User talk:Rupert

Search Odds

If you have more data for other buildings except Hospital, even if not enough to be statistically significant, it would be great if you added it to SO pages so that i could unite it with data we already have there and get some significance. Thanks in advance --~~~~ [talk] 07:25, 13 February 2008 (UTC)

Thats some awesome data you have there, hope to see more. --Memoman 07:43, 13 February 2008 (UTC)

I will add more data soon. I have info for every search result it's just that I only have so many alts and there are so many building types, and then there are the combinations (lights/no lights, bargain/no bargain). Ugh. I agree that adding my scant results to the other results would at least help. --Rupert 17:21, 13 February 2008 (UTC)

I like the table for your individual searches, I think it will be a more accurate assessment. Of course you could do both, your individual results and yours added to the mass.--Memoman 07:14, 14 February 2008 (UTC)

Gentlemen, I reported my current results on my user page. It's a little hard to read, but at least it was easy to create from SQL Query Browser (File->Export HTML). --Rupert 23:50, 14 February 2008 (UTC)

Dammit! DAMN IT! I just found a bug in my regular expressions for detecting building lights! If a generator was running low on fuel then my scraper considered it lights off. I just fixed it but I had to throw away all 12,000 of my search results! That's okay. I'll bounce back. --Rupert 03:03, 16 February 2008 (UTC)

x_x --~~~~ [talk] 15:36, 16 February 2008 (UTC)

Okay, I added some fail safes. The building name is extracted in two different ways and must be exactly equal (never had a problem with this code, it's just in case). The lights/no lights are extracted two different ways and must be exactly equal (the second method is a lot simpler; I should have thought of it the first time). And the item detection regexps are not allowed to miss (never had a problem there either, unless it too slipped past me unnoticed). If any of those fail in the slightest way no corrupting result is submitted to my DB and bells and whistles go off. I started the DB again, this time focusing on Fire Stations. --Rupert 04:47, 17 February 2008 (UTC)

So, how many alts do you have? 1.5K of searches in 2 days o_O --~~~~ [talk] 19:04, 18 February 2008 (UTC)

I have lots of alts scattered across Malton now. It's annoying when they end up dying and I have to wait for revives. But most are alive right now, in sort-of safe neighborhoods. I stopped playing the game a while ago and instead just search, search, search. I keep them far away from each other so they don't interfere with each other. I use proxies to get around the ip limit (that's super annoying by the way, both ip limits and proxies) --Rupert 19:21, 18 February 2008 (UTC)

Which one's easier to read? - i like html-tabled more because it's easy to copypaste it --~~~~ [talk] 22:08, 19 February 2008 (UTC)

Added nice pretty tables with colors... oooo... ahhhh... --Rupert 02:33, 21 February 2008 (UTC)

Who'd be crazy enough to search a ruined building anyway?

well, i made about 350+ searchs of ruins so far (overall). mind that i don't use proxies to overcome iplimit, my characters have other goals except searching, etc. --~~~~ [talk] 21:30, 21 February 2008 (UTC)

Knives

also in fort infirmaries it's "scalpel" in mansions - "kitchen knife". there miht be even more variations --~~~~ [talk] 18:42, 24 February 2008 (UTC)

variance

You have a doubtful conclusions made based on percentages that you put with 2 decimal digits, while their variance is more than 1% with such number of searches. Most of clonclusions in "more verbage" sections is very doubtful without variance being estimated and brought as arguement. For example, about 300 searches that reveal something in 75% cases (225) assuming binomial distribution and taken average as estimated mean has variance 2.5% --~~~~ [talk] 17:28, 7 March 2008 (UTC)

Interesting, and you may be right. Even I have doubts about the data and my investigation is still on going. My next step is to take the characters from XYZ and have them search ABC and vice versa, then see if the baises remain the same or change. I wrote a little php program to simulate 300 searches using rand with 25% chance of finding 'something' vs 'nothing' and I ran the simulation 12 times and got similar variance to what I posted:

nothing: 223 0.74333333333333
nothing: 222 0.74
nothing: 228 0.76
nothing: 232 0.77333333333333
nothing: 228 0.76
nothing: 231 0.77
nothing: 219 0.73
nothing: 237 0.79
nothing: 221 0.73666666666667
nothing: 227 0.75666666666667
nothing: 237 0.79
nothing: 226 0.75333333333333

So perhaps it all is just normal variance. The above certainly takes away most of the thunder from my data but I still get the feeling that there is some sort of bias going on. In any case, bias or no bias, the only data that has any value to other players is the 'average' for all characters, and that remains unchanged at 75%. --Rupert 18:24, 7 March 2008 (UTC)

Love your numbers

Nice work. I've been tooling around with stats.csv, mostly trying to get a feel for trends in the zombie/human ratio, and I was wondering if you had any ideas concerning a problem I've got. I want to figure out the rate at which people are getting killed and getting revived, but the hour by hour data on reviving bodies, dead bodies, survivors, zombies, and active players doesn't seem to be enough. Any thoughts on how to handle this? -#XX - Jack Sparta ^TMG 16:16, 15 November 2008 (UTC)

I don't think you have enough information. (new human players) + (revivifying bodies standing up) - (human deaths) = delta(standing humans), right? You know #4. You might be able to figure out #1, but #2 and #3 are beyond your grasp. Zombies are forever getting jabbed in the neck, increasing the revivifying body count. Those that stand up increase the human class counts (the only way that class count can be increased barring new humans), but those that add to the pile come from the zombie class (but they will remain zombie class until they stand up) or they could come from the human->zombie->revivifying class in less than an hour (but that will still increase the zombie class count). ... I'd really have to think about it. You could come close but I think there is always going to be a lack of some important piece of info. Come up with as many equations as you can and count the number of unknowns, just like highschool algebra. Count the pieces of info from the stats page that you have good confidence in.

Yeah, that's the problem that keeps blocking me. A 100 human drop could be from 200 (kills + idles) and 100 (revives + new) or 400 and 300. Best I've got so far is the obvious change in (dead bodies + zombies - survivors - reviving bodies) as a rough estimator of the gap between kills and revives. It seems to behave like you'd want it to behave, even in cases where someone is killed and revived (or vice versa) within one hour. Idles and new players can be dealt with too. Anyway, I'll keep thinking about this. If you think up any way to estimate needles used or kills achieved, please drop me a line. -#XX - Jack Sparta ^TMG 19:22, 15 November 2008 (UTC)