I will backtest and see but I was thinking of doing it on the daily time frame. My hope is to avoid the HFT noise but we'll see when I test live.
Most of the posts here on ET and discussion centers around getting into a trade. My personal WAG is that 90% of the posters here on ET devote 90% of their time and energy and research to getting into a trade - and that IMHO is pure misguided naivety and a tragic misunderstanding that keeps most ET Members from ever giving themselves a real chance at success. In terms of a successful strategy to me, at least, that means a comprehensive approach that includes - dare I say it and the boring and cliche - position management ( taking a profit and taking a loss in a systematic fashion that mitigates discretionary bad habits and self-sabotage ), and brutal introspection in terms of a truly acceptable risk tolerance. It seems that these attributes to a successful strategy are glossed over or paid mere superficial lip service. A "trade" means that you've not only "entered" the market - but you've taken a profit or a loss at the conclusion of the trade's lifespan. Taking on too much risk to the point that it dominates your emotions and biases your trade selection approach is no good either. A successful strategy to me requires consistency - and that means a rules-oriented holistic and comprehensive approach to the trade lifecycle. Exerting 90% of one's energy and focus on trade entry in my opinion is akin to saying that the most important part of the human experience is just being born and that the rest of your life will simply take care of itself somehow.
Your example of complex news not coded to be machine readable would be one. (#3) I think there is another example of predictable disadvantage to algos which relates to #6. They are weak at complex/fuzzy/intuitive work and strong at speed/calculation work. Therefore (I assume) strategies are going to coalesce around: A) pure arb/stat arb B) automated news dissemination C) microstructre analysis in the product / correlated markets - next tick prediction stuff D) momentum ignition / sweeping / flipping / spoofing / type strategies E) market making Which of these algos could be weak vs a manual trader? I'd go for market making. Restrictions are maximum position size / inventory permitted by the risk profile, policy to exit positions by the close as you mentioned, limited ability to avoid adverse selection. One would assume that market making algos will bias their quotes according to inventory, and in some conditions withdraw from quoting on one side of the market when risk limits are reached. In dire cases crossing the market to liquidate positions. And there will be several of these in competition but doing similar things based off similar inputs. In some circumstances, an intuitive manual trader could identify condition where type E is longer or shorter than desired, A&B (speed) are not relevant, where C would operate as usual, and if D is present it actually helps the trade. In other words, if the manual trader was good at intuiting when enough market participants would cross the market, causing an imbalance and further similar trades, forcing type E to withdraw/bias quotes/cross the market to exit while this new flow continues, then this would be one of the best places for a manual trader at speed/fee disadvantage to trade. It occurs this knowledge would also be very useful in designing a momentum ignition strategy. If one were able to consistently and accurately read the liquidity of the product and under what specific conditions to pay to sweep the book. This strategy would be almost an inverse of #1, trading far less frequently, but with an expectancy per roundtrip an order of magnitude larger than your example for #1. You're getting paid for being better informed about future order flow than automated market makers. Not easy or I wouldn't write this here.
Many complain about the influx of low latency automated systems in the markets these days. Personally, I think it is great. Many of these HFT strategies are doing the same things, which creates predictable situations that can be taken advantage of by an automated trader (and maybe a manual trader) with less infrastructure. I.e. the single or small group trader. Think of it as a swarm mentality. When something goes negatively for one strategy type, you can assume many others are in the same boat. This causes a race for the exits at speeds that most can never take advantage of. However, the backdraft of those events can be traded. Just need to spot the footprints. (This works in reverse as well. If something goes positive, and there is a race to gobble up shares, you cannot participate at their speeds, but can trade the back end of it.)
How fast is "fast but not overwhelmingly fast" and do you think this would be potentially accessible to retail as well as 3) and 6)? In other words, is this on the order of microseconds (direct connection required), milliseconds (execution platform, e.g. TT), seconds (good Starcraft player) or longer? I would guess that most quickie stat arb would be sweeped up in microseconds as stale quotes are picked off, but maybe I'm wrong.
Either direct connect or a custom TT setup where the gateways and the API are all on the same server. It's not an officially supported configuration and most FCMs won't really like it but people specialize in setting those up. At around 150 mics it beats any other non direct connect solution I know of. That's about as slow as you'd want to be. Not the very top tier but at the top of the second tier.
Good Starcraft players can make about 5 rational decisions converted into action per second. A huge advantage for some types of manual trading. I should sponser a tournament.
I'd download that and play to keep in practice, but don't think it's a great game despite it's huge following. But yeah, quick thinking and reflexs are required for intraday trading.