Market prediction based on macroeconomic indicators - page 52
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
The picture is nice, but could we instead calculate at each step the product of the predicted change by the actual change, sum over the whole period and divide by the same product, but in which the predicted and actual changes are taken modulo?
0.963834
Adjusted data attached
0.963834
Corrected data attached.
Thank you. Although it's difficult to conduct experiments without column names (at least in a separate file).
But the indicator is super good. If you could forecast something tradable like indices, rather than GDP with such accuracy, you could make a fortune.
Here, let's think together. I wanted to choose a simple step function:
out = -1 if input < threshold, +1 if input > threshold
...
That's exactly how about 40 years ago a chief officer in a village in Vladimirovka, Astrakhan oblast, received a salary of 3 roubles 62 kopecks. His seniority on the day of pay was not more than 25 years and not less than 25 years, but exactly 25 years :)
PS The value of 3.62, which coincides with the cost of half a litre of vodka, was adopted by the developer by default.
Predictions of the GDP Atlantean Fed:
Taken from
https://www.frbatlanta.org/research/publications/wp/2014/07.aspx
There's also an article there explaining the prediction model:
https://www.frbatlanta.org/-/media/Documents/research/publications/wp/2014/wp1407.pdf?la=en
The GDPNow model has little history so far (since 2011 I believe) and can hardly be compared with professional Blue Chip predictions. But the creators of the model show the following accuracy of their predictions depending on the number of days before the release of GDP:
Well, let's compare.
Finished the linear GDP predictions. Here are two quarters ahead:
There are 4 predictors in the model, although 3 is sufficient. After 3-4 predictors, the remainder looks like noise. Predicting the S&P500 with the same method as GDP works very poorly. I don't even show it here. I also quickly tried non-linear transformations with a step function as I described earlier. It works worse than linear regression.
Waiting for the release of the new GDP value at the end of April. Resting for now.
Here's the forecasting on the first 4 points:
On the first 6 points, and the "accurate" forecasting that you are trying to do, I don't believe in:
We continue to enter points, but, the basic forecast does not change, although, the premonition of a fall is becoming more and more real:
Just before the fall, the model does not believe in a catastrophic fall:
Even after the fait accompli of the fall, the model invites the indicator up, which will happen later and correctly predicts its value for today:
GDP is back to the forecast, as it needed to prove and show. This indicates that the crisis was a sham:
Finally, we enter all the data:
Here's a prediction on the first 4 points:
On the first 6 points, and I don't believe in "accurate" forecasting, which is what you are trying to do:
We keep entering points, but, the basic prediction does not change, although, the premonition of a fall becomes more and more real:
Just before the fall, the model does not believe in a catastrophic fall:
Even after the fait accompli of the fall, the model invites the indicator up, which will happen later and correctly predicts its value for today:
GDP is back to the forecast, as it needed to prove and show. This indicates that the crisis was a sham:
Finally, we enter all the data:
But if your model could predict the crash of 2008 and other crashes, then you could either save your savings, get a profit, get a job in a hedge fund with good pay, or ... And so predicting an almost flat line 2% average GDP growth is something you can do without models, knowing that is the US government's goal. Throw away your model. There is no use in it, neither in trade, nor in predicting the economy. Broaden your horizons, try new models and theories.
Vladimir, please don't jump to conclusions and bury the model. You set the task - see the solution! It is natural to assume that if the hedge funds had owned the model, they would have run it every quarter as new data became available. So, if they had run it 6 quarters before the crisis, they would have revealed the picture you are talking about:
After entering 3 values that don't seem to portend any threat, the model reveals an imminent collapse:
After a quarter, it becomes more obvious:
The USgovernment is taking unprecedented measures:
The collapse is unstoppable:
The scale of the crisis is staggering:
Withenormous effort, the crisis is overcome:
The economy is recovering, but the spectre of crisis is present:
After a quarter, the ghost of the crisis disappears:
But the specter reappears:
A fallis looming:
A fall, followed by a recovery:
Without words:
Finally, the latest fall should recover:
This kind of analysis needs to be done after every quarter, which I don't have time for right now.
Lots of pictures, not much use. Show only one graph: what the model predicted at each point in the odometer, for 1, 2, etc. quarters ahead. Picking a section of history and showing 2-3 tracts and then interpreting them loosely won't work that way. Let's do it scientifically, with numerical predictions for each quarter.
You are trying to simplify the task of predicting the economy by fitting 2-3 trajectories to 2-3 points of GDP. GDP is not a rock that is thrown and the laws of physics try to determine where it will be in some time.
GDPNow continues to revise US GDP for the first quarter downwards. Their prediction is now 0.1% growth for the first quarter.
This is the prognosis