UPDATE - READ IT ALL; WATCH FOR YOUR NAME

READ ON BLOG IF YOU DON'T LIKE E-MAILS

GSD BLOG: http://grandstrategydynamics.blogspot.com/


1. Satjot, I added you to our mail list - if you don't want to see comm e-mails tell me, don't want to spam people.


2. Rajesh has taught Robert how to test games. Robert will transfer knowledge to Xiao, et.al.


3. Mo, Ramakrishnan, Lin, Huang, and Wei have solved the Initialization problem (minus the part about enrolling Quarantine Zone II players, and some mathematical details involving a Cezaro Mean, and a few other hazies?) Other than this... I have written the doc, and Wei has approved and edited, RR has yet to view and edit, pending Wei's corrections to be e-mailed shortly!!!


Zhe stands ready to etch the solution. The Document is withheld pending Edits.


4. Mo, Wei, Crone, Huang, had a discussion concerning the Gametype Problem


a) Gametype problem is the following - what do you do about different gametypes where accordingly you collect different types of data... hence bringing issues when players are sorted/ranked on the same continuum with different sources of the data depending on gametype (CTF, deathmatch, w/e).


We have a proposed multidimensional symmetry solution with transforms allowing for interconversion to fill in the missing spots. NOT being a HUGE math initiate, but merely a fan, I will use one of my (apparently now infamous?) analogies. In Jurassic Park, to recreate Dinosaurs, Amphibian DNA is used to replace missing sections of Dinosaur genes. The concept is to use the base gametype (i.e. deathmatch usually) as the Complete Amphibian DNA, and then use this data to "Fill In" the missing sections of data on the other gametypes allowing us to maintain separate DISTs on separate gametypes, but also allowing for them to intercommunicate and inform each other.


In a nutshell, that's the current solution, and I like it thus far, mostly because it's the only good one I've heard. The nature of the multi-K transforms concerns me though... While presuming symmetry, we can C/P many of them... but it should be better vetted out.


5. Mo, Wei, Crone had a discussion about the Party Problem (Which Ramakrishnan and Mo claim to have solved weeks prior)

a) RECALL our system was built for ONE player to enter Matchmaking, but many systems allow for TEAMS ("Parties") to play, thus allowing Friends to play together. This creates some Fundat issues, the main one being that this has broken GSD as the team-builder, as the player has built his (or God forbid... Her) team. (Sorry not a good personal experience with female gamers...)


Ramakrishnan and Mo's solution (involves building separate profiles for different combos of players and extracting that data from the single-person majority data, then calling this data whenever teams of specific players play together) has come under fire, though no alternative was proposed that withstood first order challenges.


6. Crone has developed a working solution the ID Player Problem (prevent use of accounts by others)... While this is NOT 5-week plan vital, it is necessary to have a defense in place against such an obvious system breaker. His Proposal is attached. I expect Comments on it from Pelletier, Xu, and Santos, former members of Team [DES].


7. Comrade Wei reports all Testing theory is complete and that Implementation may begin. His redefined success metric is the differential between all vars and the reduction therein. His new testing implementation requires some mods given discovery of 100 bot skill levels. RR should present ALL known details concerning BOTs and BOT SKILL LEVELS to this team.


8. Fresh MATLAB Conscripts available for the Front. Assigned to Chen and Ramakrishnan (1 to Lin), awaiting deployment orders.

Jeffery Chiang & Sara McArdle (Work together) (sam2148@columbia.edu, jyc2115@columbia.edu)

Felix Yu (fy2117@columbia.edu)

Subhrendu Sarkar (ss3295@columbia.edu)


All have been tested for MATLAB proficiency.



9. Remaining Unsolved PROBLEMS

a) Shock-Stability Problem (Nothing on this yet!)

b) New Player Problem (Proposed Solutions approaching consensus) (Volunteer Fast-track solution if n << N, Otherwise Use Quarantine Systems; faster method?)

c) Gametype Problem (Multidimensional solution proposed; lightly contested)

d) Party Problem (Mo, Ramakrishnan Claim Solution; Contested)

e) Drunk Player Problem and all associated issues (ID Problem) (Proposed Solutions avail.)

f) Testing Problems (Seem to be mostly solved... Need to be written up (Pelletier, Xu, Wei, Santos, Crone are assigned this duty, under direction of Wei)

g) Anything else?

h) Quality Control Module and Learning Algorithm remain nebulous and, quoting one officer, "vaporware." But they are often used to justify holes in current logic, as they should be.


10. As always, we march on. 5 Weeks. Every Casualty Allowed. Not 1 Step Back.