Your question is a fundamental one that has been debated extensively in academic literature. It is the "make or take" question. You have already identified two major factors in the decision process, the time to execution and the probability of execution. The probability of limit order execution is dependent on the distance of the limit order price and the inside market, the position of the order in the order queue, the stochastic order cancellation rate, the stochastic arrival and size of liquidity-taking orders, and other lesser factors. If someone were to answer your question with some percentage, their answer would be highly subjective and dependent on randomness. How exactly limit orders are queued and executed is very specific to the exchange in question. Some electronic order books use a price/time matching algorithm. Some also have an auction algorithm working in tandem. Market orders are always given priority over limit orders in the markets I am familiar with. You might consider a combined approach. The fact that you are considering a limit order strategy at all implies that you have the luxury of time. You could submit a series of limit orders for a specified maximum time limit, then cancel and resubmit market orders for the unfilled remainders. However, I think you will be better off selecting one strategy or the other, and avoiding this kind of complexity. Is the difference between the bid-ask spread really the difference between profit and loss for your strategy? If it is, how sure are you that you are not trading white noise? -segv
Paul, Have you considered trading longer timeframe strategies? Stat arb is very competitive and as a retail trader it makes it that much more difficult because you have inherent disadvantages that the IBanks and institutions don't (such as increased latency and higher commissions). If you're attracted to arbitrage because of its close-to-zero risk attribute, there are other ways to achieve similar results without the risk of slippage and commissions ruining your edge. For example, trading several non-correlated strategies across many timeframes, instruments and markets, results in a smooth equity curve with very small drawdowns. Like me, you prefer high-frequency trading. I have been able to test and optimize my edges over many years of data and thousands of trades. This results in systems with high statistical significance due to the large trade count, and lower drawdowns and a smoother equity curve due to trading uncorrelated strategies/markets. Your strong background in computer science will give you an advantage over most other retail traders trying to reach the same goal. By hanging out on these forums you'll pick up a lot of the jargon and gain knowledge in the underpinnings of how markets and exchanges work (on both a technical and economic level). You seem to be on the right track, don't be afraid to ask questions and don't give up! - Paccc
I noticed something while trading the US Jobs news release this morning on my paper account, the arbs seem much nicer than arbs I detect during normal non news times. They appear larger and stay around longer than normal 'spike' opportunities which only last a tick or two. Does anyone know if this is just an effect of the higher volitility in the market at the news release or do big institutional traders either suspend or change thier arbing for news releases to make the arb act differently?
my guess is, it is due to delayed data. iow, what you see is not what you get. during bursts of high volume, there is often delay.