Lol. That is my average round trip slippage. Go figure.
How do HFT strategies combat slippage
- Thread starter traderzhang
- Start date
Lol. That is my average round trip slippage. Go figure.
HFT firms have their servers running in the exchange and pay a lot of money for that benefit. They also develop trading FPGA boards to shave off microseconds from the reaction time. It's quite a different ballgame 
If your limit orders don't have a limit price that make them immediately execute, then you will get garbage backtest results if you in your sim assume they always get filled and/or instantly. Furthermore, adverse selection is a thing. Modelling limit order execution is pretty complicated and in most cases I would just recommend going with market orders and accepting the slippage. Of course, if you're intending to do HFT then you have to go into those complications.
Last edited:
I trade multiple intraday strategies with holding periods of a few minutes. I only do S&P500 stocks, only market orders, long and short. Position sizes are $25k.
When I total my daily total return, some days it underperforms theoretical and some days it out performs theoretical. When averaged over a longer term, the slippage averages out to about zero.
You have to look at long term averages because there can be one off bad fills or good fills every day. And even within the S&P500 there is a wide liquidity range - for example super liquid like NVDA vs not too liquid like FICO.
When I total my daily total return, some days it underperforms theoretical and some days it out performs theoretical. When averaged over a longer term, the slippage averages out to about zero.
You have to look at long term averages because there can be one off bad fills or good fills every day. And even within the S&P500 there is a wide liquidity range - for example super liquid like NVDA vs not too liquid like FICO.
HFT firms have their servers running in the exchange and pay a lot of money for that benefit. They also develop trading FPGA boards to shave off microseconds from the reaction time. It's quite a different ballgame
Don't hear much about FPGA. Any references you could point me to, I'd appreciate.
My current mind mapping that is associated with the subject;
I'm no expect so take this with a grain of salt, but with FPGA you implement the trading algos straight in the hardware, with the main idea of lowering the latency caused by the network hardware and software TCP/IP stack to shave off micro-/nanoseconds. You can get these FPGA development boards for something like $1000 though I don't know what's actually the best fit for HFT. The programming is done with a hardware description language (HDL) like Verilog/VHDL, which are C-like languages for hardware programming. Anyway, if you are not already running your servers glued to the exchange servers with sub-ms latencies, developing for FPGA is a moot point (the benefits are lost in the network latency).
You can look at xilinx for info. Chipscope. I remember impulse C was used years ago
if you use mkt order, how could it be that your average slippage is zero? With mkt order, you will pay the ask-bid spread for each trade.
Robert Morse
Sponsor
Slippage is not defined as the mid point. Slippage refers to the difference between the expected price of a trade and the price at which the trade is actually executed. If you expect to pay the ASK and you do, your fill was not slippage. I also find it hard to believe that 100% of the trades are executed that way at any broker especially with free routes.