As Eight and other have pointed out, the problem is most "retail" trading platforms and data feeds don't support millisecond precision. The possible solutions: 1) Use 1-tick bars 2) Configure the platform to run your trading strategy on timer (say, every 100 ms) or on arrival of every tick.
OTR is where you begin. LeeD gave you the solution and it shifts you from inductioin to deductiv reasoning. The two variables are involved: V and P. It may be true for the person who pointed out that 200 to 300 ms is his limit. This is not true for others and there is a reason. this brings us to where you can be potentially. As you see on the OTR panes the display does have a time axis but the view is not indexing at a fixed velocity. The velocity varies. You use that and its next derviative(acceleration) to have six data streams. Here are programming means to continually compare all six. This is done in the 10 and 100ms range easily. Next you set up parallel data feeds for the components of the original instrument you are monitoring. These 12 streams relate to the original six. For making money and doing it at the granularity (OTR) of the market and the market components at any frequency, ID' ing the profit segment signals is the seminal matter. This involves when signals are generatedand when signals are not generated. When signals are not generated is long intervals. When "end effect" signals are generated is known well before they appear since they can only appear on certain combinations of the 18 degrees of freedom you are working with. These signal window is very brief; its opening and closing is like a shutter which in coding is simply a gate. These are obvious through back testing a deductively derived test model and by NOT compiling inductively based data collections which you now do. How long is an "end effect? It is well under 10 ms. How long is the lead time to that event? it is well over 100 ms. How are these ratios determined? It is found out through the examination of the chains of the three subsets of the 18 degrees of freedom and there least common cycling rate. Why does this occur and why are the harmonic implications? The nature of the opportunity is the latency range of the participants as has been pointed out to you. The participanets are also weighted by their influencing through their BP. At this point you are able to isolate these folks and separate them from other traders. How often does the harmonic dominance shift between the odd or even synthesis? Within the band width you have previously established, running twin filters allow you to see how just the Amplitude Modulation is used to give you the dominant carrier for a given period. You use frequency modulation filtering to observe the non stationarity of the bandwidth feeback system that acts as an AI for you and keeps you one the signal centered bandwith. There is kurtosis as well. use it. By this point since you only use staccato "end effect" signals, you already know that you know what proifit segment you are leaving and what profit segment you are entering all based on segment sentiments which are revealed to you by who is the "minority" with respect to your initial 6 streams and by looking at the ratio of majority to minority as determined by the majority/minority six streams each. Navy research in the mid 50's provides a rich resource for you. Search TRM. Work was done on reeves analgue computers which wewre set by 10 turn reostats (largely resistance components of RLC filters).
StreamBase, founded by MIT engineers, follows a similar approach. âWe hired 40 engineers and have them think really hard,â Palmer said. âFive years ago, five milliseconds was actually fast. Now, five milliseconds is really slow. It means thereâs a bug in your system.â
Trying to suck in another newbie? I tested buying YOUR "0 to 7 turn" per page 8 of YOUR paper (attached) on the P,V relation on 1000 stocks from 2000 to 2005 -- a total of 5000 stock-years -- using spydertrader's code for the scoring and exiting 5 days later and got the equity curve below.