Testing real-time forecasting systems - page 81
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
That's for 1 hour of calculations ;)
The position closed on the stop:
For now, no comment, it's clear as it is:
Moving on...
option:
(while I was calculating and correcting the error, 5pc of 15 minutes passed)
there is also this option
baseline forecast:
>> baseline forecast:
Sergey, your forecasts are good, but only at the initial stage. My opinion, countdown should be within half a day, a day maximum... It will be a very workable option for real and profitable trading.
Sergey, your forecasts are good, but only at the initial stage. My opinion, the countdown should still be done within half a day, a day maximum... It will be a very good variant for a real and profitable trading.
This is true. Now I'm working on a more precise identification, because among the possible "attractors", there is always a working one, you just need to choose it correctly. So far, the method of maximum likelihood (in my version), is a little bit malfunctioning. Conceptually, I have understood where the main error lies - in the inaccurate definition of the length of the historical time series, maximally affecting the future, i.e. the "memory" of the series. The estimate turns out to be a bit biased. I will think about it.
This is true. Now I'm working on a more precise identification, because among the possible "attractors", there is always a working one, you just need to choose it correctly. So far, the method of maximum likelihood (in my version), is a little bit malfunctioning. Conceptually, I have understood where the main error lies - in the inaccurate definition of the length of the historical time series, maximally affecting the future, i.e. the "memory" of the series. The estimate turns out to be a bit biased. I will think about it.
>> By the way. Here's a question, has there been a test on past data? After all, you can run a year in a week (assuming machine resources are available) and try to isolate the grain that is gold... Sorry if I'm being silly, the model must have been built over a period of months and not from scratch.
By the way. Here's a question, has there been a test on past data? After all, you can run a year in a week (assuming machine resources are available) and try to isolate the grain that is gold... Sorry if I'm being silly, the model must have been built over a period of months and not from nothing.
I test the model in MathCAD. As the forecast is for one - five days, I may not bother with modelling of ticks and other stuff (moreover, it's not so easy). Intersection of a trading level on the planning horizon with the price - respectively, plus TP in the piggy bank, or minus a number of lost points. One of the serious problems (in spite of good results in forecast execution) is a stop loss, it is not clear what to do with it. Most often it is +/- SCO. I trade manually using this system at the same time. I have started to use MT, it proved to be not so easy. May be I was mistaken somewhere, but there is a discrepancy with MathCAD calculations in accuracy, starting from the third decimal place. This is a lot, I think I must have made a mistake somewhere.