Trends in crypto market data: market analytics
For a long time we’ve observed crypto market data vendors taking existing business models from the ‘traditional’ market data vendors and making a similar tool / API for crypto traders. From raw data normalization and aggregation to tensorcharts to social media semantics data, all are methods of gathering and consuming market data already familiar in the traditional securities, commodities and forex markets.
Earlier this spring we conducted a survey with 170+ crypto traders to understand what they think of analytics built on top of raw market data feeds, an area that we have long speculated to grab the attention of the mainstream crypto traders during 2020. This blog post goes over the main questions and results of that survey. So let’s start!
Structure of the survey
We had three specific analytics in mind whose applicability to various crypto traders we wanted to understand. The measurement of each feature was done to give us information on two respects:
a) How interesting is the data?
b) What is the financial value of the data?
We also wanted to understand what kind of traders are answering the question. For this we asked three questions:
a) What trading strategies do they use?
b) How many weekly trades do they do?
c) What are the average size of their trades?
Measured analytics
The three analytics that we chose were selected because they had come up in our conversations with crypto traders earlier this year. After an initial technical evaluation of each we decided that it is within the realm of possibility to build them, so obtaining information on their real usefulness to the trader is meaningful.
Wash Trading
The first feature that we selected was wash trading, the topic that has been well documented and debated in the crypto community. Originally the discussion about the impact of wash trading started with the Bitwise report, and since sprouted to become a lively debate with it’s proponents and opponents. We described the following statement to the respondents to address how wash trading can be removed from the market data:
Crypto exchanges are known to have inflated trading volume numbers. One way to get closer to the actually traded volume is by removing wash trading. We remove it by excluding trades that occur inside the spread — between highest bid and lowest ask.
Order book queue position
The topic of order book queue positioning came up in our conversations with crypto funds where some of them found it difficult to identify opportune moments to move their orders in the limit order book. We gave the following description of this topic to the respondents:
When entering an order in the order book of an exchange it is not always known what is the position of your order in the queue at that price level. It is possible to track the order to understand how much volume there is in the same price level before it gets executed.
Exchange money flows
The final topic we measured was to address the ‘black box’ nature of the crypto exchanges and brokers, and to shine light on them by monitoring the financial flows to their wallets. We described the topic as follows:
Money movement towards exchange and OTC desk wallets can have an impact on the cryptocurrency prices. Multiple wallets of the same company can be clustered together and transactions to the clusters can be sent as a notifications.
Demographics
The demographic questions were left optional as we did not want to leave out traders who weren’t comfortable giving info about their trading habits. 80% of the respondents, however, answered the demographics questions which was great as it really helped us to understand what feature is interesting to whom. For obvious reasons, the following data only includes the respondents who filled the demographics data.
Trading strategies
When indicating the trading strategies it was possible for the respondents to select more than one. To keep things simple we measured the frequency of each individual trading strategy. We observed that among the respondents the most frequent trading strategy was long hold (27%) followed by technical trading (14%), swing trading (12%), day trading (11%), momentum trading and long short (9% each), and fundamental trading and arbitrage (7% each).
The distribution of trading strategies overall seemed quite even, with a small spike in long hold. Because the respondents could select multiple strategies we also looked at how prevalent was it to have more than a single strategy. We found our that the majority used a single trading strategy (51%). This was followed up with 2 (21%), 3 and 4 (10% each), 5 (4%), 6 (2%) and 7 (1%).
Number of trades per week
We asked the respondents to estimate how many trades they do in an average week. We found out that the largest group was 1–10 trades /week (46%). It was followed by 1 or less (25%), 10–100 (22%), more than 10,000 (3%), and 100–1,000 and 1,000–10,000 (2% each).
Average trade sizes
Finally we asked the respondents to estimate the average size of the trades in a similarly fashion. We found out that the largest group for trade sizes were $1,000–10,000 (39%) followed by $100–1,000 (34%), $10,000–100,000 (14%), $10–100 (11%), $100,000–1M (2%), and larger than 1M (1%).
Results
It is clear that there are multiple types of traders. In order to derive distinctions between the level of trading activity we defined two groups: a) all traders, and b) active traders. We were left with 60 traders in the active traders group after the filtering. This amount we still considered significant enough so that we could draw meaningful conclusions.
The purpose of this grouping is to separate the (semi)professional traders from the retail traders as our goal was to find out what is the change in traders preference when their trading activity increases. The separation was done using the demographic data such that:
- The traders who selected “Long Hold” as their only trading strategy were excluded from the active trader’s group. This was done to remove traders with only a passive strategy as in our experience strategies that solely rely on long hold are based more on purely fundamentals of the asset rather than particulars of the market it is purchased on. As our features focus on exchange analytics we wanted to remove traders who solely rely on asset fundamentals.
- Traders that had weekly trading volume of less than $100,000 were excluded from the active traders group. The trading volume was calculated by selecting the higher range from the estimated trading volume.
For example, if estimated amount of trades is 1–10 and trade size is $1,000–10,000 we get a range of $1,000–100,000 of weekly trading volume. In this case the trader would be qualified as an active trader as the highest amount in the range is $100,000.
Finally, to measure the change in the trader’s interest and perceived value of the analysis we included relative delta to indicate the change in values between all traders and active traders.
Importance of the analysis
We observed that there is a clear interest in all of the features in both of the groups, active and non-active traders. Approximately 80% of all traders indicated interest in all of the 3 features. The interest level remained approximately the same for Wash Trading and Order Book Queue analysis for active traders, around 80%. The interest in exchange wallet flows, however, was significantly larger among active traders, at 95%. Basically this means that almost every single trader in the higher range of $100,000 weekly trading volume finds exchange wallet flows as an important feature.
The relative delta remained insignificant for wash trading and order book queue positioning, indicating that as the weekly traded amount increased there was no significant increase in the importance of the features to traders. However, there was a significant 21% increase in the importance of the exchange flows analysis as the weekly traded amount increased.
Value of the analysis
In order to understand the financial value of the given feature we asked for the willingness to pay $100 or more for the feature monthly. Again, the values among all traders were surprisingly similar for all the features, around 30% willingness to pay for each of the features. Among active traders there were some differences. The largest willingness to pay was in wash trading analysis at 39%, followed by order book queues at 35% and exchange flows at 31%.
The relative increase in willingness to pay was the largest for wash trading, indicating that as the weekly traded amount increases the willingness to pay increases by 33%. Also, willingness to pay for order book queue analysis increased by 14% as the traded amount increased.
Improvements of the research
In these results we left out the specification by trading strategy with respect to all traders and active traders. The reason for this is that for some trading strategies we didn’t receive enough responses to be sure that the results are not result of a couple anomalies. In order to improve the results we would need to ensure that we have a larger number of answers for some of the trading strategies.
Conclusions
Based on the answers we can conclude that the most interesting feature is exchange wallet flows while wash trading and order book queue positioning did have clear interest too. Perhaps one of the reasons for this is that the exchange wallet flows is a novel data source, that offers a great potential for real time predictive analytics. Our research has shown that it can indicate very significant price movements in the markets.
Interestingly, the willingness to pay, however, increased on the other two features, while for the exchange flows it remained approximately the same. This shows one thing for sure: the interest and willingness to pay are not directly correlated. There can be many reasons for this, but perhaps one to explain it is that wash trading is a well known issue and order book queue positions are already available in traditional markets. This means that the newest and most ‘exotic’ analysis among the surveyed ones was the exchange wallet flows. As we all know, what is new is usually much more exciting than what we already know.
Another rationale for this could be that because the exchange wallet flow data is new its position in the trading value chain is not defined well enough yet. Not understanding clearly how the data can be used and what its implications for the trading algorithms are can reduce the willingness to invest in the solution as it can be considered more experimental data.
Conversely, for the wash trading and order book queue data the use case is very clear. Basically, the first is a curated data feed that removes noise (fake trades) to be better capture the signal (actual market activity). The second, an order tracking system that keeps probabilistic track of orders in given price levels. These both can have clear and direct implications for trading profitability where that is not necessarily the case for exchange wallet flows.
If you want to hear more about these features, or learn about our work you can contact me at ekin@chainslayer.io or visit our website.
