i don't understand... why would underpredicting or over predicting matter?

150/100= 1.5

50/100= .5

distance from 1 in both = .5 =50%
right?

nvm 

changed my mind. see next post

wait what about this scenario

big christmass week:

prediction: 700k
real figure: 1000k

points method 1: 30
points method 2: 300
average: 155

----------------

quiet week 1:

prediction: 210k
real figure: 300k

points method 1: 30
points method 2: 90
average: 60

------------

Using method 2 or even averaging it with method 1 would make the huge christmas weeks much more important relative to the earlier weeks as the number of consoles you'll be off will be much greater. In the scenario above the christmass week is 2.5 times as influential to the final figure over 10 weeks (that is if you plan to add all the points up at the end instead of looking at the predicted 10 weeks total)

So I'd now say that it's best to just use method 1

Why isn't it just 1 point per percentage wrong?

predicted 50k
actual 100k

50% inaccurate= 50pts

predicted 99k
acutal 100k

1% inaccurate= 1pt

so If I take the average of everyones predictionsafter they have finallized em & then put mine up as the average, I'm bound to win muwhahahaha!

@theprof and non-gravity.

You have mis-understood.


First @ non-gravity:

You didn't work it out correctly:

-------

big Christmas week:

prediction: 700k
real figure: 1000k

points method 1: 30 (tick)
points method 2: 300 42.8%.
average: 155 (30+42.8) / 2 = 36.4... therefore you would get 36 points

----------------

quiet week 1:

prediction: 210k
real figure: 300k

points method 1: 30 (tick)
points method 2: 90 42.8 % (same as the other one)
average: 60 (same as the previous prediction)

------------

As method 2 does not involve converting 1k into 1% as you thought.... it is just dividing by the other number.... here:

Method 1:

(real figure - prediction) / real figure * 100

Method 2:

(real figure - prediction) / prediction * 100



Now onto theprof00

i don't understand... why would underpredicting or over predicting matter?

150/100= 1.5

50/100= .5

distance from 1 in both = .5 =50%
right?

---------

That is method 1 yes... but there is a different way of looking at it... which comes about because for people who under-predict, it is unlikely they will ever get more than 50 points from a prediction.... you have to predict as low as half of what the real figure ends up as to get 50 points.... over-predicting it is quite easy to get 50 points by being a little overconfident. Last year there were a number of predictions that were more than double what the real figure turnd out to be, so they gained over 100 points.... a feat which is actually impossible when under predicting unless you predict negative numbers.


With method 2... it is possible to get 100 points when both over and under-predicting.... if you predict either half of the real, or double it, then you would get 100 points.


That's all I was asking, because neither of the two methods is totally fair by itself, Method 1 is definately more correct though because it is comparing directly to the real sales, instead of to the predictions (see above post for the maths bit)

I should have thought you would be all for me trying to make it fairer when you have such high predictions on all 5 consoles.

I am fine with leaving it as it is though, it's a perfectly ok way to work it out and I don't need to do extra work.

That's not true, it is more likely to leave you in a mediocre middle of the pack position..... it should prevent you from losing miserably though.

ooooh I get it

so which way are you deciding on then?

dividing by the prediction, or dividing by the outcome?

overpredicting on all five consoles!?!?!?

You guys are predicting almost 20% drop in sales compared to last year!