Through the CoinMarketCap and LiveCoinWatch websites, we can find that OKEX is the largest exchange in the world with a daily trading volume of 1.7 Billion. For a long time, there have been rumors of fake data from exchanges, but they have been dismissed because there is no solid evidence.
The world is afraid of seriousness. A programmer named Sylvie did a test by himself. The guy's original intention was just to test the liquidity of digital assets.
testing method
Collect order books for each platform and measure the impact of selling a digital currency worth $50,000 (rich people!!!) on the price of each digital currency. This price change is called Slippage (slippage), and then corresponding to the market trading volume of the exchange, appropriately increase or decrease the amount of sales, and then look at Slippage
expect
Slippage and transaction volume should generally have an inverse relationship. Of course, the inverse coefficient of each currency will be different. The logic is: Trading pairs with huge trading volume -> There must be fierce bidding between sellers and buyers -> Orderbook will be thick -> Price spread will be reduced == Slippage will be smaller
Ideal is full, the reality is very skinny
The results show a huge discrepancy between different exchanges (Discrepancy)
It can be seen from the above picture: Bitfinex, Kraken, GDax's Slippage are roughly similar. And OKex's Slippage is much higher than them.
The guy was a little confused. He removed those trading pairs with more than 4% of Slippage, and got the picture below. Same thing. The difference is so big that it has to be Logmatic (logarithmized)
This time, the guy was completely stunned. He logged into the OKex console.
Did you see that, the volume is a steady sine wave. At the same time, the diagram of the P network is as follows:
The guy decided to fight to the end, how to measure the degree of fraud of each exchange? It's 90%, 95%...
algorithm
- Collect data for each trading pair on some trusted exchanges: Bitfinex, GDAX, Poloniex, Bistamp, Gemini and Kraken.
- Run a regression test on these data to find the relationship between slippage and transaction volume. It must be simplified here, but as a general evaluation, it should be good
- Compare the difference between the trade volume claimed by OKex and the predicted volume in step 2
The guy reduced the sales volume from $50,000 to $20,000 according to the Volatility of different trading pairs
This is the regression curve graph drawn by the guy:
Plot of predicted volume and OKex claimed volume:
OKCoin about 92.9% off all OKex’s volume is most likely fabricated.
Huobi 81.8% of made-up volume
HitBTC HitBTC is slightly less liquid than the reference exchanges
Binance
The guy calmed down and thought about some other reasons for this result
- API Matters Binance provides APIs with limitations. May cause an effective brushing strategy to not work for Binance. It can be seen that a good API can indeed improve the liquidity of the exchange
- Handling fees may also have an impact. High fees mean traders have no incentive to bid, reducing spread
- The guy only collected the 24-hour average price and did not pay attention to control the variance. It looks like the result is robust though.
- Icebergs and hidden orders. Some exchanges allow to hide limited orders, but Bitfinex also provides this feature and behaves like other clean exchanges, so hidden order effects should be ignored
- Different user groups may have different trading behaviors on different trading platforms. Although based on my algorithmic trading experience, users' trading behavior is determined by continuity.
Boy's conclusion (Conclusion)
- 3 billion daily trading volume does not exist
- The digital asset market is in an absolute bear market
- Bittrex is cleaner, so can Cryptopia and Kucoin
- Due to lack of data, the guy didn't test Korean exchanges, BitHumb looks good, Coinnest or Upbit don't know
The tools used by the guy (Tool/Data)
- CCXT Python library
- Collected data is here
The most interesting thing is that Teacher Li Xiaolai also liked:
my thoughts:
- For digital currency-based quantification, be very careful about data sources. It is best to collect and verify it yourself.
In this regard, code farmers have natural advantages, otherwise the spring of code farmers is coming. On paper, I feel shallow at the end, and I absolutely know that this matter has to be done
- Simple arbitrage with bricks should still be done, but the application of complex quantitative models to digital currencies should be carefully verified.
My question:
Doing such an experiment should cost a lot ($50,000 for different currencies). Not something a student or working class person can afford easily. Will there be support from interested parties behind such research to attack Chinese exchanges and seize the share of trading volume and the right to speak? A bit of a conspiracy theory.
Favorite sentence:
The digital currency market needs regulation
Crypto need regulation!
Refer to this article: Chasing fake volume: a crypto-plague
古人学问无遗力，少壮工夫老始成。
纸上得来终觉浅，绝知此事要躬行。