need help with power ranking

[roondog248]

I have created a Advanced Pitching Database that compares any MLB starting pitcher's advanced metrics to the league average, and also ranks those metrics.
I have taken the rank of all the important metrics, and averaged them together for an overall ranking, but I know that this is not accurate by any means because it's not weighted.

In order to create an accurate power ranking,
I would have to assign point values to each metric in order to compare relative skill across a variety of otherwise unrelated sub categories.

I was told that I need to use a simple exponential growth curve that assigns between 0.0 and 1.0 in each metric.
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
The point scale needs to follow a y=x^2 graph.

I only need to use the metrics from 10 columns, so once I figure out how to find the point value for each metric, I would multiply it by .10 (for 10 percent) and add the entire row together, and then rank from that number.

I believe I need to use the =Growth function
I'm just unable to figure out how to execute this task.


Any help would be greatly appreciated.

Thanks.

 

[NARAYANK991]

Hi ,

I am not able to understand your requirement ; as far as I can see , you have performance data of players on various measures ; you are interested in combining 10 of these measures to get one consolidated measure.

In order to combine the 10 measures to get 1 , you are looking to weight the 10 measures in some fashion.

Can you explain why you thought of the GROWTH function ?

Do you have the weights you wish to assign to each measure ?

Are all the measures measuring performance in the same manner ? What I mean is for one measure , a higher value may mean better performance , while for another , it may be that a lower value may mean better performance.

So if one measure varies on a scale of 0 to 10 , another might vary on a scale of 20 to 70 ; do all the measures need to be reduced to have the same scale ?

Narayan

 

[roondog248]

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 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.....

"Okay, well 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."


1) I assumed the growth function had something to do with the simple exponential growth curve
2) I have 19 metrics total, but only need to use 10 of them (the ones I believe are most important) which (D5,E5,F5,K5,M5,O5,Q5,S5,T5,V5)
3) yes, some are higher the better some are lower the better.
(e.g. K/9 or strikes per 9 innings = higher the better, ERA or earned run average = lower the better)
4) I believe each metric needs to have an assigned value based on it's deviation from the mean, and that value should be between 0.0 and 1.0.
I'd assume there would be a way to control the ascending / descending issue, like in the =rank formula - 1 for ascending (largest number ranked 1) 0 or blank is descending in (smallest number ranked 1)

thanks