I'm interested in wondering how high frequency trading works in the real world. (because I have no idea) Just by playing around with numbers from minute by minute stock feeds, you can see that longing if the stock just went up and shorting if the stock just went down will result in about .01% return per trade. (you wouldn't need to trade every minute, just would have to be able to close out trades quickly if the market moved the wrong way right away). Obviously you need large trade volume per trade to make these trades profitable (vs transaction costs), is this just unrealistic? Are stocks liquid enough for this type of trading? Imagine you could create a high frequency trading system, how much would it need to be able to return per trade to be sensible? For those of you who are in HF trading, are you working more on arbitrage opportunities than consistent returns per trade?
It works the same way, no matter what the frequency is. Buy low, sell high. Making 1 cent per trade on 100 trades yields the same profit as making 10 cents per trade on 10 trades. Of course, the higher the frequency, the more sensitivity to transaction costs.
I'm interested in wondering how high frequency trading works in the real world This is how it works. 1. Become a broker 2. Convince large number of traders that the high frequency trading is a good thing 3. Collect a lot of commissions and fees. 4. Get rich!
although that is an interesting view on high frequency trading, I'm talking more from the view of a company possibly within the exchange that has low transaction costs. Can you trade 1 million dollars every minute and expect liquidity to hold up in even the biggest stocks? If you can make .01% on enough money even spreads. how does this hold in the real world. If i could make .01% return on every trade (on average) it would only take a few hundred trades to double your money. (even after transaction costs).
Unless you're a programmer, don't even think about it. You could pay someone to program something for you but IMO HF trading requires a lot of tweaking and if you can't write code then you'll constantly be bothering the programmer. The time frame of the majority of my HF trades are between milliseconds to seconds. Backtesting on that time frame is useless most of the time and it just requires real time testing until you get it right -> which requires changes to code.
To be even MORE precise, it would take 6,932 PROFITABLE trades with no unforeseen events or gaps. Good luck with that.
My question was more on the feasibility of being able to execute trades with 10000+ stocks at a time. Thanks for the math correction. Does anyone actually do HF trading? What is you average return per trade AFTER spread and slippage (without commissions factored in). Commissions aren't really that important if you can trade with large volume.