On Wednesday, November 30, 2022 at 10:12:15 PM UTC-7, Titus G wrote:

> I don't understand how there can be somewhat predictable, (or
> "remarkably poor"), random numbers as I had always assumed that
> unpredicability was a subset of randomness and that all random number
> sequences were random.

Until very recently, in order to make public-key ciphers secure, computers
didn't have access to any means of producing actually random numbers
at all. And these still aren't used for things like playing Solitaire.

Instead, the "random number" function seen in BASIC is an example of what
_will_ be used.

This is 100% predictable; it's only designed to *look* random, so that it will
usually be adequate, for example, in proposing how to take a "random" sample
of, say, the plants growing in a field for an experiment.

For a solitaire game, the starting "seed" number used for the (pseudo-)random number
generator would be generated from the exact second (gained from the computer's
real-time clock) the game was started... so that it's possible to play a new game
every time.

John Savard

On Thursday, 1 December 2022 at 06:30:58 UTC, Quadibloc wrote:
> On Wednesday, November 30, 2022 at 10:12:15 PM UTC-7, Titus G wrote:
>
> > I don't understand how there can be somewhat predictable, (or
> > "remarkably poor"), random numbers as I had always assumed that
> > unpredicability was a subset of randomness and that all random number
> > sequences were random.
> Until very recently, in order to make public-key ciphers secure, computers
> didn't have access to any means of producing actually random numbers
> at all. And these still aren't used for things like playing Solitaire.
>
> Instead, the "random number" function seen in BASIC is an example of what
> _will_ be used.
>
> This is 100% predictable; it's only designed to *look* random, so that it will
> usually be adequate, for example, in proposing how to take a "random" sample
> of, say, the plants growing in a field for an experiment.
>
> For a solitaire game, the starting "seed" number used for the (pseudo-)random number
> generator would be generated from the exact second (gained from the computer's
> real-time clock) the game was started... so that it's possible to play a new game
> every time.

A sequence of "random numbers" from a
pseudo-random-number formula ought to be
unpredictable by examining the output, even when
you know what the formula is, and a good one is
unpredictable. However, if the starting condition
or "seed" is known, then the output is always the
same. Randomness is achieved by using a random
"seed".

I had a "home computer", Sinclair ZX Spectrum
I think is where I did this, with the BASIC language
and 65535 random decimal fraction numbers,
since those were the possible values of the seed.
I used random numbers to plot points on the screen,
x coordinate and y coordinate, and I got neat stripes
slightly angled from vertical. So that wasn't as "random"
as it could have been.

On Thursday, December 1, 2022 at 4:41:06 AM UTC-7, Robert Carnegie wrote:

> A sequence of "random numbers" from a
> pseudo-random-number formula ought to be
> unpredictable by examining the output, even when
> you know what the formula is, and a good one is
> unpredictable.

In that case, "good" pseudorandom number generators
are hardly _ever_ used.

> However, if the starting condition
> or "seed" is known, then the output is always the
> same. Randomness is achieved by using a random
> "seed".

> I had a "home computer", Sinclair ZX Spectrum
> I think is where I did this, with the BASIC language
> and 65535 random decimal fraction numbers,
> since those were the possible values of the seed.
> I used random numbers to plot points on the screen,
> x coordinate and y coordinate, and I got neat stripes
> slightly angled from vertical. So that wasn't as "random"
> as it could have been.

Except for having only 65,535 possible seed values,
instead of, say, 4,294,967,296 possible seed values,
that is the sort of thing that is normal and standard for
pseudorandom number generators; typically, linear
congruential RNGs are used, not _even_ MacLaren -
Marsaglia, never mind anything actually cryptosecure,
which is what it would take for a sequence to be
unpredictable by examining the output when the
algorithm is known.

John Savard

On Thursday, December 1, 2022 at 12:12:15 AM UTC-5, Titus G wrote:

>I am not a gamer but some decades ago wore out a
> few packs of plastic playing cards playing Spider without takebacks.
> As an afterthought. That number of games played sharpened the speed and
> appearance of my shuffling skills which probably developed into a
> routine. So in theory, does that mean that the sequence of cards in a
> shuffled pack after many, many, iterations would become close to
> predictable?

As Marting Gardner would say paradoxically in his Mathematical Games columns, shuffling is a skill that depends on the clumsiness of the shuffler.

It can be shown that a perfect rifle shuffle -- divide the deck exactly in half, alternate one card from each half -- will return the deck to its original or "seed" configuration after 14 repetitions.

The other shuffle, "overhand," where you hold the deck in one hand and shake the top card into the other repeatedly, would, if perfect, return to the original state if done twice. On first learning that "14 times" thing above, and the subsequent warning that something close to seven shuffles would be best to assure "randomness," I did rifle shuffles of a deck one card at a time and examined the deck's randomness after each. I also came up with a psuedorandom repeatable "realistic" clumsy overhand shuffle: 1. take the top card from the deck, note the number on it, place it at the bottom of the new deck. 2. take off that number of cards from the top, note the number of the bottom card, place them on the new deck. 3. repeat #2 until you can't. This mimics my clumsy shuffle with six or seven shakes of different clumps of cards, but takes quite a lot of repetitions to get the deck away from its original state. As suits don't enter into it, you can divide the deck into red and black with the cards otherwise random and watch how slowly it happens.

--
-Jack

On Thursday, December 1, 2022 at 8:13:31 AM UTC-5, Quadibloc wrote:
> On Thursday, December 1, 2022 at 4:41:06 AM UTC-7, Robert Carnegie wrote:
>
> > A sequence of "random numbers" from a
> > pseudo-random-number formula ought to be
> > unpredictable by examining the output, even when
> > you know what the formula is, and a good one is
> > unpredictable.
> In that case, "good" pseudorandom number generators
> are hardly _ever_ used.
> > However, if the starting condition
> > or "seed" is known, then the output is always the
> > same. Randomness is achieved by using a random
> > "seed".
>
> > I had a "home computer", Sinclair ZX Spectrum
> > I think is where I did this, with the BASIC language
> > and 65535 random decimal fraction numbers,
> > since those were the possible values of the seed.
> > I used random numbers to plot points on the screen,
> > x coordinate and y coordinate, and I got neat stripes
> > slightly angled from vertical. So that wasn't as "random"
> > as it could have been.
> Except for having only 65,535 possible seed values,
> instead of, say, 4,294,967,296 possible seed values,
> that is the sort of thing that is normal and standard for
> pseudorandom number generators; typically, linear
> congruential RNGs are used, not _even_ MacLaren -
> Marsaglia, never mind anything actually cryptosecure,
> which is what it would take for a sequence to be
> unpredictable by examining the output when the
> algorithm is known.
>
> John Savard

John and others in this thread would do well to read the
Wikipedia articles on "Random Number Generation" and
"Hardware random number generator". Otherwise this is the
ignorant talking to the ignorant.

Donald Knuth famously said "Anyone who considers arithmetical methods
of producing random digits is, of course, in a state of sin."

Making cryptographically secure random numbers
on a computer is hard. Physical sources of
randomness must be used. This includes stuff like
mouse movements, disk seek times, fun stuff like
lavarand. (Lavarand.com is actually insecure,
since the same bitstream is broadcast to everyone, and
one may assume some people have been recording it since
day one).

There are also electronic methods that produce random data
based on quantum principles, described in the Wiki articles
I listed. You'd think that would solve the problem.

The problem is that physical methods tend to be slow. Data must
be gathered, preferably from multiple sources, mixed together
and whitened to produce an unbiased bit stream.

Since 2011, Intel has included a RDRAND instruction in its processors. This
claims to use a quantum electronic method to produce random numbers.
It is still slow compared to pseudo random number generators.

The question is "Do you trust Intel?". If I was a non-US State actor,
I'd be pretty suspicious. There's reason to believe the USG has messed
with random number generators in software libraries in the past.

pt

On 2022-12-01, Titus G <noone@nowhere.com> wrote:
> On 1/12/22 04:34, Chris Buckley wrote:
>> On 2022-11-30, WolfFan <akwolffan@zoho.com> wrote:
> snip
>
>> I've played many thousands of Spider games over the years. It's a
>> very well-balanced optimization game - I enjoy it for the same
>> reasons that I enjoyed the Civ games. It may take a while before you
>> win your first game (especially if you don't allow takebacks), but
>> you can push the optimizations to the point where you are winning
>> over a third of your games (with no takebacks).
>>
>> Watch out for poor implementations - those floating around based on
>> the Windows 7 free version have a remarkably poor RNG. I use
>> SolSuite which I like a lot, but it's Windows only.
>
> I don't understand how there can be somewhat predictable, (or
> "remarkably poor"), random numbers as I had always assumed that
> unpredicability was a subset of randomness and that all random number
> sequences were random. I am not a gamer but some decades ago wore out a
> few packs of plastic playing cards playing Spider without takebacks.
> As an afterthought. That number of games played sharpened the speed and
> appearance of my shuffling skills which probably developed into a
> routine. So in theory, does that mean that the sequence of cards in a
> shuffled pack after many, many, iterations would become close to
> predictable?
> I use Linux and see that Spider is in the AisleRiot package. How would I
> determine which Spider software had a better RNG? Thank you.

Any modern (<20 years?) implementation from scratch on 32 bit machines
is likely to be fine. It's easier to use the system RNG than roll
your own and those are fine nowadays, assuming the implementation
handles RNG seeds fine, does a reasonable approach to shuffling, and
doesn't do ridiculous things with multi-deck solitaires like shuffle
the decks separately.

The main problem is that some solitaire implementations have been
around for a long time and date back to 16-bit or even 8-bit days.
Code space was a major concern and they had their own RNGs, which were
often poor, yielding obvious patterns in the results which can be
deadly in Spider. In addition, just adapting those implementations for
32 or 64 bit machines or new operating systems may introduce new bugs
or misfeatures. This ranges from reasonably innocuous problems such
as Microsoft Freecell which until very recently (Windows 11?) only had
32K distinct games, because the RNG seed only had 15 bits, to more
serious problems.

In particular with Spider, if you download the Windows 7 version of
Spider to play on a Windows 10 machine (well-known compatibility
packages exist) as I did, you may eventually notice that you're
playing the same deal more than once. There was one particular deal
that I eventually realized was showing up about once every 30 games in
one time period! The problem seemed to come and go but was quite
annoying.

I would expect AisleRiot to be fine. Enough people have looked at
its open-source code over the years.

Chris

Titus G <noone@nowhere.com> writes:
>On 1/12/22 04:34, Chris Buckley wrote:
>> On 2022-11-30, WolfFan <akwolffan@zoho.com> wrote:
>snip
>
>> I've played many thousands of Spider games over the years. It's a
>> very well-balanced optimization game - I enjoy it for the same
>> reasons that I enjoyed the Civ games. It may take a while before you
>> win your first game (especially if you don't allow takebacks), but
>> you can push the optimizations to the point where you are winning
>> over a third of your games (with no takebacks).
>>
>> Watch out for poor implementations - those floating around based on
>> the Windows 7 free version have a remarkably poor RNG. I use
>> SolSuite which I like a lot, but it's Windows only.
>
>I don't understand how there can be somewhat predictable, (or
>"remarkably poor"), random numbers as I had always assumed that
>unpredicability was a subset of randomness and that all random number
>sequences were random.

Generating a truely random number is rather difficult. Most random
number generators are "Pseudo Random Number Generators" (PRNG) which
have repeatable sequences, given the same starting point. So
a lot of care goes into selecting a random starting point (seed), which
is very very hard.

There was a paper back in the mid 90's from an SGI engineer where
they used a video camera on a Indy to periodcally snap an image
of a working lava-lamp and generated the seed from that image.

Today, modern processors have a "true" random number generator
that builds an entropy pool from which random seed number can
be selected. But it takes a long time to generate sufficient
entropy to generate a large number of random values, so they
use the true number as a seed to a PRNG such as:

https://en.wikipedia.org/wiki/Mersenne_Twister

"pete...@gmail.com" <petertrei@gmail.com> writes:
>On Thursday, December 1, 2022 at 8:13:31 AM UTC-5, Quadibloc wrote:
>> On Thursday, December 1, 2022 at 4:41:06 AM UTC-7, Robert Carnegie wrote:
>>
>> > A sequence of "random numbers" from a
>> > pseudo-random-number formula ought to be
>> > unpredictable by examining the output, even when
>> > you know what the formula is, and a good one is
>> > unpredictable.
>> In that case, "good" pseudorandom number generators
>> are hardly _ever_ used.
>> > However, if the starting condition
>> > or "seed" is known, then the output is always the
>> > same. Randomness is achieved by using a random
>> > "seed".
>>
>> > I had a "home computer", Sinclair ZX Spectrum
>> > I think is where I did this, with the BASIC language
>> > and 65535 random decimal fraction numbers,
>> > since those were the possible values of the seed.
>> > I used random numbers to plot points on the screen,
>> > x coordinate and y coordinate, and I got neat stripes
>> > slightly angled from vertical. So that wasn't as "random"
>> > as it could have been.
>> Except for having only 65,535 possible seed values,
>> instead of, say, 4,294,967,296 possible seed values,
>> that is the sort of thing that is normal and standard for
>> pseudorandom number generators; typically, linear
>> congruential RNGs are used, not _even_ MacLaren -
>> Marsaglia, never mind anything actually cryptosecure,
>> which is what it would take for a sequence to be
>> unpredictable by examining the output when the
>> algorithm is known.
>>
>> John Savard
>
>John and others in this thread would do well to read the
>Wikipedia articles on "Random Number Generation" and
>"Hardware random number generator". Otherwise this is the
>ignorant talking to the ignorant.
>
>Donald Knuth famously said "Anyone who considers arithmetical methods
>of producing random digits is, of course, in a state of sin."

I liked the pre-computer method with a pencil, closed eyes,
and the random number tables in the CRC math handbook :-)

<snip>

>There are also electronic methods that produce random data
>based on quantum principles, described in the Wiki articles
>I listed. You'd think that would solve the problem.
>
>The problem is that physical methods tend to be slow. Data must
>be gathered, preferably from multiple sources, mixed together
>and whitened to produce an unbiased bit stream.
>
>Since 2011, Intel has included a RDRAND instruction in its processors. This
>claims to use a quantum electronic method to produce random numbers.
>It is still slow compared to pseudo random number generators.

ARM has relatively recently added similar instructions to the
ARMv8 architecture, however they delegate the actual generation
to the SoC vendor who includes the ARMv8 IP in their implementation,
and while I am familiar with how our implementation generates entropy,
I can't make that claim for other vendors using ARM IP.

On Thursday, December 1, 2022 at 10:28:34 AM UTC-5, Chris Buckley wrote:
> On 2022-12-01, Titus G <no...@nowhere.com> wrote:
> > On 1/12/22 04:34, Chris Buckley wrote:
> >> On 2022-11-30, WolfFan <akwo...@zoho.com> wrote:
> > snip
> >
> >> I've played many thousands of Spider games over the years. It's a
> >> very well-balanced optimization game - I enjoy it for the same
> >> reasons that I enjoyed the Civ games. It may take a while before you
> >> win your first game (especially if you don't allow takebacks), but
> >> you can push the optimizations to the point where you are winning
> >> over a third of your games (with no takebacks).
> >>
> >> Watch out for poor implementations - those floating around based on
> >> the Windows 7 free version have a remarkably poor RNG. I use
> >> SolSuite which I like a lot, but it's Windows only.
> >
> > I don't understand how there can be somewhat predictable, (or
> > "remarkably poor"), random numbers as I had always assumed that
> > unpredicability was a subset of randomness and that all random number
> > sequences were random. I am not a gamer but some decades ago wore out a
> > few packs of plastic playing cards playing Spider without takebacks.
> > As an afterthought. That number of games played sharpened the speed and
> > appearance of my shuffling skills which probably developed into a
> > routine. So in theory, does that mean that the sequence of cards in a
> > shuffled pack after many, many, iterations would become close to
> > predictable?
> > I use Linux and see that Spider is in the AisleRiot package. How would I
> > determine which Spider software had a better RNG? Thank you.
> Any modern (<20 years?) implementation from scratch on 32 bit machines
> is likely to be fine. It's easier to use the system RNG than roll
> your own and those are fine nowadays, assuming the implementation
> handles RNG seeds fine, does a reasonable approach to shuffling, and
> doesn't do ridiculous things with multi-deck solitaires like shuffle
> the decks separately.
>
> The main problem is that some solitaire implementations have been
> around for a long time and date back to 16-bit or even 8-bit days.
> Code space was a major concern and they had their own RNGs, which were
> often poor, yielding obvious patterns in the results which can be
> deadly in Spider. In addition, just adapting those implementations for
> 32 or 64 bit machines or new operating systems may introduce new bugs
> or misfeatures. This ranges from reasonably innocuous problems such
> as Microsoft Freecell which until very recently (Windows 11?) only had
> 32K distinct games, because the RNG seed only had 15 bits, to more
> serious problems.
>
> In particular with Spider, if you download the Windows 7 version of
> Spider to play on a Windows 10 machine (well-known compatibility
> packages exist) as I did, you may eventually notice that you're
> playing the same deal more than once. There was one particular deal
> that I eventually realized was showing up about once every 30 games in
> one time period! The problem seemed to come and go but was quite
> annoying.
>
> I would expect AisleRiot to be fine. Enough people have looked at
> its open-source code over the years.

There's a huge difference between 'Good enough for a Spider solitaire game',
and 'Good enough for secure communications'.

pt

Jack Bohn <jack.bohn64@gmail.com> writes:
>On Thursday, December 1, 2022 at 12:12:15 AM UTC-5, Titus G wrote:
>
>>I am not a gamer but some decades ago wore out a=20
>> few packs of plastic playing cards playing Spider without takebacks.=20
>> As an afterthought. That number of games played sharpened the speed and=
>=20
>> appearance of my shuffling skills which probably developed into a=20
>> routine. So in theory, does that mean that the sequence of cards in a=20
>> shuffled pack after many, many, iterations would become close to=20
>> predictable?=20
>
>As Marting Gardner would say paradoxically in his Mathematical Games column=
>s, shuffling is a skill that depends on the clumsiness of the shuffler.
>
>It can be shown that a perfect rifle shuffle -- divide the deck exactly in =
>half, alternate one card from each half -- will return the deck to its orig=
>inal or "seed" configuration after 14 repetitions.

I believe that is AKA the "Faro Shuffle" which is used by magicians
to move a specific card to a specific location in the deck with log2
shuffles (carefully selecting which card ends up on top when starting
the shuffle).

https://en.wikipedia.org/wiki/Faro_shuffle

Jack Bohn wrote:
> Titus G wrote:
>
>>I am not a gamer but some decades ago wore out a
>> few packs of plastic playing cards playing Spider without takebacks.
>> As an afterthought. That number of games played sharpened the speed and
>> appearance of my shuffling skills which probably developed into a
>> routine. So in theory, does that mean that the sequence of cards in a
>> shuffled pack after many, many, iterations would become close to
>> predictable?
>
> As Marting Gardner would say paradoxically in his Mathematical Games
> columns, shuffling is a skill that depends on the clumsiness of the
> shuffler.
>
> It can be shown that a perfect rifle shuffle -- divide the deck exactly
> in half, alternate one card from each half -- will return the deck to
> its original or "seed" configuration after 14 repetitions.
>
> The other shuffle, "overhand," where you hold the deck in one hand and
> shake the top card into the other repeatedly, would, if perfect, return
> to the original state if done twice. On first learning that "14 times"
> thing above, and the subsequent warning that something close to seven
> shuffles would be best to assure "randomness," I did rifle shuffles of
> a deck one card at a time and examined the deck's randomness after each.
> I also came up with a psuedorandom repeatable "realistic" clumsy overhand
> shuffle: 1. take the top card from the deck, note the number on it, place
> it at the bottom of the new deck. 2. take off that number of cards from
> the top, note the number of the bottom card, place them on the new deck.
> 3. repeat #2 until you can't. This mimics my clumsy shuffle with six or
> seven shakes of different clumps of cards, but takes quite a lot of
> repetitions to get the deck away from its original state. As suits don't
> enter into it, you can divide the deck into red and black with the cards
> otherwise random and watch how slowly it happens.

So, in a sense it seems an underhanded overhand plausibly enables a two
timer to trump opponents with ace in the hole. ROTFL.

Danke,

--
Don.......My cat's )\._.,--....,'``. https://crcomp.net/reviews.php
telltale tall tail /, _.. \ _\ (`._ ,. Walk humbly with thy God.
tells tall tales.. `._.-(,_..'--(,_..'`-.;.' Make 1984 fiction again.

scott@slp53.sl.home (Scott Lurndal) wrote in
news:sW3iL.2762$lzK9.2608@fx35.iad:

> Jack Bohn <jack.bohn64@gmail.com> writes:
>>On Thursday, December 1, 2022 at 12:12:15 AM UTC-5, Titus G
>>wrote:
>>
>>>I am not a gamer but some decades ago wore out a=20
>>> few packs of plastic playing cards playing Spider without
>>> takebacks.=20 As an afterthought. That number of games played
>>> sharpened the speed and=
>>=20
>>> appearance of my shuffling skills which probably developed
>>> into a=20 routine. So in theory, does that mean that the
>>> sequence of cards in a=20 shuffled pack after many, many,
>>> iterations would become close to=20 predictable?=20
>>
>>As Marting Gardner would say paradoxically in his Mathematical
>>Games column= s, shuffling is a skill that depends on the
>>clumsiness of the shuffler.
>>
>>It can be shown that a perfect rifle shuffle -- divide the deck
>>exactly in = half, alternate one card from each half -- will
>>return the deck to its orig= inal or "seed" configuration after
>>14 repetitions.
>
> I believe that is AKA the "Faro Shuffle" which is used by
> magicians to move a specific card to a specific location in the
> deck with log2 shuffles (carefully selecting which card ends up
> on top when starting the shuffle).
>
> https://en.wikipedia.org/wiki/Faro_shuffle

The number of magicians who can do that is very, very small. The
number who can simulate it with slight of hand is much, much
larger.

--
Terry Austin

"Terry Austin: like the polio vaccine, only with more asshole."
-- David Bilek

Jesus forgives sinners, not criminals.

On Thu, 1 Dec 2022 18:12:11 +1300, Titus G <noone@nowhere.com> wrote:

>On 1/12/22 04:34, Chris Buckley wrote:
>> On 2022-11-30, WolfFan <akwolffan@zoho.com> wrote:
>snip
>
>> I've played many thousands of Spider games over the years. It's a
>> very well-balanced optimization game - I enjoy it for the same
>> reasons that I enjoyed the Civ games. It may take a while before you
>> win your first game (especially if you don't allow takebacks), but
>> you can push the optimizations to the point where you are winning
>> over a third of your games (with no takebacks).
>>
>> Watch out for poor implementations - those floating around based on
>> the Windows 7 free version have a remarkably poor RNG. I use
>> SolSuite which I like a lot, but it's Windows only.
>
>I don't understand how there can be somewhat predictable, (or
>"remarkably poor"), random numbers as I had always assumed that
>unpredicability was a subset of randomness and that all random number
>sequences were random. I am not a gamer but some decades ago wore out a
>few packs of plastic playing cards playing Spider without takebacks.
>As an afterthought. That number of games played sharpened the speed and
>appearance of my shuffling skills which probably developed into a
>routine. So in theory, does that mean that the sequence of cards in a
>shuffled pack after many, many, iterations would become close to
>predictable?
>I use Linux and see that Spider is in the AisleRiot package. How would I
>determine which Spider software had a better RNG? Thank you.

The wargame /War in Europe/ has had two Computer versions. This is
relevant, by the way.

The first was a DOS version. One of the flaws the second (Windows)
version corrected was that the first saved the last random value from
the previous turn and used it as the seed for the next. This may not
sound too bad, but it meant that /every single result/ was the same if
the turn were replayed.

This was certainly considered a defect in the RNG -- not the generator
itself, but in how it was used.

This made it possible for a player to play it enough to memorize the
results and then vary the order in which the rolls were applied to
optimize his results, leading to "local rules" that actions requiring
a random number would be taken in a defined order. For example, on the
Eastern Front, all combats might have to be resolved in order from
North to South. No shifting the good rolls to one point in the line
and busting it open.

This had its advantages for research: I used this in a solitaire game
to actually /capture Moscow/ and then explore what a game without
Russia would look like. I ended up with two opposing lines of units
stretching from Morocco to Algeria, and going nowhere. IOW, a draw.
--
"In this connexion, unquestionably the most significant
development was the disintegration, under Christian
influence, of classical conceptions of the family and
of family right."

On 1 Dec 2022 15:28:29 GMT, Chris Buckley <alan@sabir.com> wrote:

>On 2022-12-01, Titus G <noone@nowhere.com> wrote:
>> On 1/12/22 04:34, Chris Buckley wrote:
>>> On 2022-11-30, WolfFan <akwolffan@zoho.com> wrote:
>> snip
>>
>>> I've played many thousands of Spider games over the years. It's a
>>> very well-balanced optimization game - I enjoy it for the same
>>> reasons that I enjoyed the Civ games. It may take a while before you
>>> win your first game (especially if you don't allow takebacks), but
>>> you can push the optimizations to the point where you are winning
>>> over a third of your games (with no takebacks).
>>>
>>> Watch out for poor implementations - those floating around based on
>>> the Windows 7 free version have a remarkably poor RNG. I use
>>> SolSuite which I like a lot, but it's Windows only.
>>
>> I don't understand how there can be somewhat predictable, (or
>> "remarkably poor"), random numbers as I had always assumed that
>> unpredicability was a subset of randomness and that all random number
>> sequences were random. I am not a gamer but some decades ago wore out a
>> few packs of plastic playing cards playing Spider without takebacks.
>> As an afterthought. That number of games played sharpened the speed and
>> appearance of my shuffling skills which probably developed into a
>> routine. So in theory, does that mean that the sequence of cards in a
>> shuffled pack after many, many, iterations would become close to
>> predictable?
>> I use Linux and see that Spider is in the AisleRiot package. How would I
>> determine which Spider software had a better RNG? Thank you.
>
>Any modern (<20 years?) implementation from scratch on 32 bit machines
>is likely to be fine. It's easier to use the system RNG than roll
>your own and those are fine nowadays, assuming the implementation
>handles RNG seeds fine, does a reasonable approach to shuffling, and
>doesn't do ridiculous things with multi-deck solitaires like shuffle
>the decks separately.
>
>The main problem is that some solitaire implementations have been
>around for a long time and date back to 16-bit or even 8-bit days.
>Code space was a major concern and they had their own RNGs, which were
>often poor, yielding obvious patterns in the results which can be
>deadly in Spider. In addition, just adapting those implementations for
>32 or 64 bit machines or new operating systems may introduce new bugs
>or misfeatures. This ranges from reasonably innocuous problems such
>as Microsoft Freecell which until very recently (Windows 11?) only had
>32K distinct games, because the RNG seed only had 15 bits, to more
>serious problems.
>
>In particular with Spider, if you download the Windows 7 version of
>Spider to play on a Windows 10 machine (well-known compatibility
>packages exist) as I did, you may eventually notice that you're
>playing the same deal more than once. There was one particular deal
>that I eventually realized was showing up about once every 30 games in
>one time period! The problem seemed to come and go but was quite
>annoying.

I ran into something similar when trying to have my Squeezebox select
the next album randomly. I finally conceptualized it this way:

1) It was using a bucket sort.
2) It was selecting each bucket using a simple RNG that started at the
same bucket each time and then, of course, continued on in the same
order. It disguised this by keeping track of which albums in that
bucket had already been played and playing the first that had not.
3) Every 80 or 90 albums, it stopped working because, despite not
running Windows or DOS and not running on segmented hardware, it still
leaked memory.
4) When it re-started, it restarted at the same bucket and, since the
list of what had already been played was gone, at the same album. So
80-90 of my 726 Albums played every time and the rest were never
heard.

Note: for a while, I randomized the list myself (that is, I used the
ODS randomize function on a copy of the list), each time a made a new
copy after working through the old copy. But, eventually, I realized
that I really wanted a fixed list so I did the work necessary to
produce one that I could live with. It isn't random but the various
musical Genres (Humor, Rock, Harpsichord ... the list goes on) were
mixed together to produce variety.

>I would expect AisleRiot to be fine. Enough people have looked at
>its open-source code over the years.
--
"In this connexion, unquestionably the most significant
development was the disintegration, under Christian
influence, of classical conceptions of the family and
of family right."

Titus G <noone@nowhere.com> writes:

> On 1/12/22 04:34, Chris Buckley wrote:
>> On 2022-11-30, WolfFan <akwolffan@zoho.com> wrote:
> snip
>
>> I've played many thousands of Spider games over the years. It's a
>> very well-balanced optimization game - I enjoy it for the same
>> reasons that I enjoyed the Civ games. It may take a while before you
>> win your first game (especially if you don't allow takebacks), but
>> you can push the optimizations to the point where you are winning
>> over a third of your games (with no takebacks).
>>
>> Watch out for poor implementations - those floating around based on
>> the Windows 7 free version have a remarkably poor RNG. I use
>> SolSuite which I like a lot, but it's Windows only.
>
> I don't understand how there can be somewhat predictable, (or
> "remarkably poor"), random numbers as I had always assumed that
> unpredicability was a subset of randomness and that all random number
> sequences were random. I am not a gamer but some decades ago wore out a
> few packs of plastic playing cards playing Spider without takebacks.
> As an afterthought. That number of games played sharpened the speed and
> appearance of my shuffling skills which probably developed into a
> routine. So in theory, does that mean that the sequence of cards in a
> shuffled pack after many, many, iterations would become close to
> predictable?
> I use Linux and see that Spider is in the AisleRiot package. How would I
> determine which Spider software had a better RNG? Thank you.

Almost no computer random number generators are random -- when people
are speaking more carefully, they call them "pseudorandom". From a
given "seed", a random number generator produces, completely
deterministically, a result and a new seed (in some very early RNGs, the
result was also the new seed).

The original UNIX RNG was notorious for generating alternate even and
odd numbers -- as many, many people discovered when trying to write
trivial coin-flipping games. It was *terrible*.

I would expect Spider under Linux to use the C library RNG, which is just
fine for nearly any purpose -- generating random numbers for games or
simulations, for instance. On the other hand, it isn't nearly good
enough to generate the keys used for secure communications.

On Thursday, December 1, 2022 at 8:00:13 AM UTC-7, pete...@gmail.com wrote:

> Donald Knuth famously said "Anyone who considers arithmetical methods
> of producing random digits is, of course, in a state of sin."

If he did say that, he was quoting John von Neumann.

John Savard

On Thursday, December 1, 2022 at 8:00:13 AM UTC-7, pete...@gmail.com wrote:

> John and others in this thread would do well to read the
> Wikipedia articles on "Random Number Generation" and
> "Hardware random number generator". Otherwise this is the
> ignorant talking to the ignorant.

I read "Random Number Generators" by Birger Jansson. Does that count?

Of course, that book was written long before the Pentium III came out. It
had a picture of a hardware random number generator on the cover, though.

A set of three icosahedral dice, with the digits 0 through 9 repeated twice
on them, of three different colors.

They were made by the Japanese Bureau of Standards... *since not only
was this book written before the Pentium III came out, it was written before
Dungeons and Dragons was invented*.

John Savard

On Thursday, December 1, 2022 at 3:25:13 PM UTC-7, Quadibloc wrote:
> On Thursday, December 1, 2022 at 8:00:13 AM UTC-7, pete...@gmail.com wrote:
>
> > Donald Knuth famously said "Anyone who considers arithmetical methods
> > of producing random digits is, of course, in a state of sin."

> If he did say that, he was quoting John von Neumann.

The section "Random number generation" by Jack Moshman
from Volume 2 of Mathematical Methods for Digital
Computers provides a bit more of the quote:

"anyone who consideres arithmetical methods of producing
random digits is, of course, in a state of sin. For, as been
pointed out several times, there is no such thing as a random
number - there are only methods to produce random numbers,
and a strict arithmetic procedure of course is not such a
method."

John Savard

On 2/12/22 05:12, Jibini Kula Tumbili Kujisalimisha wrote:
> scott@slp53.sl.home (Scott Lurndal) wrote in
> news:sW3iL.2762$lzK9.2608@fx35.iad:
>
>> Jack Bohn <jack.bohn64@gmail.com> writes:
>>> On Thursday, December 1, 2022 at 12:12:15 AM UTC-5, Titus G
>>> wrote:
>>>
>>>> I am not a gamer but some decades ago wore out a=20
>>>> few packs of plastic playing cards playing Spider without
>>>> takebacks.=20 As an afterthought. That number of games played
>>>> sharpened the speed and=
>>> =20
>>>> appearance of my shuffling skills which probably developed
>>>> into a=20 routine. So in theory, does that mean that the
>>>> sequence of cards in a=20 shuffled pack after many, many,
>>>> iterations would become close to=20 predictable?=20

I am guessing that the insertion of the "=20" is a different software
interpretation of a system character perhaps where I have reworded
something or corrected spelling?

>>>
>>> As Marting Gardner would say paradoxically in his Mathematical
>>> Games column= s, shuffling is a skill that depends on the
>>> clumsiness of the shuffler.
>>>
>>> It can be shown that a perfect rifle shuffle -- divide the deck
>>> exactly in = half, alternate one card from each half -- will
>>> return the deck to its orig= inal or "seed" configuration after
>>> 14 repetitions.
>>
>> I believe that is AKA the "Faro Shuffle" which is used by
>> magicians to move a specific card to a specific location in the
>> deck with log2 shuffles (carefully selecting which card ends up
>> on top when starting the shuffle).
>>
>> https://en.wikipedia.org/wiki/Faro_shuffle
>
> The number of magicians who can do that is very, very small. The
> number who can simulate it with slight of hand is much, much
> larger.

The fact that it is possible without slight of hand is fascinating to
me. I doubt whether I could have even split a deck exactly in half!

On 2/12/22 04:28, Chris Buckley wrote:
> On 2022-12-01, Titus G <noone@nowhere.com> wrote:
snip
>> I use Linux and see that Spider is in the AisleRiot package. How would I
>> determine which Spider software had a better RNG? Thank you.

snip

Thank you for that interesting detail.

> I would expect AisleRiot to be fine. Enough people have looked at
> its open-source code over the years.
>
> Chris

If I become addicted again, is there a Spider's Anonymous?

On 2/12/22 04:40, pete...@gmail.com wrote:
snip

Regarding random number generation.
> There's a huge difference between 'Good enough for a Spider solitaire game',
> and 'Good enough for secure communications'.
>

My computing 'career' was ended whilst inter-computer communications
were physical (and I still have arduinos connected to Windows XP :-)) so
that explains my ignorance with regard to standards of RNGs which I have
encountered occasionally but little understood. The only games I have
written were simple class exercises in Pascal using an unknown to me
RNG. This thread has been interesting, thank you and others.

On 2/12/22 04:34, Scott Lurndal wrote:
> Titus G <noone@nowhere.com> writes:
>> On 1/12/22 04:34, Chris Buckley wrote:
>>> On 2022-11-30, WolfFan <akwolffan@zoho.com> wrote:
>> snip
>>
>>> I've played many thousands of Spider games over the years. It's a
>>> very well-balanced optimization game - I enjoy it for the same
>>> reasons that I enjoyed the Civ games. It may take a while before you
>>> win your first game (especially if you don't allow takebacks), but
>>> you can push the optimizations to the point where you are winning
>>> over a third of your games (with no takebacks).
>>>
>>> Watch out for poor implementations - those floating around based on
>>> the Windows 7 free version have a remarkably poor RNG. I use
>>> SolSuite which I like a lot, but it's Windows only.
>>
>> I don't understand how there can be somewhat predictable, (or
>> "remarkably poor"), random numbers as I had always assumed that
>> unpredicability was a subset of randomness and that all random number
>> sequences were random.
>
> Generating a truely random number is rather difficult. Most random
> number generators are "Pseudo Random Number Generators" (PRNG) which
> have repeatable sequences, given the same starting point. So
> a lot of care goes into selecting a random starting point (seed), which
> is very very hard.
>
> There was a paper back in the mid 90's from an SGI engineer where
> they used a video camera on a Indy to periodcally snap an image
> of a working lava-lamp and generated the seed from that image.
>
> Today, modern processors have a "true" random number generator
> that builds an entropy pool from which random seed number can
> be selected. But it takes a long time to generate sufficient
> entropy to generate a large number of random values, so they
> use the true number as a seed to a PRNG such as:
>
> https://en.wikipedia.org/wiki/Mersenne_Twister

I was enjoying your post until I followed the link. I was interested to
read its use by Linux and Free Pascal but then volunteered for a full
frontal lobotomy to ease my tired brain.

On 02/12/2022 02:38, Scott Lurndal wrote:
> "pete...@gmail.com" <petertrei@gmail.com> writes:
>> On Thursday, December 1, 2022 at 8:13:31 AM UTC-5, Quadibloc wrote:
>>> On Thursday, December 1, 2022 at 4:41:06 AM UTC-7, Robert Carnegie wrote:
>>>
>>>> A sequence of "random numbers" from a
>>>> pseudo-random-number formula ought to be
>>>> unpredictable by examining the output, even when
>>>> you know what the formula is, and a good one is
>>>> unpredictable.
>>> In that case, "good" pseudorandom number generators
>>> are hardly _ever_ used.
>>>> However, if the starting condition
>>>> or "seed" is known, then the output is always the
>>>> same. Randomness is achieved by using a random
>>>> "seed".
>>>
>>>> I had a "home computer", Sinclair ZX Spectrum
>>>> I think is where I did this, with the BASIC language
>>>> and 65535 random decimal fraction numbers,
>>>> since those were the possible values of the seed.
>>>> I used random numbers to plot points on the screen,
>>>> x coordinate and y coordinate, and I got neat stripes
>>>> slightly angled from vertical. So that wasn't as "random"
>>>> as it could have been.
>>> Except for having only 65,535 possible seed values,
>>> instead of, say, 4,294,967,296 possible seed values,
>>> that is the sort of thing that is normal and standard for
>>> pseudorandom number generators; typically, linear
>>> congruential RNGs are used, not _even_ MacLaren -
>>> Marsaglia, never mind anything actually cryptosecure,
>>> which is what it would take for a sequence to be
>>> unpredictable by examining the output when the
>>> algorithm is known.
>>>
>>> John Savard
>>
>> John and others in this thread would do well to read the
>> Wikipedia articles on "Random Number Generation" and
>> "Hardware random number generator". Otherwise this is the
>> ignorant talking to the ignorant.
>>
>> Donald Knuth famously said "Anyone who considers arithmetical methods
>> of producing random digits is, of course, in a state of sin."
>
> I liked the pre-computer method with a pencil, closed eyes,
> and the random number tables in the CRC math handbook :-)
>
> <snip>
[SNIP]
Amusingly, way back when such things were common, it was shown - using
this method - that Unicorns are most active at 4:00AM. :-)

Cheers,
Gary B-)

On Fri, 2 Dec 2022 16:21:25 +1300, Titus G <noone@nowhere.com> wrote:

>On 2/12/22 05:12, Jibini Kula Tumbili Kujisalimisha wrote:
>> scott@slp53.sl.home (Scott Lurndal) wrote in
>> news:sW3iL.2762$lzK9.2608@fx35.iad:
>>
>>> Jack Bohn <jack.bohn64@gmail.com> writes:
>>>> On Thursday, December 1, 2022 at 12:12:15 AM UTC-5, Titus G
>>>> wrote:
>>>>
>>>>> I am not a gamer but some decades ago wore out a=20
>>>>> few packs of plastic playing cards playing Spider without
>>>>> takebacks.=20 As an afterthought. That number of games played
>>>>> sharpened the speed and=
>>>> =20
>>>>> appearance of my shuffling skills which probably developed
>>>>> into a=20 routine. So in theory, does that mean that the
>>>>> sequence of cards in a=20 shuffled pack after many, many,
>>>>> iterations would become close to=20 predictable?=20
>
>I am guessing that the insertion of the "=20" is a different software
>interpretation of a system character perhaps where I have reworded
>something or corrected spelling?

Bing has problems finding it.

[https://ascii.cl/] suggests that, it the 20 is decimal and ASCII, it
is "DC4". If it is hex and ASCII it is, of course, a space.

However, it is probably something else altogether.

<snippo>
--
"In this connexion, unquestionably the most significant
development was the disintegration, under Christian
influence, of classical conceptions of the family and
of family right."

Titus G <noone@nowhere.com> writes:

> On 2/12/22 05:12, Jibini Kula Tumbili Kujisalimisha wrote:
>> scott@slp53.sl.home (Scott Lurndal) wrote in
>> news:sW3iL.2762$lzK9.2608@fx35.iad:
>>
>>> Jack Bohn <jack.bohn64@gmail.com> writes:
>>>> On Thursday, December 1, 2022 at 12:12:15 AM UTC-5, Titus G
>>>> wrote:
>>>>
>>>>> I am not a gamer but some decades ago wore out a=20
>>>>> few packs of plastic playing cards playing Spider without
>>>>> takebacks.=20 As an afterthought. That number of games played
>>>>> sharpened the speed and=
>>>> =20
>>>>> appearance of my shuffling skills which probably developed
>>>>> into a=20 routine. So in theory, does that mean that the
>>>>> sequence of cards in a=20 shuffled pack after many, many,
>>>>> iterations would become close to=20 predictable?=20
>
> I am guessing that the insertion of the "=20" is a different software
> interpretation of a system character perhaps where I have reworded
> something or corrected spelling?

https://en.wikipedia.org/wiki/Quoted-printable