I tested my strategy on the top 200 stocks from the S&P 500, over 20 million trades tested since Jan 2000, or whatever year the ticker was first listed. The dollar volume yielded was amazing, but the average net profit was only 0.04%. ~20 trades per day per stock. This tiny margin makes me very vulnerable to slippage. How to the HFT strategies that execute millions of trades per week survive? Could 0.04% avg net profit per trade actually be profitable?
It’s very expensive to run HFT strategies, so they trade liquid instruments. When you trade thin instruments, your fill rate will be very important and you are likely to experience significant adverse selections.
The tickers I tested are all extremely liquid (TSLA, NVDA, RIVN, PLTR, etc). Would slippage be a realistic concern for these types of stocks? Should I expect to lose the average spread on each trade?
Your strategy will either provide or consume liquidity. If you are worried about slippage then I assume it will take liquidity. Yes, it would be a concern because the prices move fast and you may not get filled within acceptable slippage. That’s where the latency becomes important.
I'm actually using limit orders in my strategy backtests, so providing liquidity, but in real life, I assumed (maybe incorrectly) that I'd have to switch to market orders to get filled?
sell it to hft shops. they might be interested go live in small size and see adverse selection in real time
I have worked with a few automated HFT firms putting on larger positions for either for a few hours or a few days. They do extensive testing to determine the best way to get in and out with the least amount of market impact. The path is going to be different for each strategy depending on size and time frame. There is no holy grail answer. What we offer is a access to Exchanges, ECNs, SMART routes, Dark pools and ALGOs. Some also add Options for more leverage.