bid-ask spread average + $0.0035/share is reasonable for back-testing... Which is what you will pay on several million shares/month. As you know... The bid-ask spread varies a lot form $0.01-0.02 on liquid stocks... To $0.05 to $0.10 or more on less liquid stocks. A good trader exploiting a decent spread might have 50% of profits FROM the spread... And 50% from quant analysis. So the spread must be your friend and profit center. If you are losing money on the spread... Then you have no chance short-term trading. Also.. Your profits margins are very important and should be at least 40-50%... Because if your profits <<<< fees... Then that will not likely sustain a business in the long run... And your variance will be higher than optimal. Also... Back-testing has countless limitations. You should be trading as soon and as much as possible... Because it's the LOSSES that will move you up the learning curve... not the profits.
I would talk to a few prop. firms and ask for a remote server to be installed in their system with DMA access. What you're looking at is having the core trade engine inside the remote server, and supplementary engines or app. stuff like GUI and etc. on your own computer.
There are exactly and ONLY 2 ways to make money trading: (1) Exploiting a market inefficiency. (2) Exploiting the bid/ask spread using Limit Orders... In effect, getting paid for providing liquidity where it's needed. I generally trade stocks with daily volume around 100K... So the bid-ask spread is typically $0.03 to $0.15. I traded 31,000,000 shares in fiscal 2007... And I would estimate that my profits were spit about 50/50 between the 2 above categories. I use quant analysis to buy/short a stock that is "inefficiently priced"... Immediately hedge it with something very similar... And then scalp, scalp, scalp to exploit the bid-ask spread. When markets go nuts like during the subprime selloffs... Market inefficiency goes thru the roof... And bid-ask spreads also widen... So volatility is always my friend.
100% NYSE Listed routed SMART... But only about 75% (and dropping) executes on the NYSE... Versus > 92-93% one year ago.
A question in my naivete: A what timescale would something like interactive brokers scale out? Wrt. bid/ask spread - one simple way around this is to use limit orders which may make your backtesting somewhat more complicated. My experience with market orders is that the price I get compared to my last recorded price differs a lot - not necessarily to my loss.
While I have your attention, I would like to ask another question related to trading costs. If your intraday equities ATS does a fair amount of shorting, how do you properly account for transaction costs on the short sale in your analysis? Obviously there is no more uptick rule. There is a restriction on which stocks can be shorted (depending on your broker) but after accounting for that, is it realistic to expect that shorting trading costs would be the same as going long (ignoring interest) ?