Author |
Topic: Lottery numbers & non repeating values tool (Read 1135 times) |
|
PatrickM
New Member
member is offline
Posts: 14
|
 |
Re: Lottery numbers & non repeating values tool
« Reply #3 on: Apr 10th, 2017, 3:38pm » |
|
Quote:Even faster would be to do this, but some people don't like to see multiple exit points from a function:
Code: DEF FN_checkduplicates(x%(), i%, n%) LOCAL I%,J% IF n% <= i% THEN = FALSE FOR I% = i% TO n%-1 FOR J% = i%+1 TO n% IF x%(I%) = x%(J%) THEN = TRUE NEXT NEXT = FALSE
|
|
This surprised me, I didn't know you could do that - I thought it would result in a 'Too many FORs' error, but now that I've tested it I see that would only be a problem on the BBC Micro and (I assume) other older versions of BBC Basic.
|
|
 Logged
|
|
|
|
Richard Russell
Administrator
member is offline
Posts: 803
|
|
Re: Lottery numbers & non repeating values tool
« Reply #4 on: Apr 10th, 2017, 4:10pm » |
|
on Apr 10th, 2017, 3:38pm, PatrickM wrote:| I didn't know you could do that - I thought it would result in a 'Too many FORs' error... |
|
Possibly, but if so I think that behaviour is unique to the 6502 versions of BBC BASIC. As far as I know every other version, including later ones from the Acorn stable (e.g. ARM BASIC V), all of mine (including the 8-bit Z80 version) and Brandy allow you to abort from inside a FOR...NEXT - or any other variety of loop - like that.
Nevertheless EXIT FOR etc. (available in BB4W and BBCSDL) is generally preferable. Its omission from earlier versions of BBC BASIC is one of the few significant limitations of the language, in my opinion. It's understandable that it wasn't in the 8-bit versions but with hindsight it should have been added at the same time as WHILE...ENDWHILE and multi-line IF...ENDIF.
Richard.
|
|
Logged
|
|
|
|
MarsFS
New Member
member is offline
Gender: 
Posts: 16
|
|
Re: Lottery numbers & non repeating values tool
« Reply #5 on: Sep 1st, 2017, 04:39am » |
|
Here's another example of how it might be done using less loops and reduced redundant comparisons. The check for dupes FN gets called before assigning a new number to the array and printing is included in the main loop. Variables have been added to allow a different amount of numbers and number range.
Another nice addition would be to sort the numbers before printing.
Code: REM Lotto number generator max%=49 nums%=7 DIM x%(nums%-1) REPEAT PROC_newnumbers key$ =GET$ UNTIL key$="q" OR key$="Q" PRINT "Done" END DEF PROC_newnumbers LOCAL t,n CLS PRINT "Here is ";STR$(nums%);" numbers for Lotto Max:": PRINT FOR t=0 TO nums%-1 REPEAT n=RND(max%) UNTIL FNnodupes(n,t-1) x%(t)=n PRINT STR$(n);" "; NEXT t PRINT: PRINT PRINT "Press any key for more numbers, Press 'Q' to quit" ENDPROC DEF FNnodupes(n,c) LOCAL t IF c>-1 THEN FOR t=0 TO c IF n=x%(t) =FALSE NEXT ENDIF =TRUE |
|
Logged
|
|
|
|
Leo
New Member
member is offline
Posts: 10
|
|
Re: Lottery numbers & non repeating values tool
« Reply #6 on: Sep 1st, 2017, 5:19pm » |
|
Here's a very simple version which does multiple lines and outputs numbers in order (without sorting).
Code: REM Simple Lottery. Multiple lines, output ordered. NMAX%=49 LMAX%=7 DIM N&(NMAX%) @%=&02 INPUT "How many lines", J% FOR I%=1 TO J% PROC_Line NEXT END REM Create and print a lottery line DEF PROC_Line LOCAL c%, i%, x% N&()=0 REPEAT x% = RND(NMAX%) IF N&(x%) = 0 THEN N&(x%) = x% c% += 1 ENDIF UNTIL c% = LMAX% FOR i%=1 TO 49 IF N&(i%) THEN PRINT " " i%,; NEXT PRINT ENDPROC |
| « Last Edit: Sep 1st, 2017, 7:08pm by Leo » |
Logged
|
|
|
|
Richard Russell
Administrator
member is offline
Posts: 803
|
|
Re: Lottery numbers & non repeating values tool
« Reply #7 on: Sep 1st, 2017, 8:35pm » |
|
on Sep 1st, 2017, 5:19pm, Leo wrote:| Here's a very simple version which does multiple lines and outputs numbers in order (without sorting). |
|
Personally I dislike and distrust all solutions that rely on repeatedly choosing a 'random' number until it's not one of a previously chosen set (or repeatedly choosing a set of numbers until some condition is met).
I dislike it because the execution time is not predictable. Suppose one was using that technique to draw 52 cards from a pack; the last card drawn is already known (it's the only one left) but a 'repeat RND(52) until not already chosen' loop might have to repeat dozens or hundreds of times before that unique number is selected. Theoretically it could take an indefinitely long time.
I distrust it because it is not obvious 'by inspection' that the method is statistically sound. It probably is, but as I'm not a mathematician I can't be confident that it really is resulting in an unbiased selection (to the degree that the pseudo-random number generator allows, at least).
So I would always prefer a method that has a predictable execution time and guaranteed statistics, such as a well understood and documented shuffling algorithm (e.g. the Fisher-Yates or Knuth shuffle).
Richard.
|
|
|
|
MarsFS
New Member
member is offline
Gender:
Posts: 16
|
|
Re: Lottery numbers & non repeating values tool
« Reply #8 on: Sep 1st, 2017, 10:20pm » |
|
Here's another version using a knuth shuffle with multiple lines and sorted and formatted output:
Code: REM Lotto number generator max%=49 nums%=7 DIM allNumbers(max%) DIM sorted(max%) REM Creat a fresh set of numbers FOR T%=1 TO max% allNumbers(T%)=T% NEXT REPEAT PROC_newnumbers key$ =GET$ UNTIL key$="q" OR key$="Q" PRINT "Done" END DEF PROC_newnumbers LOCAL T%,N%,C%,J%,A$ CLS PRINT "Lotto Number Picker" PRINT "-------------------" INPUT "How many lines (1-18): ", J% IF J%<2 OR J%>18 THEN J%=1 A$="is 1 line" ELSE A$="are "+STR$(J%)+" lines" ENDIF CLS PRINT "Here ";A$;" of ";STR$(nums%);" numbers for Lotto Max:": PRINT FOR C%=1 TO J% REM randomly shuffle the numbers FOR T% = max% TO 2 STEP -1 SWAP allNumbers(T%),allNumbers(RND(T%)) sorted(T%)=0 NEXT sorted(1)=0 REM choose a random set and sort N%=RND(max%-nums%+1) FOR T%=N% TO N%+nums%-1 sorted(allNumbers(T%))=allNumbers(T%) NEXT REM print sorted list PRINT "Line ";RIGHT$("0"+STR$(C%),2);": "; FOR T%=1 TO max% IF sorted(T%) PRINT RIGHT$(" "+STR$(sorted(T%)),2);" "; NEXT PRINT NEXT PRINT:PRINT "Press any key for more numbers, Press 'Q' to quit" ENDPROC
|
|
Logged
|
|
|
|
Richard Russell
Administrator
member is offline
Posts: 803
|
|
Re: Lottery numbers & non repeating values tool
« Reply #9 on: Sep 2nd, 2017, 11:15am » |
|
on Sep 1st, 2017, 10:20pm, MarsFS wrote:| Here's another version using a knuth shuffle with multiple lines and sorted and formatted output |
|
Nice, and works without modification on both BB4W and BBCSDL!
For my own interest I ran some tests on how many calls to RND(52) might be necessary before a specific unique number is returned. The answer seemed to be that needing to call it more than 900 times is not that unlikely (the highest I saw was 996)! Obviously typically it will be fewer, but it illustrates the uncertainty in execution time that can arise from using that technique.
Richard.
|
|
Logged
|
|
|
|
Leo
New Member
member is offline
Posts: 10
|
|
Re: Lottery numbers & non repeating values tool
« Reply #10 on: Sep 2nd, 2017, 4:00pm » |
|
Although I can't disagree with the logic behind "run time is unpredictable" - the lottery case here is a complete frippery. It's just a "throwaway" toy.
Of course I wouldn't use the technique I did for anything approaching any seriousness but I stand by my choice for the lottery case because..
a) It's a complete frippery.
b) The program is very short.
c) We're only selecting ANY 7 numbers from 49 so there's no chance of this taking forever.
d) The quality of the random numbers is irrelevant because any 7 will have an equal chance of winning.
I also ran some trials (a lot).
The maximum number of lookup clashes in any single line was 6 - sounds a lot.
However, in creating 10000 lottery lines a repeat lookup was necessary in 6.7% of cases - no problem.
Henceforward I will use the dates of my children's birthdays to derive (winning) lottery lines. 
|
|
Logged
|
|
|
|
Richard Russell
Administrator
member is offline
Posts: 803
|
|
Re: Lottery numbers & non repeating values tool
« Reply #11 on: Sep 2nd, 2017, 4:43pm » |
|
on Sep 2nd, 2017, 4:00pm, Leo wrote:| I stand by my choice for the lottery case because... |
|
That's fine, with those qualifications. The trouble is that people may adapt the code for other applications without appreciating its limitations (for example they may come upon it years later as the result of a search).
My experience is that there is widespread misunderstanding about pseudo-random numbers and statistics, and although it's unlikely that anybody would use BBC BASIC for an online casino program, or something like that, it's not impossible.
Several years ago I wrote this article at the Wiki to draw attention to some of the pitfalls. It includes simple lottery draw and card shuffling programs. It's out of date to the extent that BBC BASIC no longer uses the value of TIME as a seed.
Richard.
|
|
|
|
Leo
New Member
member is offline
Posts: 10
|
|
Re: Lottery numbers & non repeating values tool
« Reply #12 on: Sep 2nd, 2017, 9:00pm » |
|
Thanks for that link Richard - very good article.
Although I'm gradually working my way through the info in the wiki, I admit I haven't read that one before. Everybody who uses BB4W should go and rummage through the wiki - there's so much excellent information in there.
I use CryptGenrandom a lot for 'professional' use. I think I posted something here that used it a short time ago - but I don't remember what right now. (My memory ain't what it used to be).
Thanks again for the info.
|
|
Logged
|
|
|
|
|
