Game Room: The Search for More Accurate Estimation - Part 2

i tried to refine the concepts i used in the previous post (basically using the average) a little.  True, using the average is still a little off (but i don't have enough relevant data points).  So, this time i decided that i would try things a little more difficult.  This is actually trying to "guess" the true leaderboard amount.  i went to my trusty Excel and came up with this...

The Irr method: 
- i take the number of weeks a game has been on the service (the number of weeks in 2010 + weeknum(today()).  So it updates the averages and weeks automatically. 
- Using that number of weeks, i have a scale (sort of), where i set the first week of release equal to 1.4 x Average.  (the average is in my previous post -leaderboard level / # weeks on service).  i have it decrease week over week by 1/(number of weeks - 1). 
- i plug the relevant "first week data" (leaderboards started in late August) in the corresponding week.  After that leaderboard
- After that i just summed the first week with the relevant (or not) series of "estimates" for subsequent weeks.  So, in some ways this sort of resembles a reverse NPV (but not quite).  What, you thought that Irr was a reference to internal rate of return?
- The numbers will change from week to week

Example:  Activision Decathalon
- 26 weeks on Game Room
- Average weekly leaderboard growth estimate 38.46
- First week numbers i have: 601 on September 16 (week 17, missing 8 weeks of data)
- Start with 1.4 number week 25.  Input the 601 number at week 17 (but do not multiply by 1.08), multiply the next week (week 18) by 1.00 (as the decrease per week is 1.04)
- Add the 600 with the sum of the weeks below it (or just sum the range)
- Optional: Realize that you forgot to added newer data and adjust your numbers
- Look at the result (1,071 truncated) and see if it fits the range 1,000-2,000.  If it does - good.  If it doesn't oh well. 

Now look for data in the other 107 ranked games - i know it sounds really exciting. 

Then you add some kind of spreadsheet link to your blog post ...   eventually .....

Look for the link in part one... i added an example of the method to page 2 (Irr Ex), page 3 is the two results i had (so far?) using that method.