My attempt at showing a working model for stumbleupon mathematics certainly got (is getting) a lot of attention with several questions being asked I thought I would at least try to answer some of them and pass some more generic comments. If you haven’t read the stumbleupon mathematics post then this will not make much sense but to recap the proposed “simplistic big model”
(Initial stumbler audience /# domain) + ((% stumbler audience /# domain)+ organic bonus – nonfriend) – (((% stumbler audience + organic bonus) + N http://ventureskills.wordpress.com/2007/09/19/stumbleupon-mathematics-for-stumblers/
So now on to the questions. 1. You said this was a simplistic model how many variables do you think there are?
Tim – Far more then 6 in the proposed “big model” our full big model included 20 variables but this was a restriction of the data rather then the model.
2. When you say modelled have you simulated and if so how accurate was it?
Tim – Yes I have run computer simulations with mixed success we never simulated a large stumble with the exact same number of visits. We also did not use N a random variable rather substituting it for an error margin.
3. What language did you model it and when it being released?
Tim – The model was simulated with Python and I have no intention of releasing the full formula I used or the script in any form sorry.
4. Why not?
Tim – The model is not accurate its close (in predicting historical stumbles) but its not accurate, if I’m close then the guys at stumbleupon may change the algorithm I do not want my name to be a swear word for Ebay engineers for years on end. If I’m not close well I have a pride thing at stake plus I did it to satisfy my curiosity it was never meant to be heavily scrutinised.
5. But you can predict how well a sumble will do?
Tim – Nope we can simulate how well a stumble did we require historical data to simulate, I have no idea who will stumble a post and therefore can’t possibly predict how it will react.
5. But you have a shrewd idea of Audience scores?
Tim – Again yes and no, I sadly didn’t pay attention to the audience scores when they were published so have had to go on a lot of guess work. I can with some success tell you the rough percentage difference in audience scores of 2 stumblers given enough data. Mind you by the time I have worked it out you could have just looked at their profiles and told me the answer with as much accuracy!
6. What effect is a thumb down do?
Tim – This was hard to collect data we did run some small tests, but obviously we didn’t want to pollute stumbleupon with tests so had to keep such things small scale and mark them down as spam ourselves, I’m not sure what the engineers who received the spam report thought It does however have an effect and is determined by audience score.
7. Does thumbs down count towards increasing Audience score?
Tim – Pass, I modelled both my gut says no, but I have no evidence to support that.
8. Are you really sad enough to simulate some one else’s algorithm and then not even release it?
Tim – hmm yes!
9. How can I game the system?
Tim – I wish I could say you can’t but that would be untrue but the best way to game stumbleupon is to use it or get people to use it on your behalf. Remember each failed attempt counts against you.
10. Can you game the system?
Tim – no more so then anybody who has played with stumbleupon for some time.
11. If I give you a vast quantity of money will you game stumbleupon for me?
Tim – You can give me the money but as stated above I’m not in a position to game the system, but my company does provide Social media optimisation services so get in touch (sorry blatent plug it was a question asked though!)
12. You pretty sure mailing lists don’t work but I know they do?
Tim – cool what floats your boat, I didn’t say they don’t work I said they loose their effect if repeatedly used on the same domain.
13. Did you know your really sad?
Tim – yes, but geeky so that’s ok.
14. So do you know how Page rank works?
Tim – Sure here’s the patent moving on please, to the best of my knowledge Stumbleupon or ebay have not applied for a patent on the stumble algorithm I may be wrong and would love to see it if they have.
Hope that covers some of the questions you might have had if not please leave me a comment and I will try to answer it.