Yes. The general idea is that you can't compete with HF players (unless you invest a lot of money into technology). Instead, you want to find longer-term strategies (from hours/open-to-close to weekly hold times) that can assume aggressive execution and have high pnl/tradevalue.
How long are you planning to hold? What is the periodicity of the movement you wish to capture? If you're looking at a 7-10 day window, then an hour is a good "tick"...
i'm looking to mimic the actions of a high accuracy day trader, so My thought was second data would be useful since that's what traders look at.
It has been my experience that shorter time frames become more challenging as slippage and commission can erode any edge you find.
I'm curious, why do you think traders look at seconds? I agree with you that for day trading, tick by tick data is what you want ... for example TnS or VAP. You can try CloudQuant for free. You can't get the data out though. Mimic away!
If you want tick data you can get it from TradeStation going back 6 months on CME futures, stocks & indexes, forex, & options. It hardly takes much $ these days to open broker accts. Plus you get all that pretty charting & live market data with no monthly fees as long as your balance stays above $2k.
Question to TF boy. Is it polite to ask before accessing someone's camera on their computer? It's improving the service actually, sorry, got it. Either way some people don't have time to check for this nonsense so its worth letting people know you do it. That way, if you are chopping someone up you know to temporily opt for a less good service for those few minutes before resuming full service quality. If t reading WSJ or on elite trader no need. This is totally legit @good2no. I actually really like the voice search btw.
The second point I would make is in regards to your non-sense justification for how the process works i.e. it only makes sense in distribution terms on massive scales, noone is individually identifiable, a single point is irrelevent anyway. I would reply that although tensor flow sounds like fluid mechanics where individual randomness can be ignored that's not how it really works or what the neural nets are doing. What you actually want is a very large spy net to find the important particles, ideally a few orthogonal ones then watch those very closely. The idea that you haven't noticed most data is not worth collecting very vigorously i find to be ungreen, since it is a waste of the world computing resources.
We are all involved in micro-structure games. Beating the very fastest at SPY v SPY in chicago for +/- a few cents is really for brokers anyway. There are so many instruments the same money is spread out over on so many exchanges, just collecting market data alone is a big challenge. You need throughput as well as latency and you don't have to only hold for 5 minutes or not be interested in anything about markets. question. 3*Leveraged equity etfs exhibit drags on performance -they are pulled towaards the centre. A large reason is the payout formula they have in the prospectus, fees, resetting leverage with options etc. Want a properly convex long vol strategy which carries positively and isn't a tail hedge. Can i do it being short this and long equities to offset the drift? Vix high strikes have too much rr gamma and don't carry free, (unless its a 5 cent ''tail hedge'') What ideas do you guys have?