No Pain No Gain?or how Idiots destroyed the Casino
The Theory That Would Not Die: How Bayes' Rule Cracked the Enigma Code, Hunted Down Russian Submarines, and Emerged Triumphant from Two Centuries of Controversy
The book is amasing. Bayes contradicts number theory. Yes, Bayes is very elegant and arousing for cult workers and gamblers. The dude, who once called it "crack cocain of modern statistics ... seductive, addictive and destructive" was *absolutely* correct. And now that same dude is recruiting the addicts to do more damage for one US corporation. Books are cool. Even when the author has no idea what they are talking about.
Let's ask the simple question. In case of a Casino, is Casino better off when the dealer stacks the deck, or is Casino better off when the pure chance stacks the deck? Simple minded people would usually say : Of course - the rigged dealer is *better* for a Casino! "Everybody knows, the game is rigged" and all that jazz. Only in reality it's not so simple (and in a way it is similiar to Automation Paradox)
Here is why.
Everything is a complex combination of probabilities, that was *very* hard to figure out before 2008. Only in 2008 they decided to simplify things (Obamacare basically destroyed the state of the art medical insurance industry of the US and because of that, the US lost set of critical defence mechanisms). Then in 2016 they simplified things even more, weaponizing the USD.
As a result, they took the once *very* complex system of US governance and dumbed it down to something very simple, like "we print money and pump oil and that's it".
They thought that they were improving things, but as a result of the simplification they opened up the doors to very primitive attacks on the US financial system. And that happened in 2020-21 (the trillions of dollars were just "transferred" and "nobody knows where")
So on one hand you are right - the rule of "risk/reward" is in general "correct".
Only it is "more correct" and "less correct".
Some of US fintech management (let's call them "idiots") decided that stacking the deck increases the defence. In fact the *opposite* is true. When the deck is stacked every day and everybody knows the way it is stacked - Casino can not possibly win.
Snowden pointed that out many years ago. He pointed to the fact that Facebook choice dialogs only have a button "Ok". (The actual choice dialog should have "Yes", "No" and "Cancel" choice. Or at least "Ok" and "Cancel")
By minimizing choice and simplifying the system you *grossly* simplify the trade. And that always backfires. It just did in 2020. Hard not to see it. It's planet scale.How Redditors Exposed The Stock Market | The Problem With Jon Stewart - has a nice diagram(s) (and overall good explanation of what was happening).
Aug 27 2022
Another moment from that book - Pierre-Simon Laplace was most likely not one dude, but a representative of a collective. Because I now understand exactly what he was doing (he was doing everything I was doing myself, only he had no computer, so he had to do it all manually - which is impossible for a single human being). The question now is "why would a group of people hide their work behind a fake representative?" (Burbaki was the same). And the answer is in this book: Infinitesimal: How a Dangerous Mathematical Theory Shaped the Modern World.. It's not the first time the planet has to deal with the Idiots and the Idiots always do the same thing to anybody, who is smarter than them. Of course smart people have to hide behind a (protected) representative. No magic in that.
I find it very charming that at some point Laplace said that because he hates gambling, he will drop it and get back to calculating planets orbits. That's because mathematics of gambling is a bit more complex actually. Because gambling adapts realtime. Cycles are shorter. And he had no computer. So he just dropped the whole thing (at the most interesting inflection point, by the way). It's all in a book. It's a great book, really. Maybe exactly because author has no idea what is she *really* talking about.
Aug 28 2022
From BTC page
The first episode of British Sherlock Holmes (the taxi driver) is a stunning example of british math versus french/german math. Hardcore european mathematicians simply write it off as "this is not math". Only it is. It is ML math. Gives a good idea about the future also.
Sep 14 2022
Kolmogorov also contradicts number theory. He contradicts it because of Deblin method. Now the problem is that Deblin method can not be found. There is Keith Devlin in Stanford. Mathematician. Looks like Gavin. Born in 1947. Kolmogorov's work is 1949. And using that approach they could not possibly shoot accurately. Mathematically impossible. Yeah!
By the way this ML stuff. It's from the 40s. Some "New Technology" indeed.
Sep 19 2022
The puzzle from Abigail Disney
Documentary filmmaker Abigail Disney, granddaughter of Disney co-founder Roy Disney: The American Dream and Other Fairy Tales
In that documentary Abigail points to the fact that salary of Disney's CEO (Robert Iger, he was removed in 2020) was 700x times the average salary of Disney rank and file staff. She was pointing to the fact that it's ridiculous (everybody agrees on that), she was asking what can be done? (The answer was "nobody knows, because if salaries raised Disney will be toasted on stock market" (questionable statement)). In general the overall idea was that "well, something might be done but it looks complex we don't know what should be done". This is because most of US management has no math education. In reality, the formulas exist how to handle these kinds of wages discrepancies, and those formulas are known for *centuries*. A friend of mine did his PHD on similar topic (in collapsing SU). But more interesting was observation of my wife, who run the numbers and calculated the leveraged reasonable ratio. It should be not 700x but 40x for a solid hierarchy introduced by Dutch into Peter The Great Russia. Peter : 1675 - 1725. So yes. Those formulas are known to Dutch since XXVII century. Also Parkinson's law. (1942).
Bring educated people into the government. It might go a long way.
Sep 24 2022
I found too many relevant books on this now. Kolmogorov still has the best math part, but Keith Devlin's (humanitarian) work is stunning. His "Unfinished Game" forced me to re-think everything. The Bayes book is written by a journalist, Keith is a mathematician. Now might need to read it all. Decisions, decisions.
Oct 08 2022
So in "Unfinished game" there are several themes, all interesting. The key one is Pascal / Fermat. I personally think that Bernoulli / Leibniz theme is more important. I actually figured out how to apply Bernoulli's approach to Python coding. Go figure. Now another important moment is Pascal's wager. Keith downplays it's importance (because it's logic and Keith's theme is "probabilities vs calculus, how one overshadowed another" (amazing observation really)) so he has no place for logic aspects in his book. Only if you look at Pascal's wager it is clear that it is the (only) French thing that effectively resonates with "To Be Or Not To Be" (which happens to be the keystone of English culture). So the only link between French and English culture that even I can see, would be the bond of Pascal's wager and "To Be Or Not To Be". Another pair from Keith's book would be Bayes / Kolmogorov. Only Keith does not make that pair and after they outlaw Russian in Eu, this link might disappear. Forever. Not only Kolmogorov is a link to Bayes, he is also a link to Turing (via complexity). And because of his works on Artillery, it looks like Kolmogorov might be the only link between French and British at all. Well, with Russian language being outlawed basically, that *will* affect the pillars of current civilization. And I think it is the goal of cancel culture anyways. They bring down the collossal figures. They bring down the pillars. Naturally, the Oil trade will be attacked and the Gold trade will be attacked. Everything that holds the planet together will be attacked.
Oct 19 2022
Dijkstra's joke about debugging and bugs is a side effect of Bernoulli's work. Basically, we live in pre-historic times. People have no idea about computing. Well, the stuff those assholes did to Turing did not help, I suppose.
Oct 23 2022
Surpirisingly few people understand the difference between Frequentism and Bayesian. 60s were key. Empirical Bayes was best.
Nov 8 2022
P-values == Frequentist. As simple as that. ZK's are based on P-values as well. Yes, I glanced through some of the key crypto math recently. Again.
Nov 9 2022
There was an interview with Alameda CEO in which she made a few mathematical statements. Simple part she got right (first leg of crypto bubble was based on a trivial highschool math). Complex part she got *wrong*. They asked her a question about SL. Her answer was boolean. Which is ridiculous, but that's how they see things on that coast. The correct answer should have been (of course) propotional. P-values again. Actually, if they just stayed with arbitrage (which they did to the tune of millions a day), they would have been just fine. They were spectacular fools, basically. "Glass Onion" explains how it works.
Nov 22 2022
Walking up the hill to the University of Wisconsin in Madison, Box told the old man, "I'll give them equal probabilities, so if I have five hypotheses, each has a probability of one-fifth". Fisher responded ... "This is what I'm going to say and then you're going to shut up". "Thinking that you don't know and thinking that the probabilities of all the possibilities are equal is not the same thing.". Page 134.
Another key quote should be how do you calculate the probability for something that *never* happened before, ever. Like the insurance company had to calculate P-value for two planes colliding in the air, before the first episode even happened. The book does not give the answer, it only tells that the guy calculated the P-value and that it actually worked. Book does not say *how* he did it. The book gives enough context to re-construct the answer "how". Bayes would have failed, naturally.
I wonder if FTX management was accurately asked a few questions before being appointed and of course the kids gave the wrong answers (they were all set up, so giving the wrong answer was a pre-condition for getting a job). MIT is pure Bayes I think. Stanford is 50/50% Bayes, I think.
Nov 23 2022
Wolfram's new kind of science. Published 2002. I could not understand that book at all. 20 years later? I totally understand it. Basically, John Conway (died 2020) gave (?) him a Game of Life (Conway had a *lot* of cool stuff to play with) and asked Wolfram to bring some structure to the Chaos of Life. Wolfram describes all the steps that he took to do that. And he did just that. He produced a Programmable Cellular automata, that is capable of producing ornaments. In one of the pages he has an amazingly important picture, one can take that one picture and produce a PHD (or even an entire book, like Wolfram's). Yes, it's recusvive. I actually knew what to look for, that's the only reason I noticed that picture. This single book is worth a little library. Wolfram is an encyclopedist. British know the importance of encyclopedias.
There are several more layers to this Life / Chaos / Structure, but let's ignore those extra layers for now.
Nov 24 2022