Reliability of "History differences ranges"

Reliability of "History differences ranges"

Postby smokers » Wed Jul 19, 2017 12:24 pm

Hello guys,

I am playing with various settings on predictions, especially on the "History differences ranges". How reliable are the results?

I have tried with various ranges and from i get confusing results from range to range, conflicting with the previous results.

Regards
smokers
 
Posts: 44
Joined: Tue Apr 11, 2017 8:21 am

Re: Reliability of "History differences ranges"

Postby Meisterkeks » Thu Jul 20, 2017 5:46 am

Hi smokers,

I use this filter for a long time for the prognosis and always have correct results (already several times synonymous with EXCEL tested).

Image
Thank you

Best Regards!

Meisterkeks
Meisterkeks
 
Posts: 221
Joined: Fri Aug 19, 2016 7:42 am
Location: Germany

Re: Reliability of "History differences ranges"

Postby smokers » Fri Jul 21, 2017 9:14 am

Hi Meisterkeks,

i understand the filtering part (did not thought choosing last 2 draws though). I will make some research on games i'm playing. Now, regarding ranges I wanted bring more tools in my estimates so that I choose smaller ranges, thus using less estimate levels, but with the right results. I have done several range selections (increasing and decreasing the ranges), however as shown in the screenshot below things aren't straight, although the only thing changing on each range prediction is only the range change. As you can see from range to range ... some prediction are contradictory.
Image
smokers
 
Posts: 44
Joined: Tue Apr 11, 2017 8:21 am

Re: Reliability of "History differences ranges"

Postby tdnl46w » Tue Jul 25, 2017 9:45 pm

If this is using the predictions feature, can you give an example or steps from within the EL program of conflicting results. The issue might be from how the program displays the results from the prediction feature.
tdnl46w
 
Posts: 357
Joined: Thu Jan 29, 2015 10:56 am
Location: USA

Re: Reliability of "History differences ranges"

Postby smokers » Thu Jul 27, 2017 6:20 am

The scope of the above table is to help me decide wether a certain level value will be in a certain interval or not, thus deciding the min and max of history differences/values.
I have setup ranges from -/+ 5 to -/+ 45. This is the only parameter that is changing from a prediction command. After each range setup change the prediction is performed again so I have proper prediction process. The logic would be to narrow the field on which the predicted values would be. Also each prediction has also a backtrack percent as setup in the prediction options, thus giving some sort of confidence values.
Therefore, if starting from a wide range , -/+ 45 then switching to narrow ranges, I can expect at some point to find some limits. A simple scenario/example would be: if a value will be in the range of -/+ 45 but not in range -/+ 20, I expect that when having the predictions on -/+ 20 it will show me > + 20 OR < -20, and also to exist linearity in predictions. For the example below I will know that my ranges will be above + 20 or below -20. Having these ranges of values, I can go then to segmentation table and find the segmentation combos that will give me values in the appropriate expected values.

However as seen in the attachment picture, row 5 for example: the prediction conflict seems to be obvious.
smokers
 
Posts: 44
Joined: Tue Apr 11, 2017 8:21 am

Re: Reliability of "History differences ranges"

Postby tdnl46w » Thu Jul 27, 2017 12:37 pm

The prediction feature finds the pattern in past winning draws and then displays the value or property found in the following draw after the pattern for the prediction. This might be a reason your getting conflicting results. So in the spreadsheet image above, those should be the values/properties that were found right after the patterns for the setups/ranges used.
tdnl46w
 
Posts: 357
Joined: Thu Jan 29, 2015 10:56 am
Location: USA

Re: Reliability of "History differences ranges"

Postby smokers » Fri Jul 28, 2017 11:53 am

you're most probably right. Anyway, predictions seem to be useful. I have tried in my game and out of 60 estimations based on ranges (and some guessed tweaks) the were only 7 missed ranges, even on some extreme ones. So they seem to help... on some manner.
smokers
 
Posts: 44
Joined: Tue Apr 11, 2017 8:21 am

Re: Reliability of "History differences ranges"

Postby tdnl46w » Thu Aug 03, 2017 5:07 am

My last reply was mainly how the prediction displays a value, but there is a lot more important information in the documentation. View->Application Log output tab and the prediction details there is how I learned exactly what the prediction feature does.

I find the prediction trends(red/blue indicator triangles) to be the most accurate. From there, I use the Trend Predictor filter on a package to see how many tickets can be reduced. There is also History->Predictions... which links some prediction info per level to the My Estimates tab that may help to reduce estimates.
tdnl46w
 
Posts: 357
Joined: Thu Jan 29, 2015 10:56 am
Location: USA


Return to Tips + Tricks

Who is online

Users browsing this forum: No registered users and 3 guests

cron