Bitmex Trollbox, a short analysis

In this article we will analyze some data collected from Bitmex, to be more specific, the positions shared by the users.

Bitmex is the largest Bitcoin futures exchange in action. Besides the enormous amount of daily trades, the Trollbox, the internal chat service, also presents a great volume of messages.

If it were only a chat service, the Trollbox wouldn’t have a lot of interesting data. However, because of a tool that makes it possible for the user to share his position in an honest way, not being able to falsify the position size or the price, we have access to data that can clarify a little the behavior of those who use Bitmex.

We can ask the following questions:

1. How many users share their positions daily?
2. Which side constitutes the majority, longs or shorts?
3. If in their majority they are winning or losing positions, which side wins?

To answer these three questions we are going to analyze the data collected from August 1st to December 31st, 2019. During this period 161147 positions were shared, of which 81974 were longs and 79173 were shorts (50,86% longs and 49,13% shorts). Below there is a graph of shared positions throughout the period:

The dashed line represents the mean.

By looking at the graph above we can have a basic idea of how many users share their positions, but it answers the first question only for the analyzed time. As our intention is a brief investigation about these data, this is not a big problem.

It seems that the second question has already been answered, the longs are the majority. However, precisely because the proportion of longs and shorts are really close, and it shouldn’t be necessarily 50–50, it is worth to look the longs and shorts quantity compared to the total number of shared positions:

Here the first graph shows the percentage of longs shared daily, while the second gives the percentage in relation to the total volume (size of the positions). The other graphs do the same thing for the shorts.

See that there are some divergences, when the number of the long positions is smaller compared to the number of shorts it doesn’t necessarily implicates that the volume of longs is smaller than the shorts. In the case of the volume, supposing all the users share their positions, we should have 1:1 in longs and shorts. We can also notice greater variability in the volumes than in the number of positions.

Then remains a question, are the open positions a reason for worrying or a source of relief for those who made their bets? Are they winners or losers? Let’s see the following histogram:

On the horizontal axis we have the profits and loses percentage.

Here the data are not grouped by day. All the 161147 ( 81974 longs and 79173 shorts) positions are present. We can see a large amount of profitable positions, which for some will confirm something quite plausible: the users that share their positions do it when they are making profit. This suspicion is also confirmed by the vast majority of positions being profitable, about 70%. And by the shorts slightly winning, it is 70% of the shorts in profit against 68% of the longs.

Calculating the z-score, we notice something curious, a great part of the users is at most one standard deviation below the mean (1.28% for the shorts and 0.9%, both means are positive indicating profit). Here the shorts have an average profit of 2.22% and an average loss of 1%, in case of the longs the average profit is 2.02% and the average loss is 1.5%.

Horizontal axis indicates the standard deviation.

What else can we conclude after answering these three questions? In fact, everything indicates that only a small share of users make their positions public, otherwise we would have a smaller variation on the percentage of longs and shorts in relation to the volume, where the graph would have to be close to 50% most of the time. And as stated before, it is expected that we see more profitable positions being shared, that is to say, when shorts are being profitable, we will see more shorts coming up, with the same happening to the longs.

But is the last statement really in agreement with our data? Let’s take a look:

We have the graphs of a 10 day simple moving average. The first two indicate, in the case of the longs, the profits average (negative in case of loss)

We can see on the graph above that there are moments when the positions average profit increase, whether shorts or longs, while the number of daily positions decreases, going against our previous statement. A possible explanation is that some users open their positions too late and the price begins to move against them making their positions less profitable and even turning them into a loss.

To finish, we can see that a cursory analysis of the positions shared on Bitmex allows us to ask some basic questions, answering them and even risking a hypothesis: lucrative positions must compose the majority.

Besides, it would be interesting to know how much time an user leaves their positions open, if profitable positions last longer than the ones that are losing and what is the funding rate influence on the decision to close a position.

One more thing, are these data really fit to measure the market temperature indirectly? For instance, would that be that the shorts and longs percentages could indicate buying or selling pressure foreshadowing a price reversion? No doubt further assessment of how significant it is the volume of shared positions will be required. But that we will see in a future article.

If you want to see live Trollbox data, besides other data which deserve some thought, check out: https://dashboard.heimdall.land/trollbox