ceemes wrote:Mike,
While I generally like the idea of assigning values to the votes, I have been crunching the numbers and believe that the system you have adopted is unintentionally unfairly biased towards the cool/awesome categories and basically discounts the fail category. (Note: Calculations were done using the actual vote count during the time of this post and are subject to change as voting proceeds)
Using a spreadsheet, I plugged in the number using your current model and here are the results.
As you can see, using the current model, this bike scores 58% of the possible points, thereby putting it firmly in the Cool section even though the majority of the voters feel it is either uncool or a fail. Using the old first past the post or simple majority system of the past, this bike would be rated Uncool.
However, I think with a bit of tweaking we can address this imbalance by simply changing the weighting system as follows:
By assigning negative values to the uncool and fail categories, we can achieve more accurate analysis of the voting. As we can see from the example, some votes cancel out other, leaving a true score which shows that this bike actually has a negative acceptance rating of -42.53% that would place it firmly in the uncool category. This better reflects the actual votes cast by the members here, although it also discounts that more actual votes for fail have been cast.
For scoring purposes, I propose the following based on percentage of potential point scored:
Awesome: 51% to 100%
Cool: 1% to 50%
Uncool: -1% to -50%
Fail: -51% to -100%
If you like, I can send you my spreadsheet, all you need to do is plug in the votes and it will do all the work for you.
Chris.
All Mike needs is the average vote and then convert that from a scale of 0 to 3 to a scale of 0 to 100 ( a percentage). In your example this number is 40.5797101449 (Uncool)
This scale can easily be divided into 4 mnemonic categories, not dissimilar to letter grades at university:
[ 00, 25 ) : Fail
[ 25, 50 ) : Uncool
[ 50, 75 ) : Cool
[ 75, 100 ]: Awesome
Easy and no confusion and the number provides finer granularity when comparing bikes within the same category.
I think your error is in your "Weighted Vote Total" as a percentage of your "Total Possible Vote" instead of simply taking 28/69 * 100 which is mathematically exactly the same as Mikes method.
Consider what you did:
Weighted Average Vote = 28/23 = 1.21739...
Weighted Vote value = 28/23 * 100/3 = 40.5797...
Total possible vote = 3*23 = 69
Percentage of total possible vote = 40.5791 / 69 * 100 = 58.81%
When we substitute we find out what is going on:
( 28/23 * 100/3 ) / ( 3*23 ) * 100 = 58.811741
The numerator can be expressed as ( ( 28*100 ) / ( 3*23 ) ) and we can now note that the term ( 3*23 ) appears twice. That is in your calculation the total possible value of 69 is used twice because you are taking a percentage of a percentage which is why 100 appears twice as well as the maximum possible vote (3*23):
( 28 * 100**2 ) / ( 3*23 )**2 = 280000 / 69**2 = 28000 / 4761 = 58.81117 = Cool.
I think Mike has it right with:
( 28/23 )/3 * 100 = 28/69 * 100 = 40.5797% = Uncool.
As to the Indian, I vote uncool, but Awesome for Mike's new methodology.
) and the new math is horribly complex (and I'm the one that has to calculate it every week... hahaha) the new system really does allow for all the votes to count.
