roondog248
New Member
- Joined
- Aug 21, 2014
- Messages
- 7
I've created a Advanced Pitching Database that compares any MLB starting pitcher's advanced metrics to the league average, and also ranks those metrics.
[FONT=Verdana, Helvetica, Arial, sans-serif]I have taken the rank of all the important metrics, and averaged them together for an overall ranking, but It is giving me inaccurate results because it's an inaccurate way to 'Power Rank'
[/FONT]So, I got some of this information from an online friend who has a website that measures sports analytics. He's an architect that understands math/excel beyond a level that I could ever comprehend.
I think I understand what needs to be done, I just can't figure out how to get there.
here is a portion of our conversation.....
"Once you decide what metrics are important and how important each one is relative to the other,
The best way to power rank is assign point values to each metric.
Greater deviation from the mean, more points.
I use a simple exponential growth curve that assigns between 0.0 and 1.0 in each metric.
Add up all the points and you get one power ranking number. Then you rank that number.
X axis is deviation from mean Y axis goes from 0.0 to 1.0
Greatest derivation from mean = 1.0 Worst deviation = 0.0
Graph looks like y = x^2 between those points.
Then you can solve for Y for every X and get your point value for that metric.
This method "cleans up" your data and converts it into a number that is more universal.
0.0 to 1.0, with the best at 1, worst at 0.
Let's say you have 10 metrics, each worth 10% in the overall power ranking,
I take the point value for that metric, multiply by .10 (for 10 percent) and add them all together.
Use that number to rank from.
The key is the conversion of your metrics into a point value between 0.0 and 1.0 that is most descriptive. I like my point scale to follow a y=x^2 graph, and my x to be deviation from mean.
It produces a number good for comparing relative skill within a specific data set in a form useful for comparing relative skill across a variety of otherwise unrelated sub categories."[FONT=Verdana, Helvetica, Arial, sans-serif]
[/FONT]
After searching the web and posting in a few forums, I'm still lost as to where I need to start.
Just need to be pointed in the right direction.
Any help would be greatly appreciated.
Thanks,
JR
[FONT=Verdana, Helvetica, Arial, sans-serif]I have taken the rank of all the important metrics, and averaged them together for an overall ranking, but It is giving me inaccurate results because it's an inaccurate way to 'Power Rank'
[/FONT]So, I got some of this information from an online friend who has a website that measures sports analytics. He's an architect that understands math/excel beyond a level that I could ever comprehend.
I think I understand what needs to be done, I just can't figure out how to get there.
here is a portion of our conversation.....
"Once you decide what metrics are important and how important each one is relative to the other,
The best way to power rank is assign point values to each metric.
Greater deviation from the mean, more points.
I use a simple exponential growth curve that assigns between 0.0 and 1.0 in each metric.
Add up all the points and you get one power ranking number. Then you rank that number.
X axis is deviation from mean Y axis goes from 0.0 to 1.0
Greatest derivation from mean = 1.0 Worst deviation = 0.0
Graph looks like y = x^2 between those points.
Then you can solve for Y for every X and get your point value for that metric.
This method "cleans up" your data and converts it into a number that is more universal.
0.0 to 1.0, with the best at 1, worst at 0.
Let's say you have 10 metrics, each worth 10% in the overall power ranking,
I take the point value for that metric, multiply by .10 (for 10 percent) and add them all together.
Use that number to rank from.
The key is the conversion of your metrics into a point value between 0.0 and 1.0 that is most descriptive. I like my point scale to follow a y=x^2 graph, and my x to be deviation from mean.
It produces a number good for comparing relative skill within a specific data set in a form useful for comparing relative skill across a variety of otherwise unrelated sub categories."[FONT=Verdana, Helvetica, Arial, sans-serif]
[/FONT]
After searching the web and posting in a few forums, I'm still lost as to where I need to start.
Just need to be pointed in the right direction.
Any help would be greatly appreciated.
Thanks,
JR
