When you backtest systems..

Reply #25

KyMystikal — Fri, 06 Jul 2007 05:06:58 GMT

there's the 77's in TN 7/05 - 977. double 0's are next

Reply #24

Thoth — Fri, 06 Jul 2007 04:46:31 GMT

I usually backtest a system or strategy throughout the entire draw histories of every state that I have (I'm talking Pick-3 here). Normally, you can calculate the probability of a system hitting before you even start the backtest. The success rate of every system normaly performs very close to its true overall probability over the long term. Finding a system that way over performs its probability in every (or most) states would indicate a true odds-breaking system.....so far, I've found non... [ More ]

Reply #23

LckyLary — Thu, 05 Jul 2007 14:31:06 GMT

As long as the game is relatively the same. It doesn't make sense for example for me to backtest Mega to when there were 52 wb 52 bb, because some of the predictions would be unplayable. It also depends on how often a game is drawn; 6 months is nothing for MM or PB but can be more than enough for daily Pickits. There is no easy way to figure out when trends change.

Reply #22

KyMystikal — Wed, 04 Jul 2007 21:54:33 GMT

I hope someone listened to me about 99 in IL.

Reply #21

KyMystikal — Wed, 04 Jul 2007 05:48:15 GMT

I don't believe in a certain numbers triggering something to fall. I believe in statistics, for instance a double 4 should fall on the KY midday pick 3 this week and a double 3 will fall on KY evening. A double 7 should fall in IL this month and it could even be 777. I'd play 99 on midday in IL and 66 on evening. While I'm talking about IL on the evening 11 should be a HOT pair it has only been out twice this year on evening and it averages about 20 times a year. In TN 33 is due on midday and 77... [ More ]

Reply #20

guesser — Mon, 11 Jun 2007 08:26:29 GMT

But there is nothing to make a Quad hit in KY this year - or next year, and then having two hit in 2009. It would still average out to one every two years. Think outside the box.

Reply #19

KyMystikal — Mon, 11 Jun 2007 04:53:50 GMT

Also a quad is due in Ky this year. A quad falls every 2 years.

Reply #18

KyMystikal — Mon, 11 Jun 2007 04:52:44 GMT

I generally back test for the current year. If I see any type of pattern then I will go back over previous years. When I go back over years I see patterns like certain numbers falling during the same months and numbers fallin during the winter and so forth. Going on that note a double 1 is due in IL midday, double 8 is due IL evening, and double 9 is due Ky evening.

Reply #17

orangeman — Tue, 05 Jun 2007 03:04:56 GMT

I test in 10 or 15 different states over a period of time because if it works in one, it will work in all of them. After all, math doesn't lie. People who write get-rich-quick books sometimes do. . . .

Orangeman

Reply #16

guesser — Tue, 05 Jun 2007 02:14:13 GMT

I backtest as far back as the ballset changes, I don't go back 'forever' because it's irrelevent.

That gives me some data to work with.

Next up as you move forward form those dates is to see if you can find any trends, because they come and go.

The last PB Red ballset change was 3/8/06. Two numbers have not been hit since the change, and seven numbers have hit only once. Contrast that with nineteen numbers that have hit four times or more since then.

The last White ball ballset... [ More ]

Reply #15

gsc5 — Thu, 31 May 2007 15:41:56 GMT

It doesn't seem to make any difference how I back test a system, when I put money on it, the hits go sour.

Reply #14

budward — Thu, 24 May 2007 17:46:42 GMT

Hey lottolaughs, I just stumbled across this poll.

When working on P3 P-4 I usually backtest one month and according to the results I either drop it or keep going back sometimes as far as one year.

When working with P-5 P-6 I usually start tracking on the day of my system start and test going forward tracking each new draw instead of backtesting. I'm not interested in past performance. You can only win in the present and/or future not the past.

Bud

Reply #13

paurths — Wed, 23 May 2007 06:40:36 GMT

I backtest against the entire drawhistory of each state. (or at least, the draws that i have in my database)

Then i will run the same test against the draws for an upcoming month.

cheers

Ricky

Reply #12

lottolaughs — Wed, 23 May 2007 02:06:36 GMT

Thanks for all your input everyone,appreciate it a lot. I'm enjoying the show over here,keep 'em coming

Reply #11

hypersoniq — Wed, 23 May 2007 02:01:23 GMT

I backtest with all available data...

a short term promising system might have 2 or 3 year gaps with no wins... I need to see the system vs. all of the history... Helps gauge when NOT to play.

For example, a sysem that consistently averages one hit a year could be played until won then dropped.

Reply #10

pacattack05 — Wed, 23 May 2007 00:16:02 GMT

It all depends on the situation. If I backtest a p-3 system for 5 consecutive days and win straight all 5 days, buying only one ticket a day, I'll know that's a really special winning system, and not just a fluke.......LOL

On the other hand...it very well could be a fluke. Those kinds of flukes are fine with me....

Actually...that's a good poll. Has anyone won straight on the p-3, 2 days consecutively?....LOL

Reply #9

Guru101 — Tue, 22 May 2007 23:10:22 GMT

Most of my methods are backtested about a month, but I got a few that go all the way back to when the matrix changed.

Reply #8

lottolaughs — Tue, 22 May 2007 21:39:39 GMT

Thanks for shedding some light!

Make the data fit the theory in other words, instead of looking at the data first and then developing the theory.

I like this way of thinking

Reply #7

lottolaughs — Tue, 22 May 2007 21:37:57 GMT

That would seem the smart way to do it. Why bother going on if nothing shows up? Good suggestion

Reply #6

Badger — Tue, 22 May 2007 20:57:45 GMT

Definitely agree. A system may very well work in one state, but not work in others. I've never figured out the why , but for some reason -- while one would think that a Pick 3 game is the same in all states, the games don't seem to act the same in all states. It's one of those little tricky things that make something look like what it isn't.....similar to the fact that you can take 20 combinations at random, call them predictions for all states and you are bound to hit something somewhere... [ More ]

Reply #5

RJOh — Tue, 22 May 2007 19:42:47 GMT

The only thing back testing has ever proved to me is almost any theory will work once and if its parameters are wide enough it will probably work again but I can't afford to cover that many ifs until it does. Sometime when you back test, you can cheat yourself because you know the results you're looking for. I prefer to forward test, then I'm sure I'm not testing for known results, beside it's extra work rewriting the data file to look as it did the date of each back test.

Since I like... [ More ]

Reply #4

four4me — Tue, 22 May 2007 18:56:24 GMT

I have back tested a few systems and usually only about 3 weeks back. A few times there were promising results. However since they change the ball sets and machines for every drawing i don't know if back testing proves anything.

There are to many variables to consider. However if a system is any good then there should be no reason to do extensive back testing. I can admit i don't think there are any systems that can produce enough winning picks. Supposed you back tested for say a month and t... [ More ]

Reply #3

derek7 — Tue, 22 May 2007 18:29:42 GMT

I don't backtest at all.

/Derek

Reply #2

Coin Toss — Tue, 22 May 2007 18:16:57 GMT

I discarded the vote on this one.

I think one of the things that make iteasy for system sellers to sell systems, whether they are worth anything or not, is that with the variety of lotteries available, the number of states with Pick 3 and 4 games and more, a person could come up with a theory and then keep searching and searching until they found a set of results that prove their theory.

This is not unique to lottery systems, you can find it for blackjack, dice, roulette, horse racing,

Reply #1

stavros — Tue, 22 May 2007 18:05:43 GMT

I generally backtest 30 days. If nothing has hit within that perioid, I don't go any further. If there are hits during that 30 days, then I go back another month or so to see if it's consistent.

When you backtest systems..

lottolaughs — Tue, 22 May 2007 17:40:01 GMT

Is there such a thing as too much backtesting?? What's your limit when you feel whether a system is working or not???

Thanks all