Perhaps new software for Lotto 6/49 and others

[johnsmithx]

Hello everyone!

Recently I've started developing a program which may or may not be eventually useful for lottery enthusiasts. Since I don't really know much about the lotteries and various strategies and systems (I am a programmer and this idea came to my mind while developing an optimized matching algorithm), I would like to humbly ask you for your opinion and perhaps an advice about the usability of this program and its features.

The initial idea was this:
In my country there is a 6/49-based national lottery where, besides an ordinary 6-numbers ticket, one can take a system bet consisting of 7 up to 15 numbers. Sure that way you can cover more combinations, but is it really worth it? So I was thinking that maybe computing all possible combinations from 6 (let's support even standard 6-numbers bet too) to 15 numbers and then comparing them against the results of the previous draws it would be possible to find combinations which (had they been used) in the past have got the highest chances to match the actual drawn numbers. Then maybe using some sort of time analysis (how evenly a combination "scored" in the past draws) it could give some clue as whether given system combination might score in the future.
(For example, if I find a combination which in the last 30 years would have scored in 15% of all cases but looking at the dates it would appear that the majority of those 15% actually happened mainly during 15-25 years ago, then such a combination is probably not very great for the future bets. On the other hand a combination with maybe slightly lower % success but scoring more continuously from far past to the very present may be a better candidate.)

This idea may or may not be actually useful, you know surely much better than me, anyway from the programming perspective this was an interesting problem because it would require a lot of computing. Standard 6-numbers bet makes 13,983,816 possible combinations, that's nothing. 7-numbers bet goes to 85,900,584 combinations, 8-numbers 450,978,066 combinations, ..., and 15-numbers system bet can be actually selected out of 1,575,580,702,584 possible combinations. Now to make it work all those combinations would have to be computed and then compared with the lottery history data - which may be thousands draws. So each computed combination must be compared to all past draws, respective % chance must be calculated, the same thing must be done with all other millions and billions or even trillions of combinations, and at the end it should print out several best combinations. And all that in a reasonable time.

This is where my then-developed matching algorithm kicks in as it was supposed to provide some speed boost, so to test it in the reality I've implemented it in this context and built this program around it.

For the time being the performance is as follows:
* 900 millions of combinations per second are computed
* 90 millions /s are processed and delivered to the matching algorithm
* the speed of actual testing depends on how many history draws the computed combinations are to be compared with:

- 2.2 millions combinations /s when comparing against 100 lottery draws
- 212 thousands /s when comparing against 1200 lottery draws
- 50 thousands /s when comparing against 5300 lottery draws

The above example is the performance per a one CPU core (3,15 GHz in this case), thus using e.g. 4 cores of 4-core CPU the speed will result in 4 higher speed. The program can be divided into thousands pieces and can be run simultaneously and totally independently on as many cores, processors and even separate individual computers as needed. I am still not satisfied with the speed, though. Also, the time analysis is not implemented yet, that will probably drag the speed down even further.

Anyway, even in the current work-in-progress developing state, all possible 1,5 trillions of 15-numbers combinations can be computed, processed and compared against 5300 lottery draws (our lottery's data from last 35 years) in no longer than a single month, using 3 CPUs, each with 4 cores. That's not that bad, I guess.

Program parameters:

$ ./lotto
Lotto ver. 1.1
==============

This program will compute all possible k-combinations of 49 numbers (1 - 49), drops combinations containing any blacklisted number (if any specified in blfile), further drops combinations not containing all whitelisted numbers (if any specified in wlfile) and then from the remaining set takes each combination and compares it against all 6/49 lottery draws (specified in dtfile) calculating the percent rate of 4-matched numbers, 5-matched numbers and 6-matched numbers. At the end the program will show the TOP-3 combinations with the highest combined percentage.
In the alternate mode (pfreq is set to 999) the program will not calculate anything, just print out all computed (and all 'to be tested' - with regards to the blacklist and whitelist) combinations.
(C) 2011 johnsmithx (johnsmithx at seznam.cz)

Usage: lotto dtfile blfile wlfile k pfreq start stop [benchmark]
  or   lotto dtfile blfile wlfile k 999 start stop

Parameters:
   dtfile       csv file, at least 1 row, 6 values per row
   blfile       csv file, either empty or up to 49-k numbers on a single row
   wlfile       csv file, either empty or up to k numbers on a single row
   k            k-combinations (6 - 15)
   pfreq        progress report frequency in % (0.01 - 50)
   start        start computing at specified progress % (0 - 99.99)
   stop         stop computing at specified progress % (start+0.01 - 100)
   benchmark    period of seconds for initial benchmark (5 - 100), default 10

So as you can see the program is technically focused and doesn't have any eye candy useless features. It's made so that all parameters can be specified as the command arguments, the output can be redirected to a text file and the program can be run as a scheduled job in the background. Currently the platform is Linux x86_64 with assembly optimizations for modern core2 processors with sse2/ssse3 extensions.

Besides the actual history data provided in the form of a simple CSV file (comma separated values, just 6 numbers at each line, plain and clear) the program can work with the blacklist and whitelist files. In the blacklist you can specify numbers you DO NOT want to appear in the tested combinations. In the whitelist you can specify the exact opposite - the numbers you DO want to have in the tested combinations. Program will always compute all possible combinations (quick part) but only those in compliance with blacklist/whitelist are passed for testing (slow part).

You also specify how often should be shown the intermediate progress report, so you can see in what point the program run is at the moment, how long is it running and how long is expected to the end.

The very important feature is specifying the start and stop point, which allows to divide program run to smaller fragments and run then either one after another or all at once. Every running instance of the program will bind itself to a single core of a single CPU.

Example of output:

$ ./lotto lotto_data_5336_1977w1-2011w39.csv lotto_blacklist.csv lotto_whitelist.csv 7 20 0 100
Lotto ver. 1.1
==============

Input data statistics: datafile 5336 rows, blacklistfile 0 numbers, whitelistfile 0 numbers
Requested 7-combinations of 1 - 49: 85900584 combinations to compute, 85900584 combinations to test
   parameters: start 0.00% stop 100.00% pfreq 20.00% benchmark 10 s

START COMPUTING at 0.00%  (2011-11-20 17:07:48 +0100)
 benchmarking: computed combination no. 430000 (43000/s), tested combination no. 430000 (43000/s)
               elapsed: 10 s, remaining: 33 m 7 s  (2011-11-20 17:07:58 +0100)

Progress 20.00%, computed combination no. 17180117 (46685/s), tested combination no. 17180117 (46685/s)
elapsed: 6 m 8 s, remaining: 24 m 31 s  (2011-11-20 17:13:56 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

Progress 40.00%, computed combination no. 34360234 (46685/s), tested combination no. 34360234 (46685/s)
elapsed: 12 m 16 s, remaining: 18 m 23 s  (2011-11-20 17:20:04 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

Progress 60.00%, computed combination no. 51540351 (46685/s), tested combination no. 51540351 (46685/s)
elapsed: 18 m 24 s, remaining: 12 m 15 s  (2011-11-20 17:26:12 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

Progress 80.00%, computed combination no. 68720468 (46685/s), tested combination no. 68720468 (46685/s)
elapsed: 24 m 32 s, remaining: 6 m 7 s  (2011-11-20 17:32:20 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

COMPUTING COMPLETED at 100.00% in 30 m 39 s  (2011-11-20 17:38:27 +0100)
 85900584 combinations computed (46710/s), 85900584 combinations tested (46710/s)

Results
=======
1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

What I would like to ask you is this:
- What do you think about this idea and the program? Can it be of any use for lottery purposes at all or is it just non-practical nonsense?
- What other features could be useful to implement?

Thanks for your answers.

johnsmithx

[Johnny's Amusements]

You have clearly put a lot of time & effort into this but the reality is there can never be a system which can predict a totally random draw. It will be interesting also to hear other members views on this, but I would not be surprised to hear the word "spam" used!

[johnsmithx]

"Spam"? As an unsolicited advertisement? Did you actually read what I wrote?

I am not advertising anything, I am not selling anything. I am just a programmer (for the last 25 years) and during my ordinary programming I just happened to work on quite interestingly fast matching algorithm and it appeared to me that it may be used on processing historic lottery data against generated combinations, so I simply took a few days effort to prove or disprove that theory. That's all. That's what programmers do if programming is really their life - they program various (stupid) ideas. And why of all possible applications the lottery came to my mind despite I do not take part in lotteries myself? Well maybe because just now in my country one particular lottery got the all-time record jackpot, maybe because the company running this lottery was in the news for the last half a year because it was in quite emotional process of changing the owner. So that's maybe why.

Time to time, when I program something privately (not for business), purely out of my own interest, I find out that it might be maybe useful for some people, so I simply ask them what they think. Few months later I started programming some kernel stuff for the Android and got some interest from the people on the android forum in my country, you should ask them if they ever thought about "spam" as they got everything from me for free. Some time ago I programmed some quite useful forensic tool for statical analysis of PE (and other) files and gave it to the public on forensic focus forum. Some other time ago I programmed some analysis tools for financial markets, etc., etc.

The initial trigger impulse may be anything, a dream, an idea, whatever, but soon I usually forget why I actually started programming this or that and just keep programming because I simply enjoy it, can't help myself. Then when the problem is solved I either drop it or, if I think it may be useful for someone, I simply find and ask such people what they think about it. I didn't know where to ask about a tool potentially useful for lottery. I found several forums and this one seemed to have the most people participating in discussions (based on number of threads here), so I dared big time to post a thread here. Maybe it was a mistake, no big deal. If that is the case, I hereby ask the moderators/administrators to delete this thread as my intentions never have been to offend anyone and at the same time I have the same, not lower, right as anyone to not be offended by others.

Have a nice day.

[Johnny's Amusements]

"Spam"? As an unsolicited advertisement? Did you actually read what I wrote?

I am not advertising anything, I am not selling anything. I am just a programmer (for the last 25 years) and during my ordinary programming I just happened to work on quite interestingly fast matching algorithm and it appeared to me that it may be used on processing historic lottery data against generated combinations, so I simply took a few days effort to prove or disprove that theory. That's all. That's what programmers do if programming is really their life - they program various (stupid) ideas. And why of all possible applications the lottery came to my mind despite I do not take part in lotteries myself? Well maybe because just now in my country one particular lottery got the all-time record jackpot, maybe because the company running this lottery was in the news for the last half a year because it was in quite emotional process of changing the owner. So that's maybe why.

Time to time, when I program something privately (not for business), purely out of my own interest, I find out that it might be maybe useful for some people, so I simply ask them what they think. Few months later I started programming some kernel stuff for the Android and got some interest from the people on the android forum in my country, you should ask them if they ever thought about "spam" as they got everything from me for free. Some time ago I programmed some quite useful forensic tool for statical analysis of PE (and other) files and gave it to the public on forensic focus forum. Some other time ago I programmed some analysis tools for financial markets, etc., etc.

The initial trigger impulse may be anything, a dream, an idea, whatever, but soon I usually forget why I actually started programming this or that and just keep programming because I simply enjoy it, can't help myself. Then when the problem is solved I either drop it or, if I think it may be useful for someone, I simply find and ask such people what they think about it. I didn't know where to ask about a tool potentially useful for lottery. I found several forums and this one seemed to have the most people participating in discussions (based on number of threads here), so I dared big time to post a thread here. Maybe it was a mistake, no big deal. If that is the case, I hereby ask the moderators/administrators to delete this thread as my intentions never have been to offend anyone and at the same time I have the same, not lower, right as anyone to not be offended by others.

Have a nice day.

I myself did not suggest that this was spam. I just politely pointed out the fact to you that some other members of this forum MAY see it as some sort of spam, that is all. As you have clearly spent a lot of time on this, I wish you all the best with it & I did not mean to offend you or anyone else.

[lengoundry]

Hello everyone!

Recently I've started developing a program which may or may not be eventually useful for lottery enthusiasts. Since I don't really know much about the lotteries and various strategies and systems (I am a programmer and this idea came to my mind while developing an optimized matching algorithm), I would like to humbly ask you for your opinion and perhaps an advice about the usability of this program and its features.

The initial idea was this:
In my country there is a 6/49-based national lottery where, besides an ordinary 6-numbers ticket, one can take a system bet consisting of 7 up to 15 numbers. Sure that way you can cover more combinations, but is it really worth it? So I was thinking that maybe computing all possible combinations from 6 (let's support even standard 6-numbers bet too) to 15 numbers and then comparing them against the results of the previous draws it would be possible to find combinations which (had they been used) in the past have got the highest chances to match the actual drawn numbers. Then maybe using some sort of time analysis (how evenly a combination "scored" in the past draws) it could give some clue as whether given system combination might score in the future.
(For example, if I find a combination which in the last 30 years would have scored in 15% of all cases but looking at the dates it would appear that the majority of those 15% actually happened mainly during 15-25 years ago, then such a combination is probably not very great for the future bets. On the other hand a combination with maybe slightly lower % success but scoring more continuously from far past to the very present may be a better candidate.)

This idea may or may not be actually useful, you know surely much better than me, anyway from the programming perspective this was an interesting problem because it would require a lot of computing. Standard 6-numbers bet makes 13,983,816 possible combinations, that's nothing. 7-numbers bet goes to 85,900,584 combinations, 8-numbers 450,978,066 combinations, ..., and 15-numbers system bet can be actually selected out of 1,575,580,702,584 possible combinations. Now to make it work all those combinations would have to be computed and then compared with the lottery history data - which may be thousands draws. So each computed combination must be compared to all past draws, respective % chance must be calculated, the same thing must be done with all other millions and billions or even trillions of combinations, and at the end it should print out several best combinations. And all that in a reasonable time.

This is where my then-developed matching algorithm kicks in as it was supposed to provide some speed boost, so to test it in the reality I've implemented it in this context and built this program around it.

For the time being the performance is as follows:
* 900 millions of combinations per second are computed
* 90 millions /s are processed and delivered to the matching algorithm
* the speed of actual testing depends on how many history draws the computed combinations are to be compared with:

- 2.2 millions combinations /s when comparing against 100 lottery draws
- 212 thousands /s when comparing against 1200 lottery draws
- 50 thousands /s when comparing against 5300 lottery draws

The above example is the performance per a one CPU core (3,15 GHz in this case), thus using e.g. 4 cores of 4-core CPU the speed will result in 4 higher speed. The program can be divided into thousands pieces and can be run simultaneously and totally independently on as many cores, processors and even separate individual computers as needed. I am still not satisfied with the speed, though. Also, the time analysis is not implemented yet, that will probably drag the speed down even further.

Anyway, even in the current work-in-progress developing state, all possible 1,5 trillions of 15-numbers combinations can be computed, processed and compared against 5300 lottery draws (our lottery's data from last 35 years) in no longer than a single month, using 3 CPUs, each with 4 cores. That's not that bad, I guess.

Program parameters:

$ ./lotto
Lotto ver. 1.1
==============

This program will compute all possible k-combinations of 49 numbers (1 - 49), drops combinations containing any blacklisted number (if any specified in blfile), further drops combinations not containing all whitelisted numbers (if any specified in wlfile) and then from the remaining set takes each combination and compares it against all 6/49 lottery draws (specified in dtfile) calculating the percent rate of 4-matched numbers, 5-matched numbers and 6-matched numbers. At the end the program will show the TOP-3 combinations with the highest combined percentage.
In the alternate mode (pfreq is set to 999) the program will not calculate anything, just print out all computed (and all 'to be tested' - with regards to the blacklist and whitelist) combinations.
(C) 2011 johnsmithx (johnsmithx at seznam.cz)

Usage: lotto dtfile blfile wlfile k pfreq start stop [benchmark]
  or   lotto dtfile blfile wlfile k 999 start stop

Parameters:
   dtfile       csv file, at least 1 row, 6 values per row
   blfile       csv file, either empty or up to 49-k numbers on a single row
   wlfile       csv file, either empty or up to k numbers on a single row
   k            k-combinations (6 - 15)
   pfreq        progress report frequency in % (0.01 - 50)
   start        start computing at specified progress % (0 - 99.99)
   stop         stop computing at specified progress % (start+0.01 - 100)
   benchmark    period of seconds for initial benchmark (5 - 100), default 10

So as you can see the program is technically focused and doesn't have any eye candy useless features. It's made so that all parameters can be specified as the command arguments, the output can be redirected to a text file and the program can be run as a scheduled job in the background. Currently the platform is Linux x86_64 with assembly optimizations for modern core2 processors with sse2/ssse3 extensions.

Besides the actual history data provided in the form of a simple CSV file (comma separated values, just 6 numbers at each line, plain and clear) the program can work with the blacklist and whitelist files. In the blacklist you can specify numbers you DO NOT want to appear in the tested combinations. In the whitelist you can specify the exact opposite - the numbers you DO want to have in the tested combinations. Program will always compute all possible combinations (quick part) but only those in compliance with blacklist/whitelist are passed for testing (slow part).

You also specify how often should be shown the intermediate progress report, so you can see in what point the program run is at the moment, how long is it running and how long is expected to the end.

The very important feature is specifying the start and stop point, which allows to divide program run to smaller fragments and run then either one after another or all at once. Every running instance of the program will bind itself to a single core of a single CPU.

Example of output:

$ ./lotto lotto_data_5336_1977w1-2011w39.csv lotto_blacklist.csv lotto_whitelist.csv 7 20 0 100
Lotto ver. 1.1
==============

Input data statistics: datafile 5336 rows, blacklistfile 0 numbers, whitelistfile 0 numbers
Requested 7-combinations of 1 - 49: 85900584 combinations to compute, 85900584 combinations to test
   parameters: start 0.00% stop 100.00% pfreq 20.00% benchmark 10 s

START COMPUTING at 0.00%  (2011-11-20 17:07:48 +0100)
 benchmarking: computed combination no. 430000 (43000/s), tested combination no. 430000 (43000/s)
               elapsed: 10 s, remaining: 33 m 7 s  (2011-11-20 17:07:58 +0100)

Progress 20.00%, computed combination no. 17180117 (46685/s), tested combination no. 17180117 (46685/s)
elapsed: 6 m 8 s, remaining: 24 m 31 s  (2011-11-20 17:13:56 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

Progress 40.00%, computed combination no. 34360234 (46685/s), tested combination no. 34360234 (46685/s)
elapsed: 12 m 16 s, remaining: 18 m 23 s  (2011-11-20 17:20:04 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

Progress 60.00%, computed combination no. 51540351 (46685/s), tested combination no. 51540351 (46685/s)
elapsed: 18 m 24 s, remaining: 12 m 15 s  (2011-11-20 17:26:12 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

Progress 80.00%, computed combination no. 68720468 (46685/s), tested combination no. 68720468 (46685/s)
elapsed: 24 m 32 s, remaining: 6 m 7 s  (2011-11-20 17:32:20 +0100)
...1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
...2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
...3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

COMPUTING COMPLETED at 100.00% in 30 m 39 s  (2011-11-20 17:38:27 +0100)
 85900584 combinations computed (46710/s), 85900584 combinations tested (46710/s)

Results
=======
1st comb: 5, 14, 21, 25, 30, 39, 48 prob: 0.6372% (32x 4, 2x 5, 0x 6)
2nd comb: 5, 18, 21, 24, 25, 26, 48 prob: 0.6372% (34x 4, 0x 5, 0x 6)
3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)

What I would like to ask you is this:
- What do you think about this idea and the program? Can it be of any use for lottery purposes at all or is it just non-practical nonsense?
- What other features could be useful to implement?

Thanks for your answers.

johnsmithx

Hi Johnsmithx
You are brilliant and I do not think you know how brilliant you are!
The health Lottery
14 Jan 2012 Draw 15) 05,14,29,41,48

your post
3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)


Groups 5,4,9,1,8
i.e. you missed two numbers by two, or should I say ten to the right.
i.e. you posted 39 it was 29 you posted 31 it was 41

3rd comb: 5, 14, 21, 25, 31, 39, 48 prob: 0.6184% (33x 4, 0x 5, 0x 6)
------------5, 14,----------29, 41, 48 The Health Lottery

Please keep up the good work and please keep posting here it will help other's