China repegs. The GDP is nowhere near the government stated 6.x %. Probably close to 4 is more accurate. This chart is almost six months old. It would be nice to update it to today. In the meantime, the US markets continue to steam higher. On what, God knows. http://www.bloomberg.com/news/artic...-fixing-to-lowest-since-2011-as-dollar-climbs
Industrial revolution but it will be scientists that is out of a job. I almost feel we should just stop all science except to put all into AI. Quantum computing is already a reality and will replace intuition. Chess was beat 10 years ago, Go was beat this year...which was the last straw for me, and so in 10 years time, I think every other field will be beat.
Classic divergence on this rally. My systems don't get it either. If markets were physics, this rally would be explained by "Dark Energy." The conspiracy theorist in me says this is China trying to paint the tape for the FED to hike. Then it would ease its' own need to manipulate the Yen/$. This rally is 'false' and here's the tell: Trader Stocks continued to grind higher Wednesday with the S&P 500, Dowand Nasdaq tracking for their largest two-day gains since early March — but the lack of participation in a crucial area of the market has one trader calling foul on the rally. "I'm taking a look at a very interesting divergence in two key markets that I think could lead to a potential false breakout in the unbelievable rally that we've seen in the S&P 500," Todd Gordon told CNBC's "Trading Nation" early Wednesday. The S&P has surged 2 percent since Tuesday's opening bell, pushing the large-cap index near a one-month high... http://www.cnbc.com/2016/05/25/this-rally-is-false-and-heres-the-tell-trader.html
For those that are setting up a database, here are some average numbers that I am seeing as it relates to disk space requirements. I am storing tick data for about fifty active symbols (for now). A single Cassandra node (for now) is easily able to easily handle the load. The disk requirements are ~1GB per week. So to store a years worth of data is approximately 55 GB per year. With todays HDs having 6TB or more, that is nothing. When I add equities and longer expiry instruments, the amount of data will probably go up by 10x. That is still only 500GB per year. This is well within the means of the retail trader.
But why will a retail trader want this, if he can't submit orders cheaply or quickly? Better to do machine learning on OHLC.
Data is the wellspring from which all empirical scientific theories are matched against reality. From tickdata, you can resample to any time bar you want, or even tickbars. If what you are storing instead is OHLC bars, you are saying there is no other interesting way to compress the data to feed algorithms. I leave all further connections to the reader.
Tick data is not the lowest level of granularity. You can actually store the limit order book, as it changes from moment to moment. For some active securities, such as the ES, this would take about 2Gb per ticker per day.
It is probably more accurate to say, a market cannot stay efficient across all time frames. But it can be very efficient on several time frames at once. If that is true, and you don't adapt to it, that means there are periods where you will either force a trade incorrectly, or you are forced to sit out of the market for extended periods of time. Therefore, preferably, you must find systems that exploit inefficiencies on multiple time frames on the same symbol. Alternatively, a different asset class on the underlying might be inefficient, but that is harder to ascertain, I think. For example, if the underlying is efficient on the daily time frame, the options on that underlying might be overpriced.
It may be profitable to think about this as saying that this is the markets way of a particular instrument being in a superposition of states. Although, it may be more complicated as the system can also be in a temporal "superstate".