WinHunter trials and triumphs, jump in we need help!!

Brad

Member
No ... I can't get the prog to run, will have to wait for thornc I think and try again tomorrow ... I'm prob doing something silly, like misplacing a character ...
 

Brad

Member
Combo,

I did not email the file cuz you CAN use .txt file as Ion pointed out.
What threw me was the error message I kept getting when initially trying .txt ... ergo the .dat file ... apparently we're to ignore the msg, doesn't affect things.

You were right about not needing to create the output file as the prog will make one. Just a note, use a different name each time ... ie: output1, output2 etc ... otherwise I suspect the results will accumulate in the same file, don't think it gets cleared and overwritten --> this to be confirmed (or you can try).

The output numbers display scrambled ... they get 'ordered' when I import them into CoverMaster.

Tried last 40 of BC49, that gave me 24 lines ... I'll post them later so we can compare if you'd like ... not playing today, just observing :cool:.
 

Brad

Member
Perkisize Proccessor

Combo,

I'm getting conflicting results when I compare WH's to my 'by hand' Perkisized #s. Have you checked this ?

eg: for draw #1150 (last draw in hist file is #1149):

by hand - 3.11.12.16.21.24.25.35.38.39.40.44 -> 12#s
WH - 3.11.12.16.20. 21.24.25.33. 35.38.39.40.44 -> 14#s

Seems like WH is eliminating one draw from group D (#1137 in this example) when it should be skipping it.

I've looked at my previous results for last 10 draws and they don't match either. Will test #1139 next.

You're prob at work but if you get a chance I'd appreciate your comparison later.

PS. Managed to reproduce your results exactly, from the settings you posted earlier, we can talk about that too if you'd like.
 
WORK IS THE RIGHT PLACE

Hey Brado

1st the numbers I had posted for 6/49 are Double play for Group.
in an hour or so I will buy by own AW 9--> $4.50.
I am putting more concentration on COLD this time...

The comparison you're talkina bout , is it for upcoming 1151 or for the past 1150 ?

Also the settings that you are talking about, u mean the long post that I did few days back ?
 
DON'T FORGET THE 20 LINES

Yo Brado

Don't forget the the different output lines for various progs like LA, Lotto 97 and Markov.
So far all failed on Feb 1..
My opinion these numbers should be there for life and see which one is gonna match multiple 4's or even 5/6 once and of course JACKPOT..
If all fail to get 4/6 quite often then it's useless..

On the other hand I wanna get EXCEL to the put the history in ascending order and see the famous pairs,triplets that haven't showed up yet and based on OON too..Then I can try to predict the upcoming ones.
But for that we would have to play at least 40 lines each draw!!
 

Brad

Member
Combo,

I'll keep tracking Markov and I think I'm finally getting a better grip on WH too, (don't crash as much ... just go into a tailspin but I can usually recover now :D ) ... matter of time tho as I've got a lot on my plate lately.
 

winhunter

Member
Re: Perkisize Proccessor

Brad said:
I'm getting conflicting results when I compare WH's to my 'by hand' Perkisized #s. Have you checked this ?

Brad,

Email your Stack & history to me via LT. I want to be working on the same page as you guys. Both of you may be onto something here, and I dont mean a bug.


BTW, I looked at the code for 0.14, it will take a while to convert it over to the new version. Im gonna focus on getting the BETA version of WINHunter out now, but I did put a Subproject Task for the conversion of the old design into the new WINHunter.



Andrew
 

Brad

Member
Andrew,

give me a few min ... don't use the BC49 file from Free Lotto Stuff btw, I'll explain later.

Ok, files are on their way.

The file in Free Lotto Stuff doesn't always load ... don't know why, gave up on trying to fix it ... the one I emailed to you has worked fine so far, I asked LT to switch.

Cheers
 
Last edited:
PERKI UNIMPORTANT

Yo Brado,

I don't think Perki stuff is important anymore.I think it was a flash in the pan thing....
Yesterday was the perfect example of our instincts plus the help of Winhunter.
I managed to predict 1,14 and WH managed to get 9,10,11,32 !!
It's just that who would have thought these consecutives would show up !!!
I think Markov, others should go back to sleep...It will be a race between L.A. and WH :agree2:
 

Brad

Member
Hee hee ... a nonbeliever, eh? :D Me thinks there's more to it, don't think we've even scratched the surface of this thing yet ... call it a gut feeling .... just wish I had a faster puter.
 

winhunter

Member
Perki Settings

Hey,

Those settings I used (0,0,0,4,10) are basically the same as using the <PreviousDwgs> Processor, using the <Simple> scoring method with the Previous Drawings value set to 10, and the Parent Filter Skip value set to 5 (origional was skip 1, and the Perki Group D was 4, which gave a total of 5 skipped dwgs.)

The PreviousDwgs Processor may prove slightly faster if you want to simply test/optimize that. Here is how I would test it....


With a single Filter/Processor (PreviousDwgs), Select the Filter and Click <WINHunter> in the popup Menu (you already know to do this, but other users may want to follow along.)

Check BOTH items in the optimizer Setup WIndow that appears. Then double click Both items (Filter AND PReviousDwgs) so that they expand. Then Check the following Items:

Filter - Drawings to Bypass
PreviousDwgs - Previous Drawings

Then double click the Previous Drawings item, and then change the default MIN/MAX values to 0/20? (A nice range of Previous Drawings to test against.)

Then click the <Run WIN Hunter> Button.

NOTE: The mathematics involved in the PreviousDwgs processor are almost non-existent after the initial (most recent) DRAW has been predicted at the start of each test pass. You see, WINHunter is optimized to calculate the Stats on the first Test in a single Test loop (entire history test.) Then it Shifts the data, only dropping and adding draws as needed, and calculating for those few draws, instead of the entire group all over again.

Brad said:
Hee hee ... a nonbeliever, eh? :D Me thinks there's more to it, don't think we've even scratched the surface of this thing yet ... call it a gut feeling .... just wish I had a faster puter.

I think you are right, we are getting close. Now IS the time for the Distribution Server! I am nearing a point where I can release another installation. Soon, very soon....




Andrew
 
Last edited:

Brad

Member
Perkisize Proccessor

Hey guys,

the new Perki proccessor is correct now, able to match my data.

Get diff results, as expected, when re-running 'homework'. The best to date is setting 0,0,0,3,10 , pass 44: JP- 36 avg picks, 176x 6of6 hits. Paused at 0,1,6,6,1.

When I 'Perked' all draws from day one (excluding initial 16 draws in '92) I got these hits: 0.-0, 1.-1, 2.-19, 3.-109, 4.-292, 5.-472, 6.-242 in 1135 draws (using default settings 3,3,4,3,3 of course).

This means that ~21% of all the draws hit for 6of6 after Perked #s were eliminated, vs ~45% theoretical ...

No as good as anticipated but if this turns out to be the best setting for using the Perkisize Proccessor alone, 1 in 5 draws will be successfully reduced (the reduction could be from ~ 7 to 15 #s).

It would be interesting to check with CAN6/49 ...
 

winhunter

Member
Brad...

For your optimize setup, do this....

Set your Min/Max's for your Perkisize Processor to 0/10. Then set your Use Qty to 10, and set your Skip Qty to whatever you want. Then let it rip! You might find that 0/0/0/3/10 isnt the best after all. Be sure to minimize the windows and turn off verbose on the Backtest window.

Also, go ahead and set the base settings for your perki to 0/0/0/3/9 to start with. Then you will see what comes out better than your best so far.



Andrew
 
SUPPORTS MY THEORY

It's indeed a flash in the pan..no need to waste time on it..
Remember the 4/8 scenario had nothing to do with Perki.
But hey thanks for the hard work :agree2:
 

winhunter

Member
BC49

This is one of the History files I have, and it starts at Aug 7, 2002. I am going to run an Optimize on the Perki over the weekend. Let's see what kinda numbers this thing comes up with. Also, I started with a skip of 0, use 50.


I myself want to see what better settings can be found, and if anything is any better....


Andrew
 

winhunter

Member
Not yet

No, I didnt get anything from LT yet. Maybe by Monday or something. IM not majorly concerned over it right now, Im still trying to test things out. There will come a time though when I will get down serious about optimizing some Filters! :dancer:



Andrew
 

Brad

Member
Just some observations

Time to dust off the old thread now that Combo's got good predictions in two separate lottos same day !

The new 4.0d version is running well now, no bugs to report so far, I really like the Tier display . I still get the occasional Cycler error but know how to deal with it so it doesn't present a big problem.

AI has produced some rather complex stacks compared to what I was getting with the other versions. Don't know this for certain but it may have something to do with the way I applied it in the last run.
In the past once I stopped the AI I put the stacks she made in a separate folder for analysing at some convenient time. However, in the last run I let her keep and dump the stacks ... just kept tabs on her occasionally. A couple of times I noticed 'no apparent action' on her part for lenghty periods but just as I was about to shut her down she resumed operations so I left her alone. Even when I had to stop her I left the stacks alone and then restarted ... think that's the key.

AI's previous stacks seemed rather simple usually containing only one processor, it was rare to see 2 in the same stack. This last run looks very different, for example one stack had 4 groups with 6 filters and 8 processors!! Maybe she's getting smarter ... or just appreciates being left to her own devices? ;)
 

Sidebar

Top