I hadn't played backgammon against a human since eons ago because of
where I live but I never thought I would get tired of playing backgammon
even if against bots.
Yet, now I can't even remember how long it has been since I quit playing against bots also, simply because it became too predictably repetitive and boring. And, yes, when the challenge also wanes for the unlucky few people who can actually consistently beat the bots (as unbelievable as it mey be
for you flock).
The two bots that were consirered the last best, have not been improved on for the past six years. Even any talks about an Alpha-BG stopped over a year ago (or more?).
There was a spark of hope that an Alpha-Hypestgammon and then an Alpha- Hypergammon could be developed but that didn't go anywhere either.
Every so often I take a loot see if there is anything interesting dicussed here
but the last time I marke a post with an intent to reply to it was in May and then I thought it would be waste of time since it would quickly degrade to stupid smartass one-liners, as all/most other discussions had done in the past, since the mentally ill gamblers (even if claimedly scientists) can't do any better to pursue them into any fruitful conclusions.
I don't know if there is more interest in other forums, if people still buy BG
bots, books, etc. but this group has long been reduced to a private mail-list for less than a handful, obcessive senile idiots to dicuss positions (which is
fine), to then defer to bots (which is not good) to decide the "correct" moves
when simple logic dictates that bots based on arbitrary, fartass algorithms are biased by an undeterminable magnitute.
It's all very sad... :(
MK
The inevitable end of backgammon? Not at all, R.B.
Take a deep breath and curse me right now. Trust
me, it will feel good!
I hadn't played backgammon against a human since eons ago because of
where I live but I never thought I would get tired of playing backgammon
even if against bots.
Yet, now I can't even remember how long it has been since I quit playing against bots also, simply because it became too predictably repetitive and boring. And, yes, when the challenge also wanes for the unlucky few people who can actually consistently beat the bots (as unbelievable as it mey be
for you flock).
The two bots that were consirered the last best, have not been improved on for the past six years. Even any talks about an Alpha-BG stopped over a year ago (or more?).
There was a spark of hope that an Alpha-Hypestgammon and then an Alpha- Hypergammon could be developed but that didn't go anywhere either.
Every so often I take a loot see if there is anything interesting dicussed here
but the last time I marke a post with an intent to reply to it was in May and then I thought it would be waste of time since it would quickly degrade to stupid smartass one-liners, as all/most other discussions had done in the past, since the mentally ill gamblers (even if claimedly scientists) can't do any better to pursue them into any fruitful conclusions.
I don't know if there is more interest in other forums, if people still buy BG
bots, books, etc. but this group has long been reduced to a private mail-list for less than a handful, obcessive senile idiots to dicuss positions (which is
fine), to then defer to bots (which is not good) to decide the "correct" moves
when simple logic dictates that bots based on arbitrary, fartass algorithms are biased by an undeterminable magnitute.
It's all very sad... :(
MK
How about a vacation to Istanbul and play a bg tournament like this one.. https://www.youtube.com/watch?v=jFUu8KTvUno&t=339s&ab_channel=GezenAdam
I hadn't played backgammon against a human since eons ago because of
where I live but I never thought I would get tired of playing backgammon
even if against bots.
Yet, now I can't even remember how long it has been since I quit playing against bots also, simply because it became too predictably repetitive and boring. And, yes, when the challenge also wanes for the unlucky few people who can actually consistently beat the bots (as unbelievable as it mey be
for you flock).
The two bots that were consirered the last best, have not been improved on for the past six years. Even any talks about an Alpha-BG stopped over a year ago (or more?).
There was a spark of hope that an Alpha-Hypestgammon and then an Alpha- Hypergammon could be developed but that didn't go anywhere either.
Every so often I take a loot see if there is anything interesting dicussed here
but the last time I marke a post with an intent to reply to it was in May and then I thought it would be waste of time since it would quickly degrade to stupid smartass one-liners, as all/most other discussions had done in the past, since the mentally ill gamblers (even if claimedly scientists) can't do any better to pursue them into any fruitful conclusions.
I don't know if there is more interest in other forums, if people still buy BG
bots, books, etc. but this group has long been reduced to a private mail-list for less than a handful, obcessive senile idiots to dicuss positions (which is
fine), to then defer to bots (which is not good) to decide the "correct" moves
when simple logic dictates that bots based on arbitrary, fartass algorithms are biased by an undeterminable magnitute.
It's all very sad... :(
MK
On September 27, 2021 at 12:02:46 AM UTC-6, tetraHydro saved my life wrote:
How about a vacation to Istanbul and play a bg tournament like this one.. https://www.youtube.com/watch?v=jFUu8KTvUno&t=339s&ab_channel=GezenAdamFor the ones who can't read Turkish, it's an almost 6 years old video
of a promotional bg tournament for a shopping center. About $10's
worth of shopping got people free entry, with some prizes for the
winners. Too noisy and wouldn't be challenging at all to play against
a hord of "shopper-level" players from the street.
Also, this wasn't "Turkey" then and is less so now. Sure there a few
dozen luxury hotels, resorts, golf courses, etc. but it has become a
mainly islamo-fascist, narco-state dictatorship of 85 million dirt eating, insect brained moslems... :(
I'm sure rich tourists, more and more from arab emirates, etc. still go there to vacation, well fenced off from the masses.
Also probably old woman and gay men from european countries may
be still going there to get fucked by horny moslem locals, who can't
have normal human sex with women, and tired of fucking donkeys,
sheep, dogs, etc.
People with average means and travel/vacation interests would have
to be out of their minds, (or greedy and brave enough to risk getting
more for less), to go Turkey. Personally, I wouldn't go for a million dollars. Not worth it. There are plenty of better places to go in the world (of course, after covid will be over).
MK
greedygammon with friends I knew from FIBS back in the day but we
lost touch, and i dont know how how reliable the program is but if
there was interest in it i would fix any bugs. I still enjoy playing the bot.
If you don't like to see people obsessively discussing
positions, just quit and remain silent.
Take your depressive ass elsewhere.
Everyone's trying to carry on through these times as best
as they can.
If you or anyone else interested wants to start with an easier variant,
try Alpha-Hypestgammon and Alpha-Hypergammon first. Even a PC
should have enough CPU power to handle those.
Someone should have the courage to do it. Yes, at first, your xg-centric universe will turn upside down but you probably will survive to enjoy it.
On September 27, 2021 at 10:58:51 AM UTC-6, tetraHydro saved my life wrote:
greedygammon with friends I knew from FIBS back in the day but weSince you are capable of developing BG software, why don't you just
lost touch, and i dont know how how reliable the program is but if
there was interest in it i would fix any bugs. I still enjoy playing the bot.
forget about the obsolete stuff and tackle the Alpha-Gammon??
Tesauro's excuse was lacking CPU power back then, which we have
more than enough of today. TD-Gammon v1 was an "Alpha-Gammon"
but limited to 1-point games. Just start from there and let the bot
make random cube decisions as well and tabulate the results. Simple.
If you or anyone else interested wants to start with an easier variant,
try Alpha-Hypestgammon and Alpha-Hypergammon first. Even a PC
should have enough CPU power to handle those.
Someone should have the courage to do it. Yes, at first, your xg-centric universe will turn upside down but you probably will survive to enjoy it.
MK
On 9/28/2021 4:09 AM, MK wrote:
Someone should have the courage to do it. Yes, at first, your xg-centric universe will turn upside down but you probably will survive to enjoy it.
Alpha-Hypestgammon isn't going to turn the XG-centric universe upside
down because I solved Hypestgammon and there were no surprises.
Isn't "alpha-gammon" just another AI bg bot like xg and gnubg?
I'm not that level of programmer as these AI people.
I did my bg game because I couldn't find any other interface that
I liked, and i had plenty of help attaching the gnubg engine to it.
On September 28, 2021 at 4:57:50 PM UTC-6, tetraHydro saved my life wrote:
Isn't "alpha-gammon" just another AI bg bot like xg and gnubg?My understanding/definition of Alpha-XYZ is a AI bot untainted
by human bias. In other words, just tell the bot the rules of the
game and let it figure out what are the best decision statistically
by trial and error, over a sufficient number of trials.
I'm not that level of programmer as these AI people.Don't over estimate them or underestimate yourself.
I did my bg game because I couldn't find any other interface thatOkay, fine. Just modify what you are doing slightly. If you can get
I liked, and i had plenty of help attaching the gnubg engine to it.
all possible legal moves from gnubg and pick one randomly yourself
or let it return a random move somehow, maybe by tricking it to think
that the winning chances are zero for the given position?? For cube decisions, don't even ask gnubg, just randomly decide to double or
not and to take or not. If you can just tabulate the results, you'll have it.
For Hypestgammon, no checker play decision is needed since they
are all "forced moves" so to speak. And just make random cube
decisions. Nothing could be easier.
Now, for Hypergammon, supposedly Sconyer has solved it 30 years
ago (according to "their" understanding of what "solved" means,
which I don't agree with but let's leave this aside foe now), with 32,000,000 possible moves (the bullshit is all over the place on the Internet, even in the Wiki). Gnubg plays Hypergammon. So, query it
for possible moves in a given position and pick one randomly. And
again, just make random cube decisions before every roll for each
side. If you can tabulate 32,000,000 moves times cube decisions
in a given length match, you will have made it.
For now, let's leave the development of regular BG Alpha-Gammon
bot to the biggest scumbags of world-class bot developers with
the help of world-class book publishing sick gamblers...
MK
I can accept that you may have misundestood from the very beginning
what I was expecting of you to do, i.e. compare Alpha-Hypestgammon
derived equities to your Hypestgammon calculated equities (that I was questioning, just the same as for Hypergammon and regular Backgammon).
On 9/28/2021 7:58 PM, MK wrote:
I can accept that you may have misundestood from the very beginning
what I was expecting of you to do, i.e. compare Alpha-Hypestgammon
derived equities to your Hypestgammon calculated equities (that I was
questioning, just the same as for Hypergammon and regular Backgammon).
Oh, I see. That's fair. Unfortunately, my colleague was doing this as
a "lockdown project" and when the lockdown was eased, he abandoned it.
I'm a bit disappointed about that and I'm hoping he'll get back to it someday.
By the way, AlphaZero is not exactly "untainted by human bias." It is..... reproduce AlphaZero Chess, there was a long process of trial and error
before they were able to get a chess engine that could beat Stockfish.
I just can't understand why human tweaking is needed... Is it to expedite
the training/finalizing of the bot? Especially in order to beat a specific adversary??
On 9/30/2021 3:50 AM, MK wrote:
I just can't understand why human tweaking is needed... Is it to expedite
the training/finalizing of the bot? Especially in order to beat a specific >> adversary??
The problem is much more basic. If you don't do any tweaking, the
training process can stall, with the program failing to get any better.
It might even get worse than it was before.
On October 1, 2021 at 5:36:18 AM UTC-6, Tim Chow wrote:
On 9/30/2021 3:50 AM, MK wrote:
I just can't understand why human tweaking is needed... Is it to expedite >>> the training/finalizing of the bot? Especially in order to beat a specific >>> adversary??
The problem is much more basic. If you don't do any tweaking, the
training process can stall, with the program failing to get any better.
It might even get worse than it was before.
"Can stall"? Did it ever stall and and is that how you know? What if
it doesn't stall?
I really can't understand anything at all from your hypothetical and
vague statement. Would you can to expand, if you can? Perhaps
even giving some specific example cases?
On 10/1/2021 6:06 PM, MK wrote:
I really can't understand anything at all from your hypothetical and
vague statement. Would you can to expand, if you can? Perhaps
even giving some specific example cases?
I don't know what you mean by specific example cases.
Of course I have observed stalling. Stalling is normal. It's *not* stalling that is hard, and that requires significant effort. If it doesn't stall, you break out the champagne. If you've tried to do any nontrivial deep
learning you would already know all this.
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 296 |
Nodes: | 16 (2 / 14) |
Uptime: | 33:46:24 |
Calls: | 6,648 |
Calls today: | 3 |
Files: | 12,193 |
Messages: | 5,328,723 |