dev build 110401

[stan]

what's new in this dev build:
- prediction results are showing in winning numbers charts
- prediction resutls are showing in history summary and diff charts
- it's possible to click any draw in a history chart to see its statistical properties and ticket panel layout
- prediction results are showing as hints in 'my estimates' window
- it's possible to predict the next draw interval value in analyzer charts (click 'details' button on the toolbar first)

[djdez]

Is it available now?

[stan]

Is it available now?

it is

[djdez]

Where because i try to update from EL5 it says no updates available

[stan]

Where because i try to update from EL5 it says no updates available

this is an update of development build, please read here: viewtopic.php?f=18&t=1632

[djdez]

Thanks Stan...am on it

[Viking]

Hi Stan!

I see you have resolved my issue 125, thank you for that.

There is only one problem and that is that most of the time the Match Previous Draws filter and the Match Winning Numbers filter do not produce the same results. To me it seems that it is the Match Winning Numbers filter that is causing the error. The test is done for the last 5 draws and I have used just one value at a time.

I have attached an image showing which values that are causing a mismatch between these two filters. The values that are ticked are the ones that are causing the mismatch. The first is from my lottery and the second is from Expert Lotto 6/49. If you compare them you will see that the first four values are right for both.

[stan]

There is only one problem and that is that most of the time the Match Previous Draws filter and the Match Winning Numbers filter do not produce the same results. To me it seems that it is the Match Winning Numbers filter that is causing the error. The test is done for the last 5 draws and I have used just one value at a time.

it is because 'match previous draws' analyzer counts the *highest* match in the previous draws. while 'matching winning numbers' filter accepts a ticket if there is at least one draw matching the filtered ticket. so 'matching winning numbers' filter accepts more tickets than analyzer table filter with the same settings.
so if you insist on having a standalone 'match previous draws' filter, please reopen that issue, thanks

[ameriken]

Stan, when doing the pattern matching and the predictions are posted, the pink and blue arrows appear to indicate whether a value has increased or decreased. I've noticed that most of these are correct, however I've seen some levels where the arrows sometimes do not accurately represent the true movement. For example, I am looking at the prior draw sum being 814, and the predicted sum is 805, however in this level a blue upward arrow is in the box. This happens in one or two levels everytime I use the function. It doesn't affect anything else, however I figured you'd want to have it corrected.

Also, I've been doing some backtesting with the pattern matching, using a sampling of 2000 draws, with the default pattern size 2 to 50 and latest draws for backtest 50. When I get the predicted values and enter them into 'my estimates' and filter a full package, the filtered results are 0. This has happened a few times, so I wonder if I am not setting something properly? Suggestions?

I'll keep working on it.

[stan]

Stan, when doing the pattern matching and the predictions are posted, the pink and blue arrows appear to indicate whether a value has increased or decreased. I've noticed that most of these are correct, however I've seen some levels where the arrows sometimes do not accurately represent the true movement. For example, I am looking at the prior draw sum being 814, and the predicted sum is 805, however in this level a blue upward arrow is in the box. This happens in one or two levels everytime I use the function. It doesn't affect anything else, however I figured you'd want to have it corrected.

email me a screenshot next time you get such results pls.
you can also check the application log (menu 'view - application log') to see more details about the prediction process. email me that log as well pls.

Also, I've been doing some backtesting with the pattern matching, using a sampling of 2000 draws, with the default pattern size 2 to 50 and latest draws for backtest 50. When I get the predicted values and enter them into 'my estimates' and filter a full package, the filtered results are 0. This has happened a few times, so I wonder if I am not setting something properly? Suggestions?

that's expected. the predictor just looks at a series of numbers, it has no way of knowing what those numbers represent or how they relate to other series of numbers (i.e. levels). so you'll be getting conficting predicitons more often than not.
the pattern matching predictor is no silver bullet:) it just tells you what happened in the past when there had been a similar pattern...

[ameriken]

Stan, when doing the pattern matching and the predictions are posted, the pink and blue arrows appear to indicate whether a value has increased or decreased. I've noticed that most of these are correct, however I've seen some levels where the arrows sometimes do not accurately represent the true movement. For example, I am looking at the prior draw sum being 814, and the predicted sum is 805, however in this level a blue upward arrow is in the box. This happens in one or two levels everytime I use the function. It doesn't affect anything else, however I figured you'd want to have it corrected.

email me a screenshot next time you get such results pls.
you can also check the application log (menu 'view - application log') to see more details about the prediction process. email me that log as well pls.

Also, I've been doing some backtesting with the pattern matching, using a sampling of 2000 draws, with the default pattern size 2 to 50 and latest draws for backtest 50. When I get the predicted values and enter them into 'my estimates' and filter a full package, the filtered results are 0. This has happened a few times, so I wonder if I am not setting something properly? Suggestions?

that's expected. the predictor just looks at a series of numbers, it has no way of knowing what those numbers represent or how they relate to other series of numbers (i.e. levels). so you'll be getting conficting predicitons more often than not.
the pattern matching predictor is no silver bullet:) it just tells you what happened in the past when there had been a similar pattern...

OK, thanks, I just wasn't sure if it was something I was doing or not (on the predictions). I've been backtesting last nights draw, going back and forth with different ways and number of draws, and finally went to 100 draws that gave me some predictions that gave me a package. Out of t 68 tickets, there were 2 tickets that matched 3/5. That's promising.

Next time I get it I'll send you a screen shot on the other deal, Ken

[ameriken]

Just sent you an email with the screenshot.

BTW, is there a recommended # of draws when doing the prediction?

The more the better?

I know more can take a long time to calculate, but I have a cash 5 game that goes back 4000 draws. Will that make the predictions more accurate, or is it better to stay with the more recent draws? Any ideas about that?

[stan]

Just sent you an email with the screenshot.

were those predictions made with 'match differences' option?

BTW, is there a recommended # of draws when doing the prediction?

it depends on what you believe in:)
if you think that the 'history repeats itself' and that there are no factors that affect the series of numbers then you should scan the whole winning numbers database when making predictions (except when predicting wn history data as those are skewed until enough draws are accumulated).
if you believe that the drawing machine, the actual balls used for drawing or any other physical factors are important then you should check only limited section of your wn database. e.g. since when a new drawing machine has been introduced. and you should perhaps let the predictor backtest only 3 or 5 latest draws to catch the very latest patterns.

besides that - experiment:)

[ameriken]

Just sent you an email with the screenshot.

were those predictions made with 'match differences' option?

BTW, is there a recommended # of draws when doing the prediction?

it depends on what you believe in:)
if you think that the 'history repeats itself' and that there are no factors that affect the series of numbers then you should scan the whole winning numbers database when making predictions (except when predicting wn history data as those are skewed until enough draws are accumulated).
if you believe that the drawing machine, the actual balls used for drawing or any other physical factors are important then you should check only limited section of your wn database. e.g. since when a new drawing machine has been introduced. and you should perhaps let the predictor backtest only 3 or 5 latest draws to catch the very latest patterns.

besides that - experiment:)

Been using the default 'absolute value', wasnt sure what the difference was between the two.

[stan]

Been using the default 'absolute value', wasnt sure what the difference was between the two.

ok, thanks for the info. the inconsistency between the predicted trend and the actual predicted value comes from the prediction algorithm. when option 'match differences' is on then the prediction algorithm behaves as in version 4.x. you can read more on the predictions in the documentation for expert lotto 4.x
'absolute values' is a slight variation of that algorithm. i'll describe the details in the documentation when version 5.1 is released officially.

btw, i suggest using the predicted trends only when making decisions for the next draw. the predicted value should be just an indicator how big the increase/decrease is to be expected in the next draw.