How many Bitcoins are bought / sold per day - best method to know that?

[d5000]

To clarify: this is not about the pure transaction volume. It is completely obvious that for every Bitcoin that is bought there is a Bitcoin that was sold.

But on the other hand there is a difference between "buying" and "selling":

- Buy / Purchase: add a buy order to an exchange with it either being executed immediately (market order) or the intention of it to be executed in the short-term. (this is important, because there are fake orders, see below)
- Sale: add a sell order with the same intention.

Sometimes what I want to know how many orders of each kind (Buy/Purchase or Sale) are added each day, with the intention to know the difference between these two numbers (measured in BTC).

The problem is that we'd need to know the real time orderbook for a precise estimation. And data about that seems to be not publicly available normally. In addition, there's the difficulty of fake orders that later are removed or placed very far away for the spot price only to try to incite either optimism or fear ("buy walls" and "sell walls"). If they were removed the same day, it would be easy to spot them if you had access to the real time order book, but some of them are there for longer.

So for an approximation I came up with the following idea in this thread:

- Take the hourly candles of the market data of a major exchange.
- For each candle, take the volume and the direction (green or red).
- If the candle is green, add the volume to the "Purchases", while if the candle is red, add it to the "Sales".

Of course this method is inexact but it becomes more exact if you use more precise candles, i.e. 5-minute or even 1-minute candles. And if you had a candle for each executed order that would be the perfect way. Because if, due to the execution of the order, the price becomes higher, then it is clear that the order was placed with the intention to buy, and vice versa.

I want to ask:

- Has someone tried the same thing and found a better method?
- Is there another flaw I'm not taking into account than the candle imprecision?

[hmbdofficial]

One thing I still don’t understand clearly how this things works why the selling of bitcoin in large quantities affects the price, because if you look at it someone is selling and another is buying, no bitcoin is lost in the process then why is that when there is a panic sell the price often get affected if for every bitcoin sold is directly linked to a new purchase.

[d5000]

One thing I still don’t understand clearly how this things works why the selling of bitcoin in large quantities affects the price, because if you look at it someone is selling and another is buying, no bitcoin is lost in the process then why is that when there is a panic sell the price often get affected if for every bitcoin sold is directly linked to a new purchase.

This is the effect of market orders  (includes also de facto market orders, i.e. limit orders with a limit which already matches an existing order).

Imagine the price is at 70000, and you create a market buy order for 1 BTC. There are only 0.5 BTC at 70000, and the next sell order is 0.5 BTC for 71000.

So with your market order you'll buy both the 70000 order and the 71000 order.

The result of you buy is that the spot price will climb from 70000 to 71000.

Market sell orders have the opposite effect: the price will go down.

Exactly that is what I want to achieve with this thread: to know the number of "buy" and "sell" market orders and defacto market orders.

[Mia Chloe]

~snip

Even if you're kinda close to it, I think the biggest flaw is that green candles don't necessarily mean buying volume exceeded selling volume and red candles don't necessarily mean the opposite. Every trade has both a buyer and a seller butt price moves because one side is more aggressive.

Personally I think a better method is to use taker buy vs taker sell volume if the exchange provides it since it gonna show which side initiated the trades.

Another thing is candles only show executed trades not all orders placed. Many buy and sell orders can be  canceled before execution, so candle data cannot accurately measure total buy and sell orders added to the order book.

[d5000]

Personally I think a better method is to use taker buy vs taker sell volume if the exchange provides it since it gonna show which side initiated the trades.

That's a good idea - thank you for the correct term. I also found the metric "Cumulative Volume Delta" which seems to be exactly what I'm searching, but until now I have only found public data for the past.

There is one catch though: Imagine the following situation:

- Trader 1 puts a limit buy order on 70000
- Trader 2 sells via market order into that limit order. Price of last trade: $70000.
- Trader 3 puts a limit buy order on 71000. There are no lower sell orders.
- Trader 4 sells via market order into that limit order. Price of last trade: $71000.

We have now two "taker sales" but the spot price increased $1000.

I wonder how common this situation is on high-volume exchanges - it could however be quite common if we take into account that often orders are placed in a tiny distance (66,777 vs 66,776.99 for example).

We can still say that there is "more interest to buy" in this case, because trader 3 created a higher limit order than trader 1. And no seller placed an order between 70000 and 71000.

The "candle method" I show in the OP takes this into account, while the "Cumulative Volume Delta" and "Maker Volume" method does not. But yeah, the candle method is less precise ...

Another thing is candles only show executed trades not all orders placed. Many buy and sell orders can be  canceled before execution, so candle data cannot accurately measure total buy and sell orders added to the order book.

Yes but that's not that important, because what I want to know is about the really executed orders (i.e. really existing "buys" and "sales"), and there are many fake orders placed all the time.

In reality, to make my definitions of "buy/purchase" and "sale" more precise, perhaps we can use the following definition:

- Buy / Purchase: if the price of an executed order is higher than the previous executed order (even if it was triggered by a market sell order).
- Sale: if the price of an executed order is lower than the previous executed order (even if it was triggered by a market buy order).

[Mia Chloe]

~snip

Well this should make more sense. In your example both trades are technically seller initiated yet the price actually rises because buyers are willing to bid higher.

I think where there is kinda an issue is that once you define a "buy" as any trade executed at a higher price than the previous trade you're no longer measuring order flow but rather price movement. And it's  gonna work if your goal is to estimate bullish vs bearish pressure.

And it's actually similar to how some volume indicators work by assigning volume to up-ticks and down-ticks. It won't tell you who initiated the trade but it does capture if the  executed volume is occurring at increasingly higher or lower prices which seems kinda closer to what you're trying to measure.