Market Behavior: BTC/USDT Frequency Distribution of Last Digit
Market Behavior: BTC/USD (and probably all others) Frequency Distribution of Last Digit before Decimal Point
I’m regularly day trading Bitcoin within the Binance market. While I give my buy/sell orders by hand, I realize the last digit that I use in my buy/sell orders affects a lot about its execution expectancy. Due to my observations about this market behavior, I decided to gather data and check whether this is my observational bias or reality, and I find out that there is significantly difference and an observable pattern in last digit of the instrument price.
While data that I use focuses on BTC/USDT, due to both algorithmic trading and psychological reasons, my argument is most probably valid in all other trading instruments.
Frequency Distribution of the Last Digit
This graph shows each minute’s Open, Highest, Lowest, and Close Price’s Last Digit before Decimal Point distribution by their appearance frequency. Data: 1-minute timeframe BTC/USDT historical data of 7 months at Binance: beginning from 2019/01/01 through 2019/08/05. Let’s zoom and simplify for analysis.
Focus on Volatile moments
Highest and Lowest Price’s appearance frequency (zoomed between 20k-45k) grouped by their Last Digit before Decimal Point. This graph shows that in terms of Lowest Price there is a significantly higher occurrence of 0’s and 5’s and significantly lower occurrence of 4’s and 9’s. Additionally, in terms of Highest Price, there is a significantly higher occurrence of 0’s and 0’s, while the significantly lower occurrence of 1’s.
This means while price shows volatile movements:
Lowest price reaches to Order bids that end with 0’s and 5’s, and do not reach to order bids that end with 9’s.
Highest price reaches to Order bids that end with 0’s and 9’s, and do not reach to order bids that end with 1’s.
I don’t know if this is due to algorithmic trading or basic human behavior, but in any case, it sounds extremely logical. Let’s assume the price as 5903 and it is in a downtrend, falling through lowest point: it touches to 5902, 5901, and when touches to 5900 statistically 30% more order (compared to all other bids, except ones that the last digit is 5) bid’s start to be executed. This volume holds price there, and as the reason 5899 becomes less likely to be executed, statistically %45 less!
Let’s assume the price as 5947 and it is in an uptrend, rising through highest point: it touches 5948, 5949, and when touches 5950 immense amount of order starts to execute, so it is less likely to cross that point to touch 5951.
Conclusion and takeaway
If you adapt your bidding habits during trading with this knowledge, there will be a difference in the execution of your orders.
PS: I’m newly learning data science from Coursera and trying to develop my data sci skills. I have way more ideas about Genetic Algorithms and algotrading, but do not have enough coding skills. If you have passion about it, please contact me @bparlan