Quote from: genuise on April 16, 2013, 07:45:32 PM

It is not a problem to multiply usd bid price and volume of btc in each bid order.

But the problem is, as nimnul explained, that simple multiplication will not get meaningful number.

The problem is that not all bid orders have meaningful, as nimnul stated a "real", price. e.g. if someone places a bid order to buy 1000 000 000 btc for 1 USD, in other words buyer expects to buy 1bn btc at 0,000000001 usd/btc. does this price make sence for youe analyzes? Such order will be never executed.
Thus such orders represents spam, not real , fake data in the orderbook.

There are a lot of such orders, and their combined volume just obscurs the true picture.

If we exclude those fake orders we will see the true picture.

It this clear?

So the questions is how to tell which orders have to be excluded and which not?

The problem is here that there is no common agreement on which order have to be counted as meaningful and which is not.

So how to find this common filtering creteria is the main question.

Do you have any idea on how can we solve it?

One way to filter it would be use an discount factor based on distance from the top order. The discount factor can be exponential or linear.
An exponential discount factor would be something like e^(-ax) where a is some scaling factor and x is the distance. A linear discount factor could be something very simple like a(x/x_top)

For example, say current top bid is $60 and there's a bid for 1000000000 BTC at $1.

The exponential discounted volume with a = 1 is 1000000000 * e^(-(60-1)) = 2.3802664e-17, i.e. negligible. If we want to give it a bit more weight we can decrease a to 0.1 which yields 1000000000 * e^(-0.1(60-1)) = 2739444.81877.

The linear discounted volume with a = 1 is 1000000000 * (1/60) = 16666666.6667. Give it more weight by increasing a.

The same thing could be used on asks. Personally I'd like to see orderbook volume charts in USD and BTC (option to show volume in USD or BTC). I'm also curious if the spam orders on both sides simply cancel out, what does the chart look like if you just multiply without discounting? If I could get some raw data I can play around with the parameters in MATLAB and see what values are reasonable.

For the depth chart it's not necessary to discount, simply convert BTC to USD and vice versa at the price of each order to make volume on each side use the same currency. Again it would be nice to have an option to toggle between BTC and USD.

Another suggestion, add another series in the VWAP chart for both buys and sells, and have a control to show/hide each chart.