We are glad to announce that CoinDr is now officially online! This time it comes with full use case for BANCA token in alpha version. Although most contents are free in alpha version of CoinDr, users can purchase access to VIP researches using BANCA token. Currently there might not be many VIP research contents on CoinDr, but there are expected to be many more in the future. As always, we will continue bug-fixing and improving of CoinDr platform during online testing.
As many users may be curious about the contents of CoinDr and their underlying models, we are going to give some brief introductions to key CoinDr reseaches.
Generally speaking, researches involve these steps: 1. Gathering of data; 2. Making of assumptions; 3. Backtesting; 4. Tweaking and Finalizing of models; 5. Applying model to current market status and presenting results. Most of our data comes from our proprietary web crawlers, third-party APIs and our own websites (e.g. CoinAI). We make sure that proper assumptions are made before backtesting is performed to prevent potential over-fitting. This is a repeated process because most assumptions are proven wrong and thrown into garbage. It also takes a very long time. Once we have a proven assumption, we pick the best model to track the parameters. Then, periodically in most cases, we will re-run those models and finish write-ups for our research papers. In this way, we make sure that the most complicated researches are presented in a user-friendly way.
The first example we want to give is the ERC-20 Market Observer. It already has the first issue uploaded to CoinDr as a free research report. This report aims to keep track of the ERC-20 market, which is one of the largest categories of the crypto market. It include three parameters we have created, size of average holding account over time, degree of wallet concentration and turnover ratio. The data of this report comes from multiple sources, including but not limited to different exchanges, etherscan as well as other public and private market data sources. Some key data has to be obtained through deep web crawlers. Then, we process the data so that it will be more meaningful. For example, in size of average holding account, we have to adjust the size of accounts according to fluctuation of market prices. We would like to see how much those accounts are holding each ERC-20 tokens over time, whether they have increased/decreased their holdings. Taking market price effect out of the analysis can make it more meaningful. According to our backtests and real tests, a rise in holdings in top accounts are usually linked to more optimistic outlook for the token. Similar work has been done on degree of wallet concentration and turnover. All the numbers shown on the report are based carefully gathered data and are processed by machine learning models that are backed by a lot of testing efforts.
Another example is Momentum Factor Tracker report. This belongs to the Deep Factor Research series. This paper is based on deep factor research. It takes “OHLCV” data on hundreds of actively traded cryptocurrencies from third-party API. Those data are then screened against a volume threshold (we do not want to include cryptos that are too thinly traded), leaving us with about 50 most actively traded ones. Then we construct momentum factors on the historical returns of those cryptos. A momentum score is then given to each of those cryptos in each period in history. Then, results are shown according to different aspects. First, we look at factor returns. This includes the amount of return on long/short portfolio (long cryptos with high score and short cryptos with low score), robustness test for the results and cumulative return by factor quantile. Then we check the IC (information Coefficient). This is a key parameter in factor research, representing the correlation between predicted price moves by the model and actual historical price moves. After that, we check the turnover of the hypothetical portfolio to see how realistic the resulting strategy is. The paper consists of Factor Return analysis, Information Coefficient analysis, and Heapmap analysis from 2018-2021 period. On the Turnover analysis, one can see mean turnover by factor quantile, top vs bottom quantile and factor rank correlation. The factor rank autocorrelation are produced on 28 days period over the last two and half years.
- Development and internal testing of CoinDr is finished.
- Alpha version of CoinDr has launched, with brand-new BANCA main website.
- Online testing of CoinDr has started.
- Research for CoinDr reports continues.
The BANCA platform serves the global cryptocurrency community. BANCA’s dynamic eco-chain uses AI and expert system that includes automatic management. The BANCA platform analyzes Big Data and delivers precise services tailored to the specific needs of our individual users.