Using Improvement Ratio

[Sooz]

Seems to me, one should consider 3 things - the cost of a ticket, the prize, and the odds of winning that prize.

Ontario49 which is a 6/49 lottery, pays $50 for a 4-hit. Tickets cost only 50 cents. The published odds for winning a 4-hit is one in 1033. Thus (hypothetically) if you buy about 1000 tickets, the cost is $500 and you're likely to win once only $50. Not a great bet.

BUT - if the GAT engine can find "6 numbers requested out of 49", where the improvement ratio for a 4-hit is greater than 10, then after buying 1000 tickerts, you're likely to win $50 ten times - which means you'll break even. Please let me know if at this point my reasoning is faulty - because I'm not accustomed to thinking in probabilities. Am I understanding the "improvement ratio" correctly?

In fact after running the GAT engine for 24 hours, the improvement ratio for a 4-hit went all the way to 27! (I could have started recording the table-numbers as soon as the imp ratio for that 4-hit column got to 11%.)
If my understanding is correct, then those 4-hit wins should (eventually after buying enough of them) more than cover the cost. ie you could truly play for free.

It would be ideal if the same GAT-number appeared under the 4 hit column AND under the 6-hit column! That did not happen.

.....Sooz

[lottoarchitect]

Well Sooz, your reasoning about improvement ratio is correct. 27 times better improvement is, well exactly that, 27 times better than classic probability of that event. Therefore, if your event is 1 in 1033, 27 times better makes this 1 in 1033/27. This assumes the GAT table will maintain this hitting cycle it has achieved over the tested draws and the reason we seek testing performance over a good test range so to rule out as much as possible this. Always look for regularity in hits. An ideal test for this is to have a big history and let the engine test against 1033 past draws. Based on classic probability you should get just 1 4-hit. You'll get a better idea of what to expect from GAT if you utilize such a big test range because it will also involve more of the cold cycles of each GAT over that test period.

It would be ideal if the same GAT-number appeared under the 4 hit column AND under the 6-hit column! That did not happen.

That doesn't mean it will not happen sometime in the future. All GATs have the potential to provide the winning combination. This shouldn't be your primary target however when dealing only with 6 numbers since this is rather difficult to achieve and being repeatable. When more numbers are involved, the story changes.

cheers
lottoarchitect

[Sooz]

I see. Thank you LottoArchitect. Well unfortunately I only have the past 220 draws for now. I’ve just emailed the Ontario Lottery Gaming Corporation to request all the past draws. We’ll see if they oblige.

So for now I’ll make “Total Real Draws to be tested for Statistical Analysis” 220. And then should “Total Statistical Data to be Used for Analysis” also be 220 for this method or can that remain a smaller number? Actually, I’m not clear on the difference between those 2 parameters?

Then to save the computer some time and energy, I’ll make min/max requested numbers 6-6. And hits category range 4-4. For this method, does that make sense?

When you say to look for regularity, do you mean, when looking at the hits graph, that the larger hits (in this case the 4-hits) should be more evenly spaced out along the graph, rather than clumped together?

It's interesting that this will not work for a 5-hit because the decrease in odds does not match the increase in winnings. The published odds for a 5 hit are one in 55,492 which is about 55 times worse odds than the 4-hit odds. But the prize only goes up 10 times - ie to $500. (We better not point this out to the OLG - I still remember when the casinos changed the blackjack rules. )
....Sooz

[lottoarchitect]

So for now I’ll make “Total Real Draws to be tested for Statistical Analysis” 220. And then should “Total Statistical Data to be Used for Analysis” also be 220 for this method or can that remain a smaller number? Actually, I’m not clear on the difference between those 2 parameters?

Total tested draws declares how many past draws will be used for the statistics generation. Start. data declares how many past draws before each tested draw of those "total tested" will be used internally for the prediction of that draw. You need at least "total tested" + 2 * Stat.Data history draws for GAT to work.

Then to save the computer some time and energy, I’ll make min/max requested numbers 6-6. And hits category range 4-4. For this method, does that make sense?

Perfect sense. All the computation speed will be concentrated at that req.numbers/hit category.