Another 4-hit!

[Sooz]

Yes, the last 2 draws have completely paid for themselves!

I'm now using total real draws = 160, and total stat data = 30. Unfortunately, this change has reduced the best improvement ratio for a 4-hit, when requesting 6 numbers out of 49, to 16.9% from the previous 27%. (And also there are fewer tables.) Nevertheless all that is needed for this lottery is an improvement ratio for a 4-hit of about 11% to have the 4-hit winners allow me to play for free.

Here's my method. I'll type it in some detail, hoping for some feedback and possible discussion.

I examine all the tables which have an imp ratio, for the 4-hit column, of over 11% for requesting 6 out of 49. There were 20 tables. Then I put those tables into the GAT list and browse through the list looking at 12 numbers requested out of 49. This is because I plan to wheel 12. (Last time I wheeled 15, this time only 12.) Of course as the requested numbers go up, the imp ratio goes down so I'll stay with 12.

For each table in the list, looking at 12 numbers requested, I want (a) few as possible zero-hits, (b) high imp ratio again for the 4-hits column (altho of course it's not as high as for 6 requested) (c) look at the graph of tested draws - are the larger hits more evenly distributed? Is a larger hit due soon?. (d) are there also some 5-hits and 6-hits? There were some but not with a great imp ratio, so I ignored those.

So I chose one of those 20 tables of 12 numbers, wheeled it with WG using filters, and requesting WG to give me 4 if 4, and of course additional coverages of 3 if 4.

The table I chose to wheel was the only one of the 20 I examined which had 4-hits!

I also played the other 19 tables' first 6 predicted numbers without wheeling them - ie 19 more tickets, each with 6 predicted numbers. No winners from those yet, but that's where the imp ratio is over 11% so I expect there will be some 4-hits there within the next 5 draws or so.

.....Sooz

[pusha]

Hi Sooz,

congrats to your hit. Is there a reason for total real draws = 160, and total stat data = 30? Which WG Filters do you use? Why don't you directly ask GAT for 12 numbers, instead only 6 and than scale them up to 12 numbers when found a appropriate GAT.


Cheers,

pusha.

[Sooz]

Hi Pusha
I'm so glad that someone is interested.

I wanted to make use of all the draws available and only had access to 220 past draws. I started with stat data =30 arbitrarily. Someone on this blog said that it doesn't make significant difference over 50. So as I add more available past draws, eventually I'll use stat data=50. Then, since maximum total real draws depends on how big the chosen stat data is, the maximum total real draws (based on stat data of 30) was 160.

My filters are fairly lenient, and I use only 3:
sums 82-219,
odds up to 5 - ie 6 numbers mustn't be all odds nor all evens but it's ok to have just one odd or one even.
position one must have numbers between 1 and 21. Position six must have numbers between 25-49. I don't worry about the intervening positions.

Reason for asking GAT for 6 numbers first - I wanted an improvement ratio of at least 10 and the imp ratio didn't get that high for more requested numbers. For 12 predicted numbers the best it could do was an improvement ratio of about 3. Why do I want an improvement ratio of at least 10? I consider 3 things - the probability of a hit, the payout for that hit, and the cost of a ticket. For this lottery (Ontario49) a ticket costs 50 cents, and the payout for a 4-hit is $50. Thus the payout is 100 times the cost of one ticket. The ordinary calculated probabilty for a 4-hit is 1 in 1033. With an improvement ratio of 10, the odds become about 1 in 100. Thus one 4-hit winner should pay for 100 tickets. With an imp ratio =10 or better, I would expect there to be 1 4-hit winner in every 100 tables.

Hopefully this makes sense?
Of course I want the big winner, but my aim is to play for free. Also the preference is that the GAT engine should find more than 20 tables, for each draw, with my required imp ratio but that's all it found.
.....Sooz

[pusha]

Hi Sooz,

it all make sense to me. Thanks a lot for sharing your approach. In the german 6/49 i can't get that good imp ratio even with over 1.600 history draws. What GAT count have you run to reach that imp? I guess ur GAT Options are 6/0 numbers and 4/4 hit right?


Cheers,

pusha.

[Sooz]

Hi Pusha

Few weeks ago when I used total real draws = 100, and total stat data = 20, the imp ratios were actually higher. Perhaps the more draws one uses, the more random it gets. I don't know. If so, then that might actually be an argument against using a lot of history. Nevertheless, I plan to continue with what worked.

For this successful method, the first time I used min/max 6-12, and hits category range 4-6. For the next time, I was just interested in the "GATS to retain" so I just set them both to 1-1 so that the engine would run faster. I think if you enter a zero, that's too much work for computer.

But in spite of those settings, I was pleasantly surprised that, although the panorama tab didn't show much, the GAT table tab still has all the information we want when we just ask. But remember, before starting, to enter the interesting previous table numbers into "GATS to retain". Then later at the GAT table tab, just tell it what table numbers you want to examine and tell it how many numbers you request out of 49.

Well the highest good table number my computer reached was over 4.5 million. For 6 numbers requested, the 4-hits had an imp ratio of 16.9. I had just left it running in the background for a couple of days. But the one I ended up choosing to wheel was only table number 344661. This one had, under 4-hits, an imp ratio of only 14. I chose it because there were significantly less zer0-hits and because the hits seemed more evenly distributed.

Have you been using all 1600 draws in your history? Do you get better imp ratios when you use LESS draws??

.....Sooz

[lottoarchitect]

Well, my two cents here to this discussion. First congrats Sooz for this new win! I'm delighted to see users find ways to use this system beyond the initial guidelines given. Now, with more tested draws or more req. numbers, the imp. ratio goes down. This is normal really. The more tested draws, the more cold cycles will be included as well which tend to lower the overall imp. ratio. Similarly, the more req. numbers, the higher the natural probability to contain more correct numbers therefore the imp. ratio again goes down.

[pusha]

Hi Folks,

well i've reached an imp ratio of 60+ with 50 total real, 10 total statistical draw with 6/12 req. number and 4/6 for a 4-hit number pool of 6, if i scale it up to 12 numbers the imp rate goes down to 3 .

If you mean by using all 1600 draws, yes i have them all in my history draws. What i've been testing is was the x-future indicator. E.g. look ahead 5 future, i've tried to get the number for e.g. draw 10 by deleting past 4, so the last draw in the history would be draw 5.
After getting the numbers and observing the indicators i take a look at the future hit prediction, by the hit amount i sum it and divide it by the x-time look ahead and i get a factor. What odd is that if the factor goes over 2,5 a 3 hit is guaranteed. BUT only if in the 5 future draw 4 haven't a 3 or higher hit...?! Let say, of that 5 future look there are 2 with 1 hit and 2 with zero, it strangely matches with the future quiet well, sometimes it differs by -1 hit...

I'll have to examine this further in more detail, but my backtesting was quiet good for now. From my 10 test i had 8 times the expected result of what GAT is predicting with some little -1/0/+1 as discussed in my other thread ("GAT trend"). What we get by that method is imo an average hit to expect if it hasn't already happened.


Cheers,

pusha.

[Sooz]

Hi Pusha,
What do you mean by the "x-future indicator"? I did go to your other blog entry and didn't understand that one either. (Sorry if I'm a little dense here ).

Do you mean the blue line on the graph? My understanding is that this blue line is simply a moving average of the past 5 hits. Sometimes it's easier to see a trend in a moving average. The moving average seems to smooth out the sharp peaks and troughs so a trend can be seen more easily.

But your description of using it as a predictor sounds more involved than that. What am I missing?

.....Sooz

[pusha]

Hi Sooz,

your understanding is fully correct. What i mean with x-future indicator are the red dots how they would hit and what their average hits are.


Cheers,

pusha.