Executing Large Crypto Orders Across Exchanges and Order Books
Summary
The document asks how to sell a large cryptocurrency position quickly while limiting market impact. It contrasts an immediate market order, which may consume multiple levels of the order book, with splitting the trade over time, which may reduce price pressure but can expose the seller to price changes during execution. The question mentions TWAP and VWAP scheduling and proposes stochastic optimal control, with live trade and order-book data available, but does not develop an implementation or execution formula.
The answer suggests using order-book depth and resistance to price movement across multiple exchanges when parallel execution is possible, allocating larger portions to venues judged more resilient. If limited to one exchange, it says execution pacing depends on pair activity and cross-exchange arbitrage responses to price differences. It cautions that arbitrage bot activity is difficult to predict. These are qualitative suggestions, not a tested optimization method; the document gives no measured performance, calibration approach, or guarantee that splitting orders will improve the realized price.
Key ideas
- A large market order can consume book depth and move the execution price.
- Splitting a trade may limit immediate impact while introducing exposure to price changes during execution.
- When multiple venues are available, the answer proposes studying order-book resilience and distributing sub-orders across exchanges.
- On one venue, suggested pacing considerations include pair activity and cross-exchange arbitrage responses.
- The answer is qualitative and does not provide a validated timing or sizing algorithm.
Tags
Full text
# Best way to buy and sell large volumes of crypto # Best way to buy and sell large volumes of crypto I had a few questions about how to properly execute a large order of crypto currency without moving the price much. I know a lot of funds employ a TWAP/VWAP algorithm to liquidate or purchase a large number of shares. But I am not sure if this would work in an HFT style setting. What I am trying to do is this. Say I have 100 BCH (bitcoin cash) and I want to sell it at the best price I can, without moving the price much in a short amount of time (under 5 mins). If I was to market sell this, I would walk the book massively and the price would fluctuate very largely due to this. Now If I was to space orders out, it wouldn't affect price much but I'm not sure if I would get the right price. I am trying to figure out a formula to calculate the optimal timing between orders, the size of the orders, and total time of execution. I am leaning towards a model I found that uses the theory of stochastic optimal control to do this, however I am not quite sure how to implement this into python. I have access to live feeds of both executed trades, and the orderbook. Any help with this would be much appreciated. ## Answer by A. STEFANI (score 1) https://quant.stackexchange.com/a/42726 It depend if you have the capability to act on multiple exchanges. - If you can: Best method is to pre-study the orderbook topology on different platforms, then sort the platform by order-book resistance to price fluctuation. Then split your large volume sell order into multiple volume sell sub-orders, and then dispatch sub-orders in parallel on different platforms. More the resistance is important, more the sell order will be important and finally, more available exchanges = Less price fluctuation - If you cannot: The timing will depend essentially on the average volume pair activity and how this specific pair is under surveillance by inter-exchange arbitrage robots. Note that the price fluctuation (due to your order execution) on your exchange imply a price difference (between your platform and others) and it is the main trigger for arbitrage robots activation. So finally the time to wait will depend essentially on this. The problematic is that robot arbitrages's activity is not really predictable.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.