A time machine that could actually deliver and tell us who would win mythical match ups... Is it more of a reality than we think?!

[billeau2]

It is fun ! - What other fun can one have with historical knowledge other than to play 'what if' ?

What if the Fench had the good sense not to behead Marie Antoinette but instead put a kitten on her lap and sent her home to Vienna. Would Austria have attacked Paris; would have they joined a coalition with Great Britain; would Berlin follow Vienna's lead and leave France alone; would Austria and Prussia allow Russia to cross its territory to assist Britain to attack France?

Would not killing the pointless, clueless little girl have resulted in Napoleon Bonaparte never rising to power? No international coalition attacking France; no collapse of the French Repunlic; no Napoleon.

See fun - imgaine if there was no 'color line' ? We would never have heard of Jack Johnson.

P.S. Unless you are a hater and want to demean fighters from the past; you can do that too. But that's not fun, just mean.

YES!! And then to construct a time line... and imagine there might be such a place. It boggles the mind pleasantly. As opposed to trying to conceptualize infinity and where things end... That throws my head into a geckoring spazz session. Its almost like there is a filter that says "Stop!"

[billeau2]

I think much of the direction AIs going in now is via machine learning - you kinda give the machine only the most basic of starting parameters and let it in effect work out for itself which information best solves the problems:



In this case it'd involve setting it lose on a database of all the fight footage ever recorded as well as databases of other fights, fighter stats - even possibly personal data about fighters - and so on and allowing it to seek out the patterns itself. I remain sceptical that you'd ever get total accuracy just cos of random unknowable factors or the general chaos of the universe but the results this method has got in other fields is pretty remarkable.

YES! because it is easy for a very fast computer to determine things that can be determined via probability which gives essentially a series of choices (3 possible positions I think) as opposed to a linear binary system.

[billeau2]

It's an incredibly fascinating subject (at least in my opinion)!

Now let's say, as you suggest, that we feed everything that has ever happened related to boxing, into this self-learning super-computer. The entire BoxRec database (of course!), every fight we have footage of, every article ever written anywhere about boxing/boxers, etc. These are not in any way related to opinion - they are FACTUAL things that have actually happened (fights that have taken place, articles that have been written, etc.).

With all this information, I wonder what our computer will make of, for example, John L. Sullivan? It will not have been able to watch any fight footage of him - but it will have gone through hundreds and hundreds af contemporary articles, where reporters were waxing eloquently about his greatness, like:

“He can strike out with either right or left and knock a man down with as much ease and grace as an accomplished lady can gently and languidly open an opera fan”.

"The arm straightens out and the blow is with a suddenness which seems paralysing to the spectators, to say nothing of the man in front of him".

“The superiority of Sullivan lies in his extraordinary nervous force and altogether incomparable skill as a boxer. In what does this extraordinary skill consist? In hitting as straight and almost as rapid as light, in the variety and rediness of his blows, in standing firmly on his feet and driving his whole weight and nervous force at the end of his fist- a very rare and high quality in a boxer, in movements as quick and purposeful as the leap of a lion”.

“He can strike more heavy blows in ten seconds than any other man in a minute and watches with self possession and calculation".

The computer could be excused for thinking, that this may possibly be the greatest fighter who ever lived! So what do we do? Do we tell the computer, that these descriptions should be taken with a grain of salt - since they were made by oldtimers, who had seen nothing yet? But if we do that, we interject our own personal OPINION - which kind of defeats the whole purpose of computer self-learning.

I would guess that a computer would figure out things like: What is the probability of Marciano being able to physically defeat everyman in a drinking establishment circa 1956...? and account for hyperbole by comparing a probability like this to "marciano could beat ten men in an alley with one hand tied behind his back." What is hyperbole? essentially it is when a claims probability is claimed a possibility in excess of such probability. What we call things like irony and hyperbole are in essence things that are used that way because they are improbable... I believe a computer could learn that if it was not binery.

[billeau2]

Just to clarify something: I am not working on the model due out next week lol. And Quantum computers are in their infancy. Yes this would be down the road a bit. In the meantime we will just have to depend on Queeny to fill in the blanks I suppose... What is the probability he will post something that will make sense? might break a quantum computer!

[Citizen Koba]

It's an incredibly fascinating subject (at least in my opinion)!

Now let's say, as you suggest, that we feed everything that has ever happened related to boxing, into this self-learning super-computer. The entire BoxRec database (of course!), every fight we have footage of, every article ever written anywhere about boxing/boxers, etc. These are not in any way related to opinion - they are FACTUAL things that have actually happened (fights that have taken place, articles that have been written, etc.).

With all this information, I wonder what our computer will make of, for example, John L. Sullivan? It will not have been able to watch any fight footage of him - but it will have gone through hundreds and hundreds af contemporary articles, where reporters were waxing eloquently about his greatness, like:

“He can strike out with either right or left and knock a man down with as much ease and grace as an accomplished lady can gently and languidly open an opera fan”.

"The arm straightens out and the blow is with a suddenness which seems paralysing to the spectators, to say nothing of the man in front of him".

“The superiority of Sullivan lies in his extraordinary nervous force and altogether incomparable skill as a boxer. In what does this extraordinary skill consist? In hitting as straight and almost as rapid as light, in the variety and rediness of his blows, in standing firmly on his feet and driving his whole weight and nervous force at the end of his fist- a very rare and high quality in a boxer, in movements as quick and purposeful as the leap of a lion”.

“He can strike more heavy blows in ten seconds than any other man in a minute and watches with self possession and calculation".

The computer could be excused for thinking, that this may possibly be the greatest fighter who ever lived! So what do we do? Do we tell the computer, that these descriptions should be taken with a grain of salt - since they were made by oldtimers, who had seen nothing yet? But if we do that, we interject our own personal OPINION - which kind of defeats the whole purpose of computer self-learning.

Well like I say I have my doubts myself, but bear in mind the computer / program would also have access to the hundreds - thousands probably - of other hyperbolic articles from the time, replete with their vivid and colourful descriptions which have so sadly devolved into todays formulaic repeititions. It would be able to cross reference and compare to the fight outcomes, so it wouldn't need telling the language doesn't always match the outcomes - it would learn itself to take contemporaneous reports with 'a pinch of salt'. Probably to an extent it would be able to work backwards from the present day, even learning how the phraseology has changed over time and the differing degrees to which hyperbole matched performance... it would also be able to deep dive and find links - for example - between the proportion of less positive articles or phrases about a fighter and their performance and search for patterns in those kinds of statistics... in effect we just can't know - the strengths of machine learning lire precisely in it's ability to look at everything and find patterns in the data that had never been considered before.

That said of course, there has to be limits... machine learning might work brilliantly in the design of bridges or ships, working out the most efficient way of unifying your public transport systems and might be able to master Chess better than any other system but I have my doubts about such an open ended quest with so many variables. Be fascinated to see someone try to do it though.

[Citizen Koba]

Just to clarify something: I am not working on the model due out next week lol. And Quantum computers are in their infancy. Yes this would be down the road a bit. In the meantime we will just have to depend on Queeny to fill in the blanks I suppose... What is the probability he will post something that will make sense? might break a quantum computer!

Pffftt.... NOW you tell us....

[QueensburyRules]

- -Fantasy baseball may be a bigger industry than MLB itself already early quantum crunching of sabremetric alchemy #s of teams composed of current MLBers.

Quantum boys currently cutting their teefs on US election cycles to inform us that we don't even need to vote since Mr. Quantum already knows the Weiner.

[Willie Pep 229]

- -Fantasy baseball may be a bigger industry than MLB itself already early quantum crunching of sabremetric alchemy #s of teams composed of current MLBers.

Quantum boys currently cutting their teefs on US election cycles to inform us that we don't even need to vote since Mr. Quantum already knows the Weiner.

Those math geeks are dominating ( and runied) fantasy football betting. You can't beat their algorithms.

[Bundana]

I would guess that a computer would figure out things like: What is the probability of Marciano being able to physically defeat everyman in a drinking establishment circa 1956...? and account for hyperbole by comparing a probability like this to "marciano could beat ten men in an alley with one hand tied behind his back." What is hyperbole? essentially it is when a claims probability is claimed a possibility in excess of such probability. What we call things like irony and hyperbole are in essence things that are used that way because they are improbable... I believe a computer could learn that if it was not binery.

Yeah, I believe you (and Citizen Koba) are right: With access to everything, the computer should be able to recognize obvious hyperbole, and separate that from more serious opinions.

As far as the printed word is concerned, the computer is dealing with FACTUAL things. It has the result of every fight we know of, and (in some cases) a description of what happened, and what people thought about it.

But how will it react to the fight films? Will there ever be a computer so sofisticated, that it can "watch" these films and form IT'S OWN opinion, as to whether the footage represent good or bad boxing? Of course it can combine what it can "see", with what it can read about the fights, and what it knows about the boxers (and their opponents) from other fights.This way it will know, that SRR (for example) was a superb boxer, and will be able to deduce, that what it can see from his best fights, represents good boxing.

But what if we let the computer know everything boxing-related... except for anything that has to do with, say, Roy Jones Jr.? This information will be "hidden" from the computer! So how will it react, when we suddenly present it with prime Jones video? Will it recognize his highly unorthodox skills as something special? Will it be able to think: "Oh man, look at that speed, those reflexes - what a great fighter"? Or is a self-learning computer that can actually recognize beauty, too much to expect (even far into the future)?

As for predicting the result of fantasy fights, this is an important point! There's of course no "right" or "wrong" way to box, and different people can have completely different opinions, when it comes to evaluating the same boxer. So how is the computer going to evaluate, what it is "watching"? Will it think Fitzsimmons looks great, because 130 year old articles say so - or can it "see", that he looks quite awful? A more contemporary boxer who is difficult to categorize, is Andre Ward. Some think, he was a terribly boring spoiler - while others saw him as a very clever (in an unorthodox way) boxer, with a high ring IQ, who was great at "reading" his opponents.

So if, for example, we ask the computer, who would win a fight at 168 lbs between Fitz and Canelo - I wouldn't trust it's opinion any more than my own!

[Willie Pep 229]

Yeah, I believe you (and Citizen Koba) are right: With access to everything, the computer should be able to recognize obvious hyberbole, and separate that from more serious opinions.

As far as the printed word is concerned, the computer is dealing with FACTUAL things. It has the result of every fight we know of, and (in some cases) a description of what happened, and what people thought about it.

But how will it react to the fight films? Will there ever be a computer so sofisticated, that it can "watch" these films and form IT'S OWN opinion, as to whether the footage represent good or bad boxing? Of course it can combine what it can "see", with what it can read about the fights, and what it knows about the boxers (and their opponents) from other fights.This way it will know, that SRR (for example) was a superb boxer, and will be able to deduce, that what it can see from his best fights, represents good boxing.

But what if we let the computer know everything boxing-related... except for anything that has to do with, say, Roy Jones Jr.? This information will be "hidden" from the computer! So how will it react, when we suddenly present it with prime Jones video? Will it recognize his highly unorthodox skills as something special? Will it be able to think: "Oh man, look at that speed, those reflexes - what a great fighter"? Or is a self-learning computer that can actually recognize beauty, too much to expect (even far into the future)?

As for predicting the result of fantasy fights, this is an important point! There's of course no "right" or "wrong" way to box, and different people can have completely different opinions, when it comes to evaluating the same boxer. So how is the computer going to evaluate, what it is "watching"? Will it think Fitzsimmons looks great, because 130 year old articles say so - or can it "see", that he looks quite awful? A more contemporary boxer who is difficult to categorize, is Andre Ward. Some think, he was a terribly boring spoiler - while others saw him as a very clever (in an unorthodox way) boxer, with a high ring IQ, who was great at "reading" his opponents.

So if, for example, we ask the computer, who would win a fight at 168 lbs between Fitz and Canelo - I wouldn't trust it's opinion any more than my own!

Some years back they fed a computer multi-millions of photos - when it was done and they tested the computer's ability to recognize- identify objects. The computer got only one question correct: "that's a cat"

It seems there an inordinate number of cat pics on the internet.

We're still a long away from a self learning computer - right now it's still 'garbage in, garbage out.