Strategic foresight systems - page 26

New comment
 

Two-week plot from the 4-hour TF. Yen and franc cannot rise indefinitely. Everyone has already got into the franc and a good moment for the sharks to eat.

 
Vizard:


Anything you like... you can make daily forecasts as well...

The main condition is to feed the data into the model before a trend break or pullback ...

The red vertical lines in the screenshot show the last data points that should be included in the model...

i.e. it is interesting to see how reversals and pullbacks are predicted ...

I will post early to see the forecast itself. True, time is short, so the confidence interval has not yet been tested. Therefore, there may be errors in determining the interval. I originally thought it should widen (as you wrote about the funnel), but so far I have not seen any effect. So the interval is like the first untested constructions so far. It looks like the breaks and corrections are not predicted in any way. The only joy - trend continuation is not predicted at these points, in four cases horizontal forecast on average(back and forth(slightly up, slightly down)).


 
-Aleksey-:

I will post early to see the forecast itself. However, time is short, so the construction of the confidence interval has not yet been tested. Therefore, there may be errors in determining the interval. I originally thought it should expand (as you wrote about the funnel), but so far I haven't seen any effect. So the interval is like the first untested constructions so far. It looks like the breaks and corrections are not predicted in any way. The only joy - trend continuation is not predicted at these points, in four cases horizontal forecast on average(back and forth(slightly up, slightly down)).

ok

ps.so as not to clutter up the branch my posts cleaned up...

good luck and profits to all ...

 

Increased the accuracy a little, but:

  • I increased the forecast horizon, now it is 10 trading days. It doesn't make sense to increase it more (it's impossible to identify the model) and it's possible that accuracy will not be high and I will have to roll back to 5 day horizon. But if it is possible, 10 days will give you a really strategic forecast
  • I will be able to make a more objective error estimation only in a few weeks, so in any case do not use the forecast for trading. The rule of thumb is old, the further the forecast horizon, the bigger the error.
  • Curious moment, I have tried to use Volterri's neural network and for all 14 quotes Volterri's kernels practically coincided although identification was performed separately. Will have to check that again.

Quote Trend type Start date Duration 21.03.
P(%)		 22.03
P(%) 		23.03
AUDJPY-02.18200 0 8.3
AUDUSD-03.03110 6.6 32.3
CHFJPY+02.092792.8 (error) 32.7 50.7
EURCHF+03.011393.6 (error) 3.8 28.3
EURGBP+02.17210 0.7 1.7
EURJPY+01.045365.9 (probably an error) 51.5 55.7
EURUSD+02.15236.5 4.3 2.7
GBPCHF-02.14240 0 0.9
GBPJPY-02.21190 0.05 1.9
GBPUSD-03.03111.5 16.6 ?
NZDUSD-02.02320 9.9 49
USDCAD- trend change01.313469.05 81.8 ?
USDCHF-02.14240.65 0 1.2
USDJPY-02.17210.7 4.6 5.3
 
Vizard:

ok

ps.so as not to clutter up the thread I've cleaned up my posts...

good luck and profits to all...

no need to "clean" anything, it's a workflow
 

So far it's not working steadily on the 10 day forecast, too much error :o(

NZDUSD looks like it is about to change trend.

 
Farnsworth:

NZDUSD looks like it is about to change trend.

If you look from the 4h TF, it is already trending up, but if you look from the dailies without correction, there is no reason to consider it a trend. If we have broken through the daily trend line, the return without the low correction will be considered as a trend. Otherwise I would consider an up move as a 4-hour trend.
 
ULAD:
If we look at it from the 4hour timeframe, the trend is up, and if we look at it from the daily without correction we have no reason to consider it a trend. If we break through the daily trend line, the return without low update and upwards will be considered as a trend. Otherwise I would consider an up move as a 4-hour trend.

I am looking from the scale of daley. The estimation principle is a bit different, but let's see, there seems to be a possibility of a trend change and for a long time.

As always a good idea comes "after the fact" - why not to combine the forecast for 5 and 10 days?

 
Farnsworth:

I'm looking at the scale of the Daily.

We have decided on that.

The valuation principle is a bit different, but let's see, there seems to be a possibility of a trend change and for a long time.

We have not decided on that:)) Everyone has his own opinion about the formulation of the trend.

As always, a good idea comes "after the fact" - why not to combine the forecast for 5 and 10 days?

I am not thoroughly acquainted with your technology. Maybe you can open the door?

The longer the period, the greater the tolerance.

 

This is the distribution of demand for observable tools today.

1...192021222324252627282930313233...50
New comment