Support NeoGAF

apujanata · Jan 18, 2008

My long time ago suggestion, which I would like to propose again :

me long time ago said:
To get a better prediction aggregate, instead of using all prediction to arrive at aggregate value, I propose that we have a NeoGAF League aggregate. The purpose of this is to see whether the NeoGAF league can arrive at better aggregate result, compared to simExchange, NeoGAF Aggregate and any other "wisdom of the crowd method"

NeoGAF League consist of the Top 30 previous year member. For 2008, that mean Top 30 2007 member. For 2009, that mean Top 30 2008.

Of those Top 30 member, every rolling month, their weighting factor will be modified up and down, depending on their previous month result. The purpose to made the prediction of members who are consistenly accurate more weight compared to those member who are hit or miss.

At the start of year, every one have weight value of 6. Every month they predict well, the weight value increase by 1. Every month they predict inaccurately, their weight value decrease by 1.

Example :
Member Apu & Donny2112. Before January 08 prediction, Apu & Donny have Weight Value (WV) = 6. In January prediction, Apu is not doing well, while Donny is doing well. Therefore, before Feb 08, Apu's WV became 5, while Donny's WV is now 7.
Once member's WV became zero, it means he is no longer counted toward League Aggregate, unless he manage to do well in the subsequent month, and managed to increase his WV into positive range again

What I have not determined yet is how to define doing well. My original thought is for them to be in the Top 30 percentage of prediction. What I mean is if, for example, January 08 have 250 participant, the definition of doing well means that League member has to be within the Top (30% x 250 =) 75 of point calculation (not raw data calculation).

If, for example, apu only managed to get position #90 on january 08, he is not doing well, and his WV will decrease by 1. If he managed to get position #60, he is doing well, and his WV will increase by 1.

I would like to hear your opinion, on whether this suggestion has merit or not, and whether there are some aspect of it that need to be changed to better achieve the purpose.

I repeat : The purpose of this NeoGAF League is :
- to have an alternate (hopefully better) aggregate value to do comparison with simExchange and NeoGAF own aggregate.
- to made 2008 prediction more interesting, since getting into Top 30 will have more impact to those who achieve it. Hopefully this will increase participation even more.
- If it is proven more successfull, we can have a better estimate of future NPD #.

tanod · Jan 18, 2008

GAF league on Simexchange

Jesse2040 · Jan 18, 2008

tanod said:
GAF league on Simexchange

Ok I see where you are going with this, but you couldn't compare yourself to The simExchange, logically it can't be done.

Picture this:
The simExchange is like a stock market that moves up and down every day. The NeoGAF estimates are like predicting "earnings per share" of a stock offered on the exchange just before earnings are released. So you really can't compare.

Look at it as The simExchange setting the market expectations. Essentially, that's what their figures are, expectations. Sometimes it's high, low, or right now. If hl2 ps3 is expected to sell 100k on TSE and it comes in at 50k, it does not mean TSE was wrong, it means people expected it to sell more.

And I think it would be great to do what you are suggesting, just to see how accurate the top gaf members can get, but comparing those results to a open market like TSE just wouldn't make sense.

The simExchange is (1) a tool for the industry and (2) a fun game to play.

But I do enjoy the comparisons every month and hope the gaf league on TSE kicks some butt this month!

Keep up the great work and congrats on the neogaf npd estimates, they were impressive!

apujanata · Jan 18, 2008

tanod said:
GAF league on Simexchange

It didn't work. Example :

Code:

65 icebergisonfire Agent Icebeezy 1,000,000.00 0.00% 
66 DarkPrince Dark Prince 1,000,000.00 0.00% 
67 wsippel wsippel 1,000,000.00 0.00% 
68 ICallItFutile ICallItFutile 1,000,000.00 0.00% 
69 Spire Spire 1,000,000.00 0.00% 
70 borsdy borsdy 1,000,000.00 0.00% 
71 Raw64life Raw64life 1,000,000.00 0.00% 
72 LaMeRz LaMeRz 1,000,000.00 0.00% 
73 mhwchung $h@d0w 1,000,000.00 0.00% 
74 Kazenone Kazenone 1,000,000.00 0.00%

All those member mentioned above has join GAF league on Simexchange Before Jan 1, 2008. Up until yesterday, none of them has done anything (witness that their networth is exactly 1 Million DKP).

As Jesse2040 explained clearly, simexchange is different than our prediction.

Sometimes, even if you know that the stock is overpriced, you might not be able to do anything about it, since your short capacity has been exceeded, which will not happen in our prediciton scenario. Or you know that the price is too low, but you have run out of cash, and therefore could not buy anything.

Anyway, thanks for the input.

tanod · Jan 18, 2008

Maybe we could get cheesemeister to separate out the top 10 predictors from 2007 in between the time predictions end and the results are released?

After the prediction deadline ends, cheese could start a new thread called "GAF estimates NPD sales" press release style.

Frank the Great · Jan 18, 2008

cool idea

Wii-za · Jan 18, 2008

I really like this idea...seems interesting...

apujanata go for it!!!

apujanata · Jan 19, 2008

tanod said:
Maybe we could get cheesemeister to separate out the top 10 predictors from 2007 in between the time predictions end and the results are released?

After the prediction deadline ends, cheese could start a new thread called "GAF estimates NPD sales" press release style.

If it is just Top 10, there is not enough diversity (not enought "crowd" to generate the "wisdom of the crowd").
Without the WV (Weight Value), there is no natural selection / performance feedback for prediction result.

I am depending on Cheesemeister or Donny2112 to separate the Top 30 predictor from the rest of the predictor, and for Donny2112 to maintain the calculation of weight value etc. It seems that they (especially Donny2112) have significant free time and can respond asap to generate result. If they are not willing to do that, I will compile them, but it may take me one or two days after the closing of prediction before I can generate the result (I have work to do, and since I have family, I may not be always available to go to NeoGAF every day).

Cheesemeister · Jan 19, 2008

It wouldn't be a problem.

Fuzzy · Jan 19, 2008

apujanata said:
It seems that they (especially Donny2112) have significant free time and can respond asap to generate result.

You saying donny2112 has no life?

apujanata · Jan 19, 2008

Fuzzy said:
You saying donny2112 has no life?

Well, I would prefer to put is as "donny2112 is at a certain age, where he has a lot of free time, and don't have to deal with life's negative issue, like dealing with wife and kids, trying to get more money to pay mortgage, work very hard so that you can get those extra money to buy games, HDTV etc, and trying to find free time to play game".

Oh, and "trying to cope with the falling stock market, especially when you have money on that market".

I am sure that donny2112 has a good life (I am no William Shatner, you know

).

apujanata · Jan 19, 2008

Edit : nevermind.

Jesse2040 · Jan 19, 2008

what he said

apujanata · Jan 19, 2008

Jesse2040 said:
*Cough*

Oops. Sorry I got carried away. My late mother (bless her in heaven) always said that I have a very imaginative mind.

donny2112 · Jan 19, 2008

tanod said:
Maybe we could get cheesemeister to separate out the top 10 predictors from 2007 in between the time predictions end and the results are released?

That's essentially the idea behind the GAF_spec grouping I use in the analyst comparisons.

tanod said:
After the prediction deadline ends, cheese could start a new thread called "GAF estimates NPD sales" press release style.

sonycowboy did that last month, and I really liked it. He skipped it this month, for some reason. :/

apujanata said:
It seems that they (especially Donny2112) have significant free time and can respond asap to generate result.

Isn't that somewhat insulting?

Fuzzy said:
You saying donny2112 has no life?

That's what I was thinking. :lol

apujanata said:
Well, I would prefer to put is as "donny2112 is at a certain age, where he has a lot of free time, and don't have to deal with life's negative issue, like dealing with wife and kids, trying to get more money to pay mortgage, work very hard so that you can get those extra money to buy games, HDTV etc, and trying to find free time to play game".

That assumption would be ... incorrect. At least most of it.

Let's just say I try to be efficient, OK? :lol

Jesse2040 said:
*Cough*

You did the simExchange topic solo this month. Last month sonycowboy did a combo simExchange/Pachter/NeoGAF thread.

For the original subject,

If the goal is to get a better prediction before NPD, then weighting as you go through the year would make sense.

If the goal is to do a GAF_spec-like grouping for comparison to the GAF Aggregate, simExchange, and Pachter, it would be giving GAF_spec an unfair advantage in the 3-month and annual totals, since we'd be automatically including "winners" from previous months in the rankings that include previous months.

If we're just talking about the first reason, I'm fine with it. Since I wouldn't be including it in the point totals for the second reason, it would just be Cheesemeister's decision, though.

GhaleonEB · Jan 19, 2008

Sounds like a great idea.

apujanata · Jan 20, 2008

donny2112 said:
Isn't that somewhat insulting?
That's what I was thinking. :lol
That assumption would be ... incorrect. At least most of it. Let's just say I try to be efficient, OK? :lol

For the original subject,

If the goal is to get a better prediction before NPD, then weighting as you go through the year would make sense.

If the goal is to do a GAF_spec-like grouping for comparison to the GAF Aggregate, simExchange, and Pachter, it would be giving GAF_spec an unfair advantage in the 3-month and annual totals, since we'd be automatically including "winners" from previous months in the rankings that include previous months.

If we're just talking about the first reason, I'm fine with it. Since I wouldn't be including it in the point totals for the second reason, it would just be Cheesemeister's decision, though.

Sorry if my post is interpreted as an attack on you. I never thought that having a lot of free time as negative. In fact, I prefer to have lots of free time myself (as long as I can feed my wife and kids, and do a comfortable living. I am a lazy man at heart). I can see that there are some stigma in the society for those that have too much free time. Anyway, I am sorry, and I'll try to be more sensitive in the future (FYI, I am not US citizen, so my point-of-view can sometimes be far removed from culturely-accepted-practice in US / any country in the world you happen to live).

My goal is for getting a better prediction before NPD, hence the weighting factor.

It can also be used as your "GAF_spec", but with the weighting factor removed (to make it more fair).

What is your opinion on Top 30 instead of Top 10 (as per Tanod's suggestion) ? I have stated my reason "more people means more crowd, and better result (hopefully)"

apujanata · Jan 23, 2008

It seems that everyone is ok with this. Should we start doing this for the Jan 07 prediction ?

Brakara · Jan 23, 2008

apujanata said:
It seems that everyone is ok with this. Should we start doing this for the Jan 07 prediction ?

Since I didn't make the cut, I think it's safe to do so.

apujanata · Jan 24, 2008

Brakara said:
Since I didn't make the cut, I think it's safe to do so.

Good.
Donny2112 or Cheesemeister, could you please post a reference to this topic when you create Jan 2007 NPD Prediction OP ? Thanks in advance.

Cheesemeister · Jan 24, 2008

It's not a huge deal. I can just generate another bolded name for the ranking comprised of the previous month's top 30 predictors.

apujanata · Jan 24, 2008

Cheesemeister said:
It's not a huge deal. I can just generate another bolded name for the ranking comprised of the previous month's top 30 predictors.

You got it wrong. The Top 30 is static (does not change monthly) for the whole 2008, and it is based on 2007 Top 30. The names are :

Code:

GhaleonEB           ;JoshuaJSlone        ;argon                      
Clever Pun          ; BuzzJive              ;donny2112           
duderon             ;rage1973           ;Heidir              
Orgen               ;Kurosaki Ichigo     sonycowboy          
Jokeropia           ;DayShallCome        ;TheKurgan           
starship            ;Frankfurter        ;AniHawk             
beermonkey@tehbias  ;apujanata           ;PhoenixDark         
djtiesto            ;Bildi               ;Earthstrike         
Hammer24            ;felipeko            ;perfectchaos007     
rostocker           ;Odysseus            ;devilscallmedad

what change monthly is the weight value (WV), and it will be calculated by Donny2112 manually, using the # you generated for him.

Once we close Year 2008, and got 2008 Top 30, that Top 30 will be used for 2009 prediction.

Brakara · Jan 24, 2008

How many people participated in all months last year? Just curious if some of those 30 are there strictly due to just participating... Maybe the top 30 should be "best average" (minimum 6-9 months participation)? Of course, I guess that's not good either (if you use people who don't participate every month).

apujanata · Jan 24, 2008

Brakara said:
How many people participated in all months last year? Just curious if some of those 30 are there strictly due to just participating... Maybe the top 30 should be "best average" (minimum 6-9 months participation)? Of course, I guess that's not good either (if you use people who don't participate every month).

You can go to donny2112 internet page to see the result from Jan 07 up to Dec 07.
You can determine yourself who is there from the start (Jan 07), and who is late comer (I am one of those).
The best average requirement is captured in the weight value.
If you should predict each and every month starting from Jan 08, you have possibility of becoming Top 30 2008,, and be part of Neogaf League.

donny2112 · Jan 25, 2008

apujanata said:
You got it wrong. The Top 30 is static (does not change monthly) for the whole 2008, and it is based on 2007 Top 30.

So it's a weighted GAF_Spec-type of listing?

I thought you were saying that the best performers from the past month could be their own group for the following month. Statistically, I'm not sure this method gains much, as, IIRC, monthly results varied widely in the 2007 GAF_spec group. Personally, I'd trust an evenly weighted GAF_Spec group more than an unevenly weighted GAF_Spec group due to the wide variances in individual performances month-to-month.

apujanata · Jan 28, 2008

donny2112 said:
So it's a weighted GAF_Spec-type of listing?

I thought you were saying that the best performers from the past month could be their own group for the following month. Statistically, I'm not sure this method gains much, as, IIRC, monthly results varied widely in the 2007 GAF_spec group. Personally, I'd trust an evenly weighted GAF_Spec group more than an unevenly weighted GAF_Spec group due to the wide variances in individual performances month-to-month.

Yeah, the reason why I don't suggest it to be based on the monthly list is because there is so much variance of the Top 30 from one month to the next (like you mentioned).

I personally believe that weighted one will give better result, but I am not sure about it. Maybe you should just calculate the "evenly weighted GAF_Spec", while I calculate the "unevenly weighter GAF_Spec", using the criteria I mentioned above (Top 30% instead of Top 30.

apujanata · Feb 21, 2008

Since donny2112 wanted to do Top 20 only (10% out of average 200 participant every month), here is the current standing of their Weight Value (beginning value is 6. +1 for those in the Top 50 for Jan 08, -1 for those NOT on Top 50 Jan 08, and 0 for those who did not submit prediction)

Code:

AniHawk		7 
apujanata		7 
argon		5 
beermonkey@tehbias	5 
BuzzJive		7 
Clever Pun	7 
DayShallCome	7 
donny2112		5 
duderon		6 
Frankfurter	5 
GhaleonEB		5 
Heidir		5 
Jokeropia		5 
JoshuaJSlone	7 
Kurosaki Ichigo	5 
Orgen		5 
rage1973		5 
sonycowboy	7 
starship		7 
TheKurgan		5

Duderon is the only Top 20 that did not participate in Jan 08, hence his WV is still 6.

This is the WV that can be used for Feb 08 prediction.

Support NeoGAF

NeoGAF league for NPD Prediction

Member

when is my burrito

Member

Member

when is my burrito

Banned

Member

Member

Member

I would bang a hot farmer!

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Similar threads