Need to develop a quick method of determining player value based on statistics (fantasy sports).
The solution should be fairly simple. Just need some heads on it and it will be fun.
There are 8 categories. Each category is worth on point. So you would think you could just do (cat1+2+3...+8)/games played. This final number will be their "quality rating" or "efficiency rating". However, this model does not work because in theory the player could get zeros in every column and a billion in one column driving his average way up. The problem there is each category is worth only one point. So his average would say he's the best value but in reality he gets you one category; one point.
How can we fix this?
I was thinking along the lines of some sort of weighted avg but each category is worth one point.
The goal is to get a somewhat accurate depiction of each players fantasy value with one final number.
The problem can probably be solved without reading the rest. But as bonus info:
Categories are: G A +\- PPP, shots, faceoffs won, hits, blocked shots.
Each category is worth an equal amount- one point.
How can I achieve one number that estimates roughly a players fantasy value based on these constructs?
Thank you very much
Just to clear up one question (since it's either ambiguous, or I'm reading it wrong) - this is rotisserie style? As in, you play an opponent each week, there are eight categories, your players accumulate totals in each category, and if your team has more in Category X than your opponent, you get one point? Then, is it all or nothing (if you win 5 of 8 categories, you get a "win"), or is it proportional to the number of categories (if you win 5 of 8 categories, you get 5 wins and 3 losses)?
One key here is that, for each category, the value of a player is going to be the incremental value beyond what could be achieved by the best freely-available player off of the waiver wire. So if you think that Sidney Crosby will score 50 goals, but the best free player would score 20 goals, then his marginal value of goals would be 30.
(Also note that what the "best freely-available player" will total is a function of league size and roster size - for instance, in a 14-team league, you'd expect lower baselines than in a 10-team league).
I don't think that would matter If a ranking method is used through.
Each player is ranked in each category and the total value of ranks/number of players in analysis (maybe 350). So the lowest value would be the best selection in theory. Then sift out who you don't want by hand.
Are there flaws in that method?
Any better ideas?
Drafts are this weekend.
I'm sure this could be way more scientific to appease this forums brains
I'm interested in the feedback.
I would even enjoy deep diving a real in depth value system at some time
I'm trying to figure out a similar sort of thing, but my system is a little different.
I take the pool of players that I'm sure will be drafted split them by D F and take all of their totals in the categories and find and average value. Then divide and determine their points and other totals compared to the averages.
And I get the ratio of a category compared to the average, then I sum them all and divide by the total number of categories.
Obviously it's far from being a super super accurate way to rank players, but it gives me a number that I know at least accounts for all my categories equally. In combination with other tools and rankings it gives me a usable list.
Edit-When I determine the ratio between their total in a category and the average which I've found limits the skewing of their number by performing very well in only one category. All the other categories where they are under the mean is assigned a negative decimal (1-ratio discussed earlier) and if its a flat 0 in a category they receive a -1.
I've found this system of assigning numbers keeps category specialists from having a value that's too high.
And I believe comparing them to the "average" player will provide far more information.
And since it was mentioned, I thought this up after watching moneyball.
Last edited by ResilientBeast: 09-19-2013 at 09:40 AM.