[SP] Fun With Escort Data II
Carol Ou
semester ends up easier than it appears to be. Payments were down across
the board... so either I was riding a lucky streak with my first set of
data (sample of 85) or the escort union read my earlier post and decided
this was easier than breaking my knees. Anyway, 201 in this sample, and
thanks to Carson and Pixie of PAG for providing some of the new data.
(Surely you didn't think I managed an additional 116 escorts (29 hours of
escorting) in three days all by myself?)
Note: These were all done on the Siege Perilous shard where there is a
limit of one escort per 15 minutes.
Profession Sample Low High Average Hourly/4 escorts
Mage 26 0* 491 308.2 1232.8
Merchant 22 0* 465 290.9 1163.6
Messenger 34 0* 485 292.8 1171.2
Noble 29 110 465 302.1 1208.4
Peasant 58 0* 500 277.5 1110.0
Wander Healer 32 125 494 287.8 1151.2
Overall 201 0 500 290.7 1162.8
*Indicates escorts which were defective and did not pay when they were
delivered to their destinations. Encountered 6 in 201 or about 3%.
The best bets with this new analysis are the nobles and mages. Random
escorting should net an average of 1162.8 gold per hour or per four
escorts. There's a new high score for escorts, a payment of 500 gold by
a female peasant. Minimum actual payment (not counting the duds who
pay nothing) remains 105.
The peasants and merchants plummeted in their average pay, while the
nobles were the big gainers. Apparently, they decided to loosen their
purse strings a bit. Overall, though, escort payment was down. The
discrepency between the highest professional average and the lowest
professional average was down as well. Previously the spread had been
279.6 (nobles) to 332.9 (merchants) or a 53.3 gold difference. Now the
spread is 277.5 (peasants) to 308.2 (mages) or a 30.7 gold difference.
A continued pattern of shrinking spreads would indicate that the
profession of the escort has no relation with the escort payment. I
guess we'll see with the next bunch of data.
I crunched some other numbers with regards to gender/profession,
destination city and origin city, but the samples of some of those are
really tiny, so I'm going to wait until there's more data before posting
an analysis involving those factors.
Ria di Sonom (Rescuer, Squire, 911) - Siege Perilous
Vera Embry (Wandering Miner, PAG) - Siege Perilous
Carol Ou
: Well, here's another installment, and the last one for a while, unless my
Forgot to mention that the streak of female escorts with facial hair
continues. All 10 female mages in this data set (up 4 from the last data
set) had at least a scrubby little mustache or a faint goatee.
Brandy
Carol Ou <o...@mail2.sas.upenn.edu> wrote in message
news:85nnng$47l$1...@netnews.upenn.edu...
<snip>
> The best bets with this new analysis are the nobles and mages. Random
> escorting should net an average of 1162.8 gold per hour or per four
> escorts. There's a new high score for escorts, a payment of 500 gold by
> a female peasant. Minimum actual payment (not counting the duds who
> pay nothing) remains 105.
>
> Ria di Sonom (Rescuer, Squire, 911) - Siege Perilous
> Vera Embry (Wandering Miner, PAG) - Siege Perilous
I have had high payments of 501 and low payments of 101. Also on SP, of
course.
Brandy (WE, LS)
Carol Ou
: I gotta say, this is starting to look to me like its pretty much
: random. Done any checks?
: Otara
I'd thought escort payments were random to begin with, and then was
informed that peasants were supposed to pay better than nobles do. So
there's that underlying lore that different professions pay differently..
which is what I'm testing. The trend does seem towards randomness
though, as I noted in that post, with there being less and less of a
spread between the average payments of each profession.
And by checks, what do you mean, specifically?
Brian Keener
> And by checks, what do you mean, specifically?
>
Did any of them pay with checks, I believe was the question. You know, if
they did, you still don't know if they'll bounce or not, so you might have a
much higher percentage of non-payers...
OK, OK, it's late, I'm still at work on my SQL database, and I've lost my
sense of humor... Is that a crime?
Brian K
rattran
achieving a substantive base (say, 50-100 of each
escort/destination/origin) you'll have a decent set of data. What you
might want to post in addition to the average payment, is the median
and standard deviation for each profession.
*shudders* That's what 4 years worth of Actuarial training gets you.
I'm almost convinced it's random, but it seems to me that Nobles and
peseants tend to pay either very high, or very low. Mages and
Merchants tend to pay me close to the average listed, about 300 each
time.
-Detlef Sierck on LS and SP [PAG]
On Sat, 15 Jan 2000 10:37:57 +1100, Otara <sp...@spammity.com.au>
wrote:
>I was just trying to figure out what kinds of statistical tests to do.
>I'm hazier than I thought I would be on what the appropriate one is.
>
>Otara
>
>On 14 Jan 2000 23:11:34 GMT, o...@mail1.sas.upenn.edu (Carol Ou) wrote:
>>I'd thought escort payments were random to begin with, and then was
>>informed that peasants were supposed to pay better than nobles do. So
>>there's that underlying lore that different professions pay differently..
>>which is what I'm testing. The trend does seem towards randomness
>>though, as I noted in that post, with there being less and less of a
>>spread between the average payments of each profession.
>>
Otara
random. Done any checks?
Otara
Otara
I'm hazier than I thought I would be on what the appropriate one is.
Otara
On 14 Jan 2000 23:11:34 GMT, o...@mail1.sas.upenn.edu (Carol Ou) wrote:
Otara
>What she's doing now is compiling a statistical database, after
>achieving a substantive base (say, 50-100 of each
>escort/destination/origin) you'll have a decent set of data. What you
>might want to post in addition to the average payment, is the median
>and standard deviation for each profession.
>*shudders* That's what 4 years worth of Actuarial training gets you.
I was just couldsnt remember what the appropriate test would be to see
if they differ significantly from each other. I guess you could look
at distribution etc as well. BTW I think thats thousands of cases
you're talking about there - I think that would be a bit of overkill
:).
And what _were_ you thinking? I know the money is good, but thats a
hell of a price to pay.
>I'm almost convinced it's random, but it seems to me that Nobles and
>peseants tend to pay either very high, or very low. Mages and
>Merchants tend to pay me close to the average listed, about 300 each
>time.
I'm certainly seeing nothing to suggest it wouldnt be random - and I
cant for the life of me think why they would have coded it in if it
isnt.
Otara.
Richard Cortese
news:MPB=OK0SGoXTHG6==sGeC4E=2i...@4ax.com...
> On 14 Jan 2000 20:36:03 -0600, rattran <rat...@renn.faire.net> wrote:
> >What she's doing now is compiling a statistical database, after
> >achieving a substantive base (say, 50-100 of each
> >escort/destination/origin) you'll have a decent set of data. What you
> >might want to post in addition to the average payment, is the median
> >and standard deviation for each profession.
>
> >*shudders* That's what 4 years worth of Actuarial training gets you.
>
> I was just couldsnt remember what the appropriate test would be to see
> if they differ significantly from each other. I guess you could look
> at distribution etc as well. BTW I think thats thousands of cases
> you're talking about there - I think that would be a bit of overkill
> :).
you a degree of certaincy for saying data is related or conversely
unrelated.
If you have ever heard something along the lines of "We can say with a 95%
degree of confidence yada yada yada" that is the t test.
rattran
>
>I was just couldsnt remember what the appropriate test would be to see
>if they differ significantly from each other. I guess you could look
>at distribution etc as well. BTW I think thats thousands of cases
>you're talking about there - I think that would be a bit of overkill
>:).
>
>hell of a price to pay.
>
>cant for the life of me think why they would have coded it in if it
>isnt.
think at least 5000 examples would be neccessary to be accurate,
though if it IS random, you'd be able to reasonably guess much
earlier.
As for what I was thinking, I love maths, and it got me a full
scholarship. Now I sell pewter goblets at 16th century re-enactments
for a living, so I can't say it was particularly useful. It is amusing
what bizarre bits of statistcs and other formulae have stayed in my
head though.
I'm waiting for more data on embark/destination, I think they have
more to do with money than anything else. I consistently get 400+ for
a Skara to Moonglow escort of any variety.
Otara
<rico...@netmagic.net> wrote:
>You are probably thinking of the ~"Student t-test" or t test anyway. Gives
>you a degree of certaincy for saying data is related or conversely
>unrelated.
>
>If you have ever heard something along the lines of "We can say with a 95%
>degree of confidence yada yada yada" that is the t test.
Nup - multiple comparison needed to check if means differ sig from
each other. And theres lots of tests that use that language.
We do stats out the wazoo in psychology, just cant remember for the
life of me what I'd use for these.
Otara
Otara
wrote:
>>If you have ever heard something along the lines of "We can say with a 95%
>>degree of confidence yada yada yada" that is the t test.
>
>Nup - multiple comparison needed to check if means differ sig from
>each other. And theres lots of tests that use that language.
I do beg your pardon Richard, it looks like a t test will do fine - I
presume each t test of a given variable checks against the means of
the other variables combined or somesuch.
Otara
Otara
>Standard deviation would give you by how much they tend to differ. I
>think at least 5000 examples would be neccessary to be accurate,
>though if it IS random, you'd be able to reasonably guess much
>earlier.
I think you could do something useful with a fair few less than that
tho - you dont really need to test every possible combination do you?
I'd just be testing for income differences between profession, and
income differences between origin points and between destination
points. I think you'd start getting a clue a bit before 5000.
I find the idea that they've got some kind of grid operating a bit
unlikely myself. If they do, no wonder theres outstanding bugs in
this game. Talk about overkill.
>As for what I was thinking, I love maths, and it got me a full
>scholarship. Now I sell pewter goblets at 16th century re-enactments
>for a living, so I can't say it was particularly useful. It is amusing
>what bizarre bits of statistcs and other formulae have stayed in my
>head though.
Fair enough. I was going to be a food technologist at one point,
deciding what chemical was needed to make asparagus the right colour
coming out of a can. I dont know what I was thinking.
>I'm waiting for more data on embark/destination, I think they have
>more to do with money than anything else. I consistently get 400+ for
>a Skara to Moonglow escort of any variety.
Island to island. Hmmm. Well shouldnt take too much to confirm that.
Should turn up as a higher mean for delivering to moonglow, unless
they have lower means from other origins as a compensation - seems
like awfully hard work to ensure that though and makes no sense
whatsoever. I really want to think they have a better sense of
priorities than that.
Otara
Richard Cortese
news:wf+AOH1WRnl+rs...@4ax.com...
> Standard deviation would give you by how much they tend to differ. I
> think at least 5000 examples would be neccessary to be accurate,
> though if it IS random, you'd be able to reasonably guess much
> earlier.
the average income is 300 gold with a standard deviation of 30, that means
~70%<71.4%?> of all samples taken will pay between 270 and 330 gold.
Can't remember how to spell it, Greek Rho or Ro... but it is one of the
factors you get when calculating SD. I seem to recall 3X(Rho) or 3XSD lets
you say ~There is only a 1 in a million chance that a new data point will
fall outside of this range.
It was useful for me since I worked on controlled drug deliver devices. We
could go to the FDA with a NDA that included a degree of certaincy that our
devices would perform within the product specifications i.e. 10 mg/device
+-20%.
SD would really only tell something about tight data sets. That is, if the
average and SD for two different groups of data were 299 +-3 and 300 +-4 you
could probably say both groups were related. In this case, range of 105 to
500 gold would stomp all over your SD. You would end up with something like
avg 300 +-140 so even 5,000 samples wouldn't help you.
>
> As for what I was thinking, I love maths, and it got me a full
> scholarship. Now I sell pewter goblets at 16th century re-enactments
> for a living, so I can't say it was particularly useful. It is amusing
> what bizarre bits of statistcs and other formulae have stayed in my
> head though.
>
rattran
the range of values seems pretty small. ie. Merchants from Skara to
Jhelom all seem to pay between 120 and 170 (145 +- 25). I think
origin/dest is the thing to check. We'll see when we have more data.
I think it's been a full 10 years since I used most of my statistics
knowledge, some bits are still creaky/dusty.