Yes. You find that when you run certain strategies, you need things like rebates, support from exchange, and direct connections to the exchange that retail brokers tend to abstract away from you. Even Genesis (one of the better retail brokers) gives you nothing more than a consolidated/modified view of the original data that a given ECN posts. That direct connectivity is so expensive that you see me begging in the auto-trader forums for slices of data just to test ideas. I've been begging people to let me colo (and pay a lease) a passive data collector in their racks. I have been spending the past 4 months trying to eek out an edge in retail trading. It's damned hard, because 1) either the execution framework is too poor, or 2) no rebate pass-through, or 3) broker will execute you through their internal framework, 4) commissions are way too high, 5) if it's a dealing desk (like in the FX world), you will never be able to passively offer and collect a spread, 6) you don't get your own orderflow to execute against or improve against, and 7) the market is quite efficient at slow time frames on most of the products tied/connect to any of the major indices. In short, the best opportunities that a person could theorize with an education straight out of school are all very well tapped. Retail trading is a VERY hard game to play. When you see these characters on this forum posting profits every day in the PnL thread, you have really revere some of them.
You mentioned that you run millisecond algorithmic strategies. I think lolatency has answered my question, but if you didn't have to make trades so rapidly, would you be able to use slower infrastructure and save cost? I guess that there is no edge in doing what you do on a longer time-scale, thus you have no choice but to pay for the infrastructure.
I appreciate everyone's comment here, and maybe I should clarify my motivation for asking this question by posting some job descriptions that seem interesting to me. Quant Trader/PM - High Frequency Equities Required Qualifications include 2+ years high frequency and intraday trading strategies with equities, a proven track record, solid hands-on technical skills and an advanced degree (PhD, MS). Statistical Arbitrage Trader / Quant Researcher Candidates should have experience building strategies from scratch. ...... Candidates should have strong statistics or computer science backgrounds and understand what it takes to build up a stat arb strategy. While it's very obvious to me that most of these high-frequency, stat-arb, quant trading jobs require advanced degrees in physics, math, or CS, it is not obvious to me how those majors are relevant to building actual strategies. Understanding the hardcore math and modeling doesn't necessarily translate into a profitable strategy. Or are the practical skills learned after entering the industry?
Slower infrastructure implies that you're carrying more risk. Liquidity and availability of liquidity are essentially functions of time, as is order flow. Retail traders, through broker-imposed constraints, are essentially boxed into running either long-time frame arbitrage strategies or are restricted to forecasting direction in equities. Both of these come with the associated problem that you have to have a larger cash outlay, pay margin interest rates, etc. Fine, you say, you'll trade leverage using options -- but then the retail trader gets skewered by the spreads. FX world, you never get the best price. There's other problems too -- not all brokers will pay the right interest rates on your positions while you hold them, or your data feeds get sliced/time-sample/filtered.
Thanks for your response. From that I suppose that most retail traders should stick to slower strategies where (1) we don't have to compete with the big guys and (2) latency has less effect.
It shouldn't be obvious to you. In most cases it's not necessary. But when you're a fund manager looking for investors, it sure sounds cool when you tell the potential client that you have math geniuses in the back room designing your systems.