-0 seems off compared to Harry…
Hmm.
"I'll text AJM a picture of my junk if he wants to use it" by ab03 on Jul 7, 2011 9:30 PM CDT
what if you set 100 as the "best" prediction
and then everyone else is based on that – not down to zero, but relative to the best.
So Elvis the best is 1.61% error. That = 100. Mine was 14.64%. That = 87 points (100 + 1.61 – 14.64)
For Harry, the best is 14.04% error. That = 100. Akalahar had 189.13 %; That = 0 points (100 + 14.04 – 189.13 = negative —> 0)
Go Rice Owls!
Would be a Matt Harrison fan, but I only like superstars
basically there should be some way to put in magnitude of differences
even while correcting for population differences.
Or use LSB average to normalize the scores. For all players you get (your RMSE – LSB average) as your score.
Go Rice Owls!
Would be a Matt Harrison fan, but I only like superstars
Very cool..
Pretty impressive LSB outscoring zips
my better is better than your better.
by rangerjake on Aug 29, 2011 3:05 PM CDT via mobile reply actions

by 































