  
| Author: | yorbrick  | | Posted: | Oct 1, 2018 10:45 | | Subject: | Re: 90-percentile as new average price ? | | Viewed: | 64 times | | Topic: | Suggestions | |
|
| | | You are jumping to conclusions. I never said prices are unfair. I'm saying
the average calculation, although mathematically correct, is not a good indication
of what a piece is actually sold for most of the time.
|
Throwing away data using some other biasing gives an even worse indication of
what it has sold for.
| | Because that is what the average amount implies: the average price to which an
item is sold for most of the time.
|
The average value reports the average (mean) value - it is exactly what it says
it is. The average is not the value an item is sold for most of the time. Although
sellers might want to know a more reasonable price to list their items at, the
average is there as a guide, not a definitive value.
| | For me, I use the average price to set my pricing. But currently it becomes unreliable,
when you have to verify how many items have been sold at a price that's not
realistic (or at least: that has something strange going on -- $1.00 for
a typicall $0.02 item, seems strange to me).
|
Why is it unreliable? It is reporting the average of all sold prices. If a $0.02
item frequently sells for $1.00 then it should be included as the sales are frequent.
If it is a one off, then it will hardly affect the average price at all if the
item is commonly sold at $0.02.
| | You are not for, because it would make the calculation less transparant. But
it doesn't have to be clear how it was calculated, as long as it's accurate.
|
There is no such thing as an accurate average when data has been removed. You
don't want an accurate average price, you want an accurate average based
on a 90-percentile, so only taking the lowest 90% of the data. What happens if
there are low outliers? Why do you want to include those, but not the high end
of the distribution. If you average over the lowest 90% then base prices on that,
then average over the lowest 90% and base prices on that, then average over
the lowest 90% and base prices on that, and iterate so on, then prices will necessarily
tend towards zero, even if the price distribution is Gaussian each month.
| | In the end, to me, it still feels the average price as a purely mathematical
average, is not very useful.
|
Any statistic like this has to be mathematical. Your suggestion is also mathematical.
The algorithm you suggest is to take the lowest 90% of prices and take the mean.
That is no less mathematical than take all the prices and average.
Averages are a bit pointless for distributions that are not Gaussian / bell-shaped,
and on BL many have either long tails left or right, or are bi-modal (or worse).
There are other ways of getting averages, such as fitting a Gaussian distribution
to price regions with dense sales, ignoring others. Or fitting the middle 95%,
or 90% or 60%, etc. You can do this, as BL supplies all the data you need to
choose your own algorithms.
A much bigger problem is that the data you are using to fix your sales prices
are worldwide, yet prices vary by region. Some regions include tax, others do
not. For example, why base your price on an average mainly coming from US sales,
when the EU market might be willing to pay more?
|
|
Message is in Reply To:  Re: 90-percentile as new average price ? - Gaston.La.Brick (1844) | [...] You are jumping to conclusions. I never said prices are unfair. I'm saying the average calculation, although mathematically correct, is not a good indication of what [...]
(69 months ago, Oct 1, 2018, to Suggestions) |
8 Messages in this Thread: Msg 1 - Gaston.La.Brick (1844) 69 months ago Oct 1, 2018 to Suggestions
 Msg 2 - Teup (6616) 69 months ago Oct 1, 2018 to Suggestions
 Msg 3 - Gaston.La.Brick (1844) 69 months ago Oct 1, 2018 to Suggestions
Msg 4 - MidwestBrick (1855) 69 months ago Oct 1, 2018 to Suggestions
 Msg 5 « - yorbrick (1185) 69 months ago Oct 1, 2018 to Suggestions
Msg 6 - Bricks_NW_UK (1367) 69 months ago Oct 1, 2018 to Suggestions
Msg 7 - Etown (1740) 69 months ago Oct 1, 2018 to Suggestions
Msg 8 - jx (13) 47 months ago Aug 12, 2020 to Suggestions
Entire thread on one page
This message and all its replies on one page
|
|
United Kingdom, England
