How to avoid the impact of slippage in programmed transactions?

First of all, let's talk about what slippage is in programmatic trading. The slippage in programmatic trading in my eyes is the spread between your expected price and the actual transaction price. You can give a calculation formula for slippage: slippage = tick level fluctuation speed * network delay time.

Since the market is always volatile, the market is not the cause of slippage. In the historical backtest and the simulated disc, there will be no slippage because there is no network delay time (but the market is still volatile, but there is no slippage). You can set a take profit and stop loss on each order in the simulated disc, and you will find that each transaction is based on the price you expect. Take profit or stop loss.

According to the above calculation formula, you can't control the fluctuation of the market, but you can control the network delay time. Be sure to be clear that the market you see on your computer is a replay rather than a live broadcast, and the instructions issued by the program according to this market also need to be passed in the middle to take effect. Therefore, the high speed of market fluctuations and the serious network delay will increase the impact of slippage. And this impact will actually have subversive results on the small-cycle trading level.

To avoid the impact of slippage, you can take the following three paths:

1. Increase the level of programmed transactions

In the process of programmatic trading, the average profit and loss points of a large-cycle trading level must be greater than that of a small trading level. If a large-level model is an average profit of 50 points and an average loss of 30 points, while a small-level model is an average profit of 5 points and an average loss of 3 points, in the historical backtest and simulated trading, there is no big difference between the two. They are both models that can achieve stable profits, but in real trading, they will be completely different. The former must be much more effective than the latter, because the scale of slippage and the average profit and loss points are not in the same order of magnitude.

2. Reduce network latency during programmed transactions

By all means, find the fastest way to connect to your programmatic trading server and reduce network latency.

3. Avoid specific market fluctuations at times of high speed

For example, some people take a completely evasive approach to non-farmers, clearing all their positions 15 minutes before the data is released. You can't control the fluctuation speed of the market, but you can hide if you can't afford to offend it. The non-farmers' announcement time is accurate to the second. If you don't hold a position at this time, no matter how big the slippage is, it will have no effect on you.

To sum up, 2 and 3 are to adjust the two multipliers of the calculation formula to reduce or avoid slippage in programmed trading. Method 1, in fact, does not reduce slippage, but only reduces the impact of slippage so that it does not affect your yield curve. Finally, slippage in programmed trading can sometimes increase your returns, which requires you to understand the way you open and close positions. In a word, if your order is an inverse tick-level potential, slippage will be beneficial to you. If your position squaring method is a tick-level potential, slippage will also be beneficial to you. At this time, your network delay is large, which is actually a good thing!

For example, step-back orders, as well as fixed-point take profits, slippage is your friend. When you have more than two trading hosts, you need to screen all orders and close positions. If slippage is beneficial to you, these instructions are placed on the slow network host to operate. If slippage is unfavorable to you, these instructions should be split to the fast network host to operate.