Hey SPB, come talk about your Stats!

[Guest]

Well if that title doesn't get his attention I don't know what will.

For everyone else, if you haven't seen it yet, SPB Corp put some very interesting statistics in this week's newspaper. Very interesting stuff.

However, I have this nagging habit - I never trust any statistics until I can verify the methodology. (This makes it very hard to read a newspaper without getting annoyed, btw.) So I'm very curious here how these statistics were generated.

I'm guessing you took a snapshot of the front page of sales once per day (around the same time) and generated the statistics off that, am I correct?

The problem I have, and have always had with any kind of approach like this, is that this specifically shows unsold offers. Meanwhile plenty of trades are coming and going almost instantly, from people refreshing their pages as fast as possible to buy any undercutting offers.

In fact I strongly suspect that turnover number dwarfs the listed number, and for certain the cost is lower. And so the behaviors we judge from looking at listings are just a shadow of what's really happening. You can infer where the trading happens by looking at the lowest price, but volume data (at least as important) is not at all accurate.

For example, one conclusion generated was "Meanwhile the volume has increase by 209% in the market." I would expect listed numbers swing wildly since it's common to see lots anywhere from a few hundred thousand to tens of millions. There may be some correlation between listed quantities and transacted quantities, but not a very consistent one.

For example, if prices swung higher, natural tendency for many buyers would be to delay purchasing on the chance they come back down. (Lower volume, higher listings, deriving from higher prices.) On the other hand, if supply overwhelmed and pushed prices down, there would be higher volume as people started buying up, since these are the best prices seen in a while. Yet the listings are filling up from the increased supply. (Higher volume, higher listings, deriving from lower prices.)

So there are explanations for how the values might correlate, but not clear enough ones to be predictive. Therefore I'd be worried about getting the wrong conclusions.

That said, a very nice effort. I've been very curious about market statistics from the beginning.

[Guest]

I thought that Editors wrote me that they will post it next week ... There are orthography mistakes in that mail . They asked for innovation, I try to give them that.



Those stats are in test. I've start it on July 1st on the paper and follow it since the 4th on a excel sheet.

The concept is simple, I take the first 30 entry of each raw material everyday between 12 to 14 and their volume.

It is esay to interpret the data for Power, Stone and Water due to a quality equivalent to 0. However, Seed, Steel and Wood pass by a retreament based on Q1 to compare each data on time.

I am not a IT engineer thus I can't make it like a stock exchange. Those data are exact at the present time "t"; when I take the shot. At the present time that represents the reality of the market.

Sending data without a small analysis it is like copy a resume without understanding the sence of it.

Information for all data:

July 4th is based 100
Average : (Amount of price * Amount of quantity) for each / total quantity
Lowest Price : Lowest price at the shotcreen time
Highest Price : Highest price at the shotcreen time
SD : Standard deviation, measure the sentivity of the material the same day
Quality : Average quality at the shotcreen time
Volume : the total amount of goods at the shotcreen time

Equivalent Q1 : Prorata of the current Quality adjusted at Q=1. It is for comparing those data at equivalent quality.

"Meanwhile the volume has increase by 209% in the market."

I haven't collect volume information on time. I 've started it on July 6th. Between July 6th noon and July 7th noon the volume it pass by X to Z so a difference of 209%.

I am improving it, plus the presentation of the data for being understable and nicer. Usally, I compare the data on Day1 to the last Day and or to Based 100 to the last day.


With the actual technology, I can't control the transaction volume and spend all my day behind my screen waiting for update data. None of us can except the Administrator. I am doing my best with that I have .

My purpose is to present under a simple financial graph (day / Highest / Lowest / Volume) the background of the market evolution , with the information that I know, collect and analyse. For the forecast nor prediction, I use a trend and I lift SD on 7 days. (exception for this week)

Currently, I can't due to the lower level of information, making that you expect for it is impossible at this step. I consider that it isn't necessary to spend 5 hours / day for making a simple chart.

I will give and send you all information that we need for evaluating my method (very simple one for being understable) but do not expect more than that. It is a game and not the real world. I don't have time for everything. However, if you suggest your help for making a spychologic speculator feature, I will support it. I am looking for something simply, considering my time and my low added value into the game, this is not much.

Any correction, criticizes are warmly welcome.

[Guest]

Ah, that's about the basis I expected, but it's good to be confirmed. Yeah I also noticed the fact that you only had 2 days of volume data, but figured it's just because you only started.

With the actual technology, I can't control the transaction volume and spend all my day behind my screen waiting for update data. None of us can except the Administrator. I am doing my best with that I have.

Oh you definitely are, I'm not criticizing you. And transaction data is something I've been hoping for since the beginning, since we can't recreate it. It's only that justified deficiencies are still deficiencies, so I wanted to be mindful.

For a little while I had an excel sheet that would take a view of the latest two offers every hour and store them. I had it running on stone, and noticed both an intra-day and longer pattern. But A) it required my input to remove duplicate entries if an offer sat for an hour (though I could have fixed that with a better algorithm). And B) every 4 hours I would have to update the settings as the login session expired. Even that was limited, but I definitely have similar motivations to you. I look forward to next week's and hope a longer range of data smooths out some rougher edges.

[Guest]

Thanks for the effort but those statistics say nothing. I just hope people dont use them for any business decisions.

[Guest]

Nope. The statistic for power pricing and amount just really2 precise. i even saw millions of 0.20c power just sold less than 2 minutes.

great job SPB.

[Guest]

Maybe the seller bought it back... its impossible to make any decent statistics without help from the programmers.