It's just fine to use the entire data set, and to use the cumulative returns of 2014 for me as an example of 70% ytdcum that cannot be beaten by your straight nine innings of three strikeouts on 9 pitches you've taken in your 0.000 avg, and now I hit the only pitch from you as a pinch hitter in the top of 9th so as a straight trading Modell don't forget that the machine learns on the data it's given just as you would only learn the information I've learned by giving you every book I used in my education so that if I have to justify my returns, they're all attributable to this understanding of machine based learning being no different than how a human would try to study to find something was done incorrectly and try to correct it. The point is that you all are saying you shouldn't correct it and the fact is very data dependent but let me leave you with this argument: as long as you have May 6th, 2010 in your model, trust me, nuclear war doesn't show up as eminent very often and if you ever see that again, I don't think Washington or New York City will be around. I go back to January 2008, and you don't need anymore data. Every time futures prices change the proportional percentage of profits will still be the same but they will not correspond to as much if the contract has declined in value, and until you all realize that just as I took the mean risk adjusted analysis of a genetic algorithm to mean as in learn by re-analyzing that's just the same thing the human brain does in a Financial Analysts role at a fund, but the Quantitative Analyst you're talking to, Quant Master, uses the values given by the portfolio optimization to fit to risk, not necessarily to the maximum profit. So when I have to justify what writing this way is all about: USE ALL YOUR FUCKING DATA! Re-optimize every so often whenever there is a structural change, such as when corn crop reports come out at 11:30, then see that that made such a difference in the future of the model going forward that you can bet if you have to re-optimize everyday with certainty, it's never going to happen! So use your best judgement in how you implement the model and plot them on the efficient frontier! Standard Deviation annuallized with the return is how to structure the frontier just the same as any other investment. And you can bet we'll have a risk on day that will be so much like the insane drop we saw in the market May 6th, there might be a 10% spike in indexes and this will be said to be a sign of hyperinflation, but nobody will care because we'll all be phenonomenally richer. So go peter and begger off, Jack Hershey! We'll see which machine makes the rule BANK ROLL CASH MONEY HELLABANK! BANK TODAy! BANK TODay! BANK To-DAY! You don't have any clue how much more knowledgable I am about this subject having run constant genetic optimization out to 27 Quintillion Trillion possibilities, then tell me if when I practice around what's in front of me, like you and me should step in a cage and go bat! I was an Springfield, Illinois All Star Ball Player, and the approach I took to learning required more practice off the field to get that good than it did listening to you telling me there's a 0.300 to your wiffs on 27 straight pitches all strikes. And we'll see that eventually, but in the greater moment, I know where the risk is arbitraged and it's in the back month of every year! I worked out better how to use some of my leverage and endorsing trading in anything other than the front month should move indexes up by the fact that as long as there are that many contracts in the back month you should short the back and long the front accordingly if you have multiple contracts to trade, and I don't plan to mention anything more than if there's any more criticism on your part about me, I will seek to have you banned! You insult a professional with those returns like he doesn't have a response and the response is: THIS IS IT! THIS IS HOW TO DO IT! Get data going back as far as you want. Train the computer based on structural change and efficient frontier analysis. There! Now is that simple enough for everyone to do? That is the only reason for trading in the first place is to manage the risk that your dollars will become so much less valuable than somebody elses, iewners of equities, , that by the time you all have a chance to even find an approach the best way is to use the best fit! PERIOD! BEST FIT! ALL THE DATA YOU HAVE. Hopefully more than 5 years. Plenty of plum but so long as May 6 2010 is there, you have all the risk that can ever possibly hit the pipes or else if you're in major metropolitan areas like Chicago, New York City, Washington, etc., You better prepare for that flash of light and hope that nuclear fire is as less painfull to burn in the blink of an eye as no matter the case may be from the radioactive scorch of incinerated atmosphere most likely following any other drop as large as that. And I have to say I won't be so merciful on my trading, but piddling around is not what goes on at my trading desks! And, I can say I have a car to show for it and still flush with cash ready to pounce on a 5 1 and 2 with 2 extra spread fiscal year hedger and a 2 1 and 1 trading first month end Last Month out Straight Hedger. Minimum equity for cta's should be determined by the amounts required to keep one unit of a CTA's model open, not by any finite standard.
Use the one-three-nine rule. If the number of system parameters you need to optimize is X, then re-optimize after every 3X trades and use as your optimization period the most recent 9X trades.
My point of view is backed by prints. I will post five consecutive days of trading. there you will see trades called by beginners. I push the buttons on the related MAT type account on a platform. you can see how compounding works by doing the calculations on the days. if noting else a person can calculate the compounding through the days as the buying power (BP changes. All these days show is that capital was doubled by beginners in 5 days. It may be known to you how many trading days exist in one year. If five days is divided into that number of days, then you have the exponnet of 2 to which 2 is raised to multiply the beginning od a yearcapital by to get the year end capital made. Obviously a trading method is not important. What IS important is compounding profits. I am going to ask some learners to do these calculations. when they do then you can read about how to do it as far as calculating is concerned. You, of course, have finished having the ability to learn. this was caused by your improperly developed system of inference (see results by reading all this guys posts). inference that is irrational cannot be used to observe anything or make ATS's.
I'm saying ubiquitous. Invariant to not by none but to ALL data you give it, even loginormally distributed prices, and that's anything that has value downto it's most maximally reduced value of 0 and this means it applies to all instruments. Will someone please get me a goddam supercomputer to compute the first billionth digit indivisible Mersene prime number?