Truly random?

[Alvin]

Well, I'm developing a Tetris game in SDL, but when it comes to
deciding the next block, I'm stuck. It's random, but when I try
something like seeding the randomizer with the time, it won't update as
fast as one block can fall, and the next to be determined. Generating
different numbers in one spur can work, but people can play Tetris for
hours (or even days), and so you can't predict how long. You could
constantly be making more with the same system as making, say 5 random
numbers out of a seed, but that would prove system intensive if the
game already uses a lot of memory (not that Tetris does, but I'm sure
there's a better way).

[Dag-Erling Smørgrav]

Re: Truly random?

"Alvin" <alvin.sipraga@ gmail.com> writes:[color=blue]
> Well, I'm developing a Tetris game in SDL, but when it comes to
> deciding the next block, I'm stuck. It's random, but when I try
> something like seeding the randomizer with the time, it won't update as
> fast as one block can fall, and the next to be determined.[/color]

You only need to seed the PRNG once, at the start of the game.

DES
--
Dag-Erling Smørgrav - des@des.no

[Randy Howard]

Re: Truly random?

Alvin wrote
(in article
<1133884774.799 489.117310@f14g 2000cwb.googleg roups.com>):
[color=blue]
> Well, I'm developing a Tetris game in SDL, but when it comes to
> deciding the next block, I'm stuck. It's random, but when I try
> something like seeding the randomizer with the time, it won't update as
> fast as one block can fall, and the next to be determined.[/color]

It sounds like you are trying to seed on every random call. You
only need to seed it once, or perhaps once per game worst case,
not every time. That would slow things down quite a bit, and
not achieve much at all.



--
Randy Howard (2reply remove FOOBAR)
"The power of accurate observation is called cynicism by those
who have not got it." - George Bernard Shaw

[Gordon Burditt]

Re: Truly random?

If, as your subject line suggests, you want truly random numbers,
forget a PRNG like rand() or random(). The PR part stands for
PSEUDO-RANDOM. Computers aren't very good at generating truly
random numbers unless they are broken or they have specific hardware
for the purpose. Some attempts at truly random number generation
time keystrokes, or listen to thermal noise or radioactive decay,
or time events from outside the computer (like network traffic).

If you need truly random numbers for cryptography, a PRNG won't
do. The difference can get you killed.
[color=blue]
>Well, I'm developing a Tetris game in SDL, but when it comes to
>deciding the next block, I'm stuck. It's random, but when I try
>something like seeding the randomizer with the time, it won't update as
>fast as one block can fall, and the next to be determined. Generating[/color]

Don't seed the PSEUDO random number generator more than once. Since
the time() call gives a time granularity of seconds on most (POSIX)
systems, and presumably blocks update faster than that, you'll get
sucky non-random numbers.
[color=blue]
>different numbers in one spur can work, but people can play Tetris for
>hours (or even days), and so you can't predict how long. You could
>constantly be making more with the same system as making, say 5 random
>numbers out of a seed, but that would prove system intensive if the
>game already uses a lot of memory (not that Tetris does, but I'm sure
>there's a better way).[/color]

Seed once. Period. If calling rand() or random() alone slows down
the display too much, perhaps you need a faster CPU. But I doubt it.
All the graphics probably takes a lot more work than just generating
a pseudo-random number.

Gordon L. Burditt

[Alvin]

Re: Truly random?

Well, it works like a charm for large numbers, but for small numbers
(say, 0 to 7), it often repeats them 3 or so times before changing. I
guess I could determine on how big the number is, so if it's > 100 and
< 200, select an right sided L block or something. Thanks very much,
though.

[Mogens Heller Jensen]

Re: Truly random?


"Alvin" <alvin.sipraga@ gmail.com> wrote in message
news:1133884774 .799489.117310@ f14g2000cwb.goo glegroups.com.. .[color=blue]
> Well, I'm developing a Tetris game in SDL, but when it comes to
> deciding the next block, I'm stuck. It's random, but when I try
> something like seeding the randomizer with the time, it won't update as
> fast as one block can fall, and the next to be determined. Generating
> different numbers in one spur can work, but people can play Tetris for
> hours (or even days), and so you can't predict how long. You could
> constantly be making more with the same system as making, say 5 random
> numbers out of a seed, but that would prove system intensive if the
> game already uses a lot of memory (not that Tetris does, but I'm sure
> there's a better way).
>[/color]

The ideal solution (which, however, requires an internet connection) is to
connect to www.random.org and download some random bytes...

They are truly random, since they come from *sampling backgound noise in the
atmosphere* (how cool is that?!).

This would also be a cool selling argument!

</funmaking>
<serious>
Check Randy Howard's answer :o)

-Mogens

[Randy Howard]

Re: Truly random?

Gordon Burditt wrote
(in article <11pbgkunf7spd7 8@corp.supernew s.com>):
[color=blue]
> If, as your subject line suggests, you want truly random numbers,
> forget a PRNG like rand() or random(). The PR part stands for
> PSEUDO-RANDOM. Computers aren't very good at generating truly
> random numbers unless they are broken or they have specific hardware
> for the purpose. Some attempts at truly random number generation
> time keystrokes, or listen to thermal noise or radioactive decay,
> or time events from outside the computer (like network traffic).
>
> If you need truly random numbers for cryptography, a PRNG won't
> do. The difference can get you killed.[/color]

He said he's writing a Tetris game. I find it highly unlikely
he needs crypto random numbers for that.

--
Randy Howard (2reply remove FOOBAR)
"The power of accurate observation is called cynicism by those
who have not got it." - George Bernard Shaw

[Randy Howard]

Re: Truly random?

Alvin wrote
(in article
<1133888616.701 661.86770@g49g2 000cwa.googlegr oups.com>):
[color=blue]
> Well, it works like a charm for large numbers, but for small numbers
> (say, 0 to 7), it often repeats them 3 or so times before changing.[/color]

Random does not mean "no repeats", especially in such a narrow
range. People often think that the output from a random number
generator should be evenly disributed. Bzzt.

If you are still reseeding it over and over, that will skew your
results. Fix that first.

If you think the random implementation on your platform is
flawed (after seeding it properly), then check out the Mersenne
Twister PRNG. It's probably overkill for a video game, but it
might make you feel better anyway.



--
Randy Howard (2reply remove FOOBAR)
"The power of accurate observation is called cynicism by those
who have not got it." - George Bernard Shaw

[Gordon Burditt]

Re: Truly random?

>> If, as your subject line suggests, you want truly random numbers,[color=blue][color=green]
>> forget a PRNG like rand() or random(). The PR part stands for
>> PSEUDO-RANDOM. Computers aren't very good at generating truly
>> random numbers unless they are broken or they have specific hardware
>> for the purpose. Some attempts at truly random number generation
>> time keystrokes, or listen to thermal noise or radioactive decay,
>> or time events from outside the computer (like network traffic).
>>
>> If you need truly random numbers for cryptography, a PRNG won't
>> do. The difference can get you killed.[/color]
>
>He said he's writing a Tetris game. I find it highly unlikely
>he needs crypto random numbers for that.[/color]

True, but he asked for truly random numbers. Perhaps he's betting
large amounts on the outcome (anyone for Tetris machines next to
the video poker machines and the slots?), enough to make organized
efforts to cheat worth the time and trouble. If you're building
gambling machines which will take bets of real money, you *DO* need
crypto-quality random numbers or you're going to go broke.

Gordon L. Burditt

[Randy Howard]

Re: Truly random?

Gordon Burditt wrote
(in article <11pbk2mb3oopp1 e@corp.supernew s.com>):
[color=blue][color=green]
>> He said he's writing a Tetris game. I find it highly unlikely
>> he needs crypto random numbers for that.[/color]
>
> True, but he asked for truly random numbers. Perhaps he's betting[/color]

I'd rather bet on the outcome of the Rose Bowl...
[color=blue]
> large amounts on the outcome (anyone for Tetris machines next to
> the video poker machines and the slots?), enough to make organized
> efforts to cheat worth the time and trouble.[/color]

Ok, but I don't see a market for "Tetris Gambling". Maybe there
is, prove me wrong. :-)


--
Randy Howard (2reply remove FOOBAR)
"The power of accurate observation is called cynicism by those
who have not got it." - George Bernard Shaw

[Walter Roberson]

Re: Truly random?

In article <11pbk2mb3oopp1 e@corp.supernew s.com>,
Gordon Burditt <gordonb.8lsbe@ burditt.org> wrote:[color=blue]
>True, but he asked for truly random numbers. Perhaps he's betting
>large amounts on the outcome (anyone for Tetris machines next to
>the video poker machines and the slots?), enough to make organized
>efforts to cheat worth the time and trouble. If you're building
>gambling machines which will take bets of real money, you *DO* need
>crypto-quality random numbers or you're going to go broke.[/color]

[OT]

My recollection is somewhat vague, as I just skimmed the material
some time ago, but if I recall what I have read correctly, the
professional casinos in places like Los Vegas do -not- use
crypto-quality random numbers, at least for some kinds of machines
such as slot machines. I seem to recall reading that they use a PRNG
but with algorithm parameters modified by a central computer.

The PNRG use has partly to do with managable auditing -- recording each
individual truly random number would take too much storage. The
ability to modify the parameters is used to adjust the payout rate
based upon the payin rate -- essentially once the intake has reached a
trigger value, the central computer adjusts the payout probabilities
upwards to make it more likely that someone will win a large prize. For
one thing, a slot machine that shows exactly the same losing line of
symbol 100 times in a row might be truly random, but it won't be
-perceived- to be random...

PNRG are losing propositions in gambling to the extent that someone
can effectively analyze the past history to predict the future outcomes.
A simple linear congruence PNRG is very risky in that respect, but
there are PRNG with large internal states, and if those states are
reseeded at intervals shorter than would be needed to extract useful
state information, then PRNG can become practical.
--
"It is important to remember that when it comes to law, computers
never make copies, only human beings make copies. Computers are given
commands, not permission. Only people can be given permission."
-- Brad Templeton

[Kenny McCormack]

Re: Truly random?

In article <1133888616.701 661.86770@g49g2 000cwa.googlegr oups.com>,
Alvin <alvin.sipraga@ gmail.com> wrote:[color=blue]
>Well, it works like a charm for large numbers, but for small numbers
>(say, 0 to 7), it often repeats them 3 or so times before changing. I
>guess I could determine on how big the number is, so if it's > 100 and
>< 200, select an right sided L block or something. Thanks very much,
>though.[/color]

(Obviously, what follows is off-topic, not portable, and blah, blah, blah,
but anyway...)

Long ago and far away, using a compiler/development environment that need
not be mentioned here, I discovered that the higher order bits seemed to
be more random than the lower ordered ones. So, that the usual method (get
the number between, say, 0 and 32767, then divide it by the top of your
range, to get a random number between 0 and your range) tended to give less
good random results. So, I would take the result of random and bit shift
it left 5 (<< 5) to get the lower ordered bits more random. Seems to work.

[mensanator@aol.com]

Re: Truly random?


Kenny McCormack wrote:[color=blue]
> In article <1133888616.701 661.86770@g49g2 000cwa.googlegr oups.com>,
> Alvin <alvin.sipraga@ gmail.com> wrote:[color=green]
> >Well, it works like a charm for large numbers, but for small numbers
> >(say, 0 to 7), it often repeats them 3 or so times before changing. I
> >guess I could determine on how big the number is, so if it's > 100 and
> >< 200, select an right sided L block or something. Thanks very much,
> >though.[/color]
>
> (Obviously, what follows is off-topic, not portable, and blah, blah, blah,
> but anyway...)
>
> Long ago and far away, using a compiler/development environment that need
> not be mentioned here, I discovered that the higher order bits seemed to
> be more random than the lower ordered ones. So, that the usual method (get
> the number between, say, 0 and 32767, then divide it by the top of your
> range, to get a random number between 0 and your range) tended to give less
> good random results. So, I would take the result of random and bit shift
> it left 5 (<< 5) to get the lower ordered bits more random. Seems to work.[/color]

Gee, I got a new asshole reamed for suggesting that very thing
a couple months ago despite it being in the FAQ.

Better sit on your helmet.

[Michael Mair]

Re: Truly random?

Kenny McCormack wrote:[color=blue]
> In article <1133888616.701 661.86770@g49g2 000cwa.googlegr oups.com>,
> Alvin <alvin.sipraga@ gmail.com> wrote:
>[color=green]
>>Well, it works like a charm for large numbers, but for small numbers
>>(say, 0 to 7), it often repeats them 3 or so times before changing. I
>>guess I could determine on how big the number is, so if it's > 100 and
>>< 200, select an right sided L block or something. Thanks very much,
>>though.[/color]
>
>
> (Obviously, what follows is off-topic, not portable, and blah, blah, blah,
> but anyway...)
>
> Long ago and far away, using a compiler/development environment that need
> not be mentioned here, I discovered that the higher order bits seemed to
> be more random than the lower ordered ones. So, that the usual method (get
> the number between, say, 0 and 32767, then divide it by the top of your[/color]

You probably are talking about talking the remainder modulo the top
of the range. "Divide" was the other thingy, the one which worked
better than plain modulo.
[color=blue]
> range, to get a random number between 0 and your range) tended to give less
> good random results. So, I would take the result of random and bit shift
> it left 5 (<< 5) to get the lower ordered bits more random. Seems to work.[/color]

You probably mean the other left. The one most people call "right".
rand() << 5 certainly will give you five quite unrandom lower bits.

-Michael
--
E-Mail: Mine is an /at/ gmx /dot/ de address.

[Jordan Abel]

Re: Truly random?

On 2005-12-06, Kenny McCormack <gazelle@yin.in teraccess.com> wrote:[color=blue]
> In article <1133888616.701 661.86770@g49g2 000cwa.googlegr oups.com>,
> Alvin <alvin.sipraga@ gmail.com> wrote:[color=green]
>>Well, it works like a charm for large numbers, but for small numbers
>>(say, 0 to 7), it often repeats them 3 or so times before changing. I
>>guess I could determine on how big the number is, so if it's > 100 and
>>< 200, select an right sided L block or something. Thanks very much,
>>though.[/color]
>
> (Obviously, what follows is off-topic, not portable, and blah, blah, blah,
> but anyway...)
>
> Long ago and far away, using a compiler/development environment that
> need not be mentioned here, I discovered that the higher order bits
> seemed to be more random than the lower ordered ones. So, that the
> usual method (get the number between, say, 0 and 32767, then divide it
> by the top of your range, to get a random number between 0 and your
> range) tended to give less good random results. So, I would take the
> result of random and bit shift it left 5 (<< 5) to get the lower
> ordered bits more random. Seems to work.[/color]

I assume you mean to the right.

And it is marginally on-topic, because the code of a PRNG with IIRC
characteristics like those you describe appears in the text of the C
standard.