BBC BASIC for Windows
Programming >> BBC BASIC language >> re RND
http://bb4w.conforums.com/index.cgi?board=language&action=display&num=1270164019
re RND
Post by leslie griffin on Apr 1st, 2010, 11:20pm

Updated my prog with the bugfree lotto generator RND.
I think anyone can run it now.
something is wrong
if i choose numbers less than 43 it seems to give to high a payout.
if i choose numbers in the 40s it pays far lower than it should.
it was showing approx 22% with my original but longwinded RND generator and error check
mode 20
dim ltt%(49)
dim lot(10,6)
dim drww(7)
dim hit(10)
dim ttal(10,7)
dim mwon(10)
gh=1
ttdr=1
for f=1 to10
mwon(f)=1
for g= 1 to 6
lot(f,g)=0
next
next
proc_setup
proc_choose6
off

repeat
proc_draw
proc_logic
proc_scrn
ttdr=ttdr+1
until gh=2
end

defproc_setup
cls
print tab(0,6)"This is a lottery simulator"
input tab(0,15)"how mamy tickets do you want Max=10 " ticket
print tab(0,8)"now choose 6 different numbers between 1 and 49 for each ticket press enter after each entry "
print tab(0,9)"3 numbers pays 10"
print tab(0,10)"4 numbers pays 50"
print tab(0,11)"5 numbers pays 1000"
print tab(0,12)" 5 numbers plus the bonus ball pays 100000"
print tab(0,13)"6 numbers pays 3 million"
endproc


defproc_choose6
for d=1 to ticket
for t=1 to 6
repeat
input tab(t*4,16+d) lot(d,t)
until (lot(d,t)>0 and lot(d,t)<50)
next
next
endproc

defproc_draw
max% = 49
num% = 7
for I% = 1 to max%
ltt%(I%) = I%
next I%

for choice% = 1 to num%
R% = rnd(max%)
drww(choice%)=R%
ltt%(R%) = ltt%(max%)
max% = max%-1
next choice%

endproc

defproc_logic
for z=1 to ticket
hit=0
for n=1 to 6
for m=1 to 6
if lot(z,m) = drww(n) then hit = hit+1
next
next
if hit=1 then ttal(z,1)=ttal(z,1)+1
if hit=2 then ttal(z,2)=ttal(z,2)+1
if hit=3 then ttal(z,3)=ttal(z,3)+1 mwon(z)=mwon(z)+10
if hit=4 then ttal(z,4)=ttal(z,4)+1 mwon(z)=mwon(z)+50 sound 3,-15,96,1
if hit=5 then ttal(z,5)=ttal(z,5)+1 mwon(z)=mwon(z)+1000 proc_bonusball
if hit=6 then ttal(z,6)=ttal(z,6)+1 mwon(z)=mwon(z)+3000000
if hit=7 then ttal(z,7)=ttal(z,7)+1 mwon(z)=mwon(z)+100000
next
endproc

defproc_bonusball
for s=1 to 6
if drww(7)=lot(z,s) then hit=7
next
endproc

defproc_scrn
print tab(12,0)"draw number 3 HIT 4 HIT 5 HIT 6 HIT 5 + BONUS MONEY WON RATIO"
for s= 1 to ticket
print tab(6,1+s) ttdr ttal(s,3) ttal(s,4) ttal(s,5) ttal(s,6)" " ttal(s,7)" " mwon(s)" "int(mwon(s)/ttdr*100)"%"
print tab(0,15)"number of years running is " int(ttdr/52)
next
endproc




Re: re RND
Post by admin on Apr 2nd, 2010, 10:13am

on Apr 1st, 2010, 11:20pm, Guest-leslie griffin wrote:
something is wrong. if i choose numbers less than 43 it seems to give to high a payout. if i choose numbers in the 40s it pays far lower than it should.

I don't know how important this is, but in the case of 'five balls plus bonus ball' the payout is 101000 (not 100000) because you've added the normal '5 ball' payout as well.

BBC BASIC's RND is quite a good generator, considering its relatively short sequence length (2^33-1), because it uses a Linear Feedback Shift Register having desirable characteristics as a Pseudo Random Number Generator.

Richard.
Re: re RND
Post by admin on Apr 2nd, 2010, 10:55am

on Apr 1st, 2010, 11:20pm, Guest-leslie griffin wrote:
it was showing approx 22% with my original but longwinded RND generator and error check

22% is far too low a result, but whether that is because your original RND was untrustworthy or because you didn't run enough trials I don't know.

You would have to run your program for an infeasibly long time to get meaningful statistics. For example the chances of winning the top prize (6 numbers) are about 1 in 14 million; so to measure an average percentage payout figure that you can trust you'd have to run something like a billion trials!

If you disable the 100000 and 3000000 prizes, it's only necessary to run the program for about 500000 trials (!) to give reasonably reliable statistics. If I do that I get around a 60% average payout irrespective of the numbers I choose. I don't see any evidence of bias.

Unless you're a mathematician, working with 'random' numbers is beset with pitfalls. You must at the very least have an understanding of how many trials you need to carry out before the results have statistical significance.

Richard.