Furiokao

silence before the strike knowledge that cuts through noise

[Furiokao @furiokao]

A Harvard professor has been giving away Wall Street's entire mental toolkit for free since 2006. Almost nobody has done the homework.

Joe Blitzstein teaches Statistics 110 at Harvard. Thirty-five lectures, all on YouTube. No paywall. No tuition.

Quant firms pay half a million a year for five mental models: expected value, base rates, ergodicity, conditional probability, signal versus noise. Every one is covered in this course, starting from a coin flip.

He opens the second lecture with this:

"Statistics is the logic of uncertainty. Math is the logic of certainty. If you're 100% certain of everything, there's something wrong with you."

Then tells students that after a few weeks, they will do calculations that three hundred years ago required consulting Isaac Newton in person.

That is not hyperbole. If you needed the odds of a dice game in the 1650s, you wrote a letter to Newton, Fermat, or Pascal. Nobody had worked out the rules yet.

Even Newton's intuition was wrong on a basic dice problem. A student three weeks into this free course can now solve it correctly.

The lectures are free on YouTube. The practice problems are free on the course website. Almost nobody who watches actually does the problems.

The problems are free. The willingness to solve them is the entire edge.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF

[Furiokao @furiokao]

@Ryomenex wait this perspective is actually a total game changer

[Furiokao @furiokao]

A Nobel Prize–winning MIT professor solved a problem worth trillions with one phone call, and the framework he used still fits on a single line.

A trader called Robert Merton about an option out of Hong Kong that didn't fit any pricing model on the desk. Nobody had seen the structure before. There wasn't even a name for it yet.

His entire framework fit on one line: any derivative, no matter how strange, can be replicated with a dynamic trading strategy, as long as you can state its boundary conditions.

"This looks like a very difficult problem. And of course, all it meant was we just had to put the right boundary conditions in. It's like having someone give you a secret sauce, like a Lego set, that you can build."

He solved it in an afternoon. He liked the example enough to tuck it into a paper as a throwaway illustration, not even the main result. That throwaway became the seed of an entire industry, exotic derivatives, a market that now runs into the trillions.

Consultants build whole practices around "structuring" a new product from nothing. Merton did it by refusing to treat an unfamiliar problem as unfamiliar. He just asked what the edges of the problem were.

The paper is a few clicks away, free on his own website. The framework fits on one line.

Almost nobody actually stops to ask "what are the boundary conditions" before deciding a problem is impossible.

That question was the entire edge.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF

[Furiokao @furiokao]

@Di_Krass_ that’s such a sharp way to approach any problem

[Furiokao @furiokao]

@Rossst_03 that is a brutal reality check for the degens

[Furiokao @furiokao]

@tolik12308 the math doesn't care about our ego, love that perspective

[Furiokao @furiokao]

One of the most important numbers on a Wall Street options book doesn't even have an agreed name, and an actual bank training manual blames it on a casino.

"Kappa is also called Vega by some uneducated traders at Solomon Brothers. They have mistaken Vega as a Greek letter after gambling at Vegas."

Jake Xia, who co-teaches MIT's math finance course, found it in Morgan Stanley's training manual on his first day at an options desk.

Vega measures how sensitive a book is to volatility. Two major banks couldn't even agree what to call it.

Xia uses this to make an uncomfortable point to his students.

Quant finance is barely thirty or forty years old, still fighting over its own vocabulary while the models keep changing underneath it.

So asking "is this term correct" is a trap. The naming is arbitrary, born from a joke about drunk traders in Vegas. The math underneath is not negotiable.

That's the exact split the article above is built on.

Five models, expected value, base rates, ergodicity, updating, sample size, don't care what your desk calls them or which bank taught them to you.

Rename Vega into Kappa out of pettiness, and the risk doesn't move an inch. Rename "stop resulting" into whatever sounds smarter, and the math still decides who goes broke.

None of this sits behind a paywall. What nobody hands you is the instinct to skip the jargon, straight to the ideas that were never up for debate.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF

[Furiokao @furiokao]

@koplenkoo this is such a sharp take on the memory bloat problem

[Furiokao @furiokao]

@Ryomenex that level of discipline is honestly kind of terrifying

[Furiokao @furiokao]

Deep Blue didn't beat the world chess champion by thinking harder. It beat him by refusing to think about most of the board at all.

Patrick Winston laid out the math behind that. Brute-forcing chess takes 10 to the 120th evaluations, more than every atom in the universe computing since the Big Bang, still short by 14 orders of magnitude.

Alpha-beta pruning exists because of that wall. The trick isn't computing more, it's skipping entire branches you never need to evaluate, because one path already guarantees a worse outcome than an alternative you've already found.

This is the same discipline underneath every mental model a quant actually uses. Nobody runs the financial equivalent of brute force, checking every outcome exhaustively. They prune.

Expected value prunes "will this work?" down to "what is this worth on average?" Base rates prune a 95 percent accurate test down to the two percent that actually matters. Signal versus noise prunes ten promising trades down to nothing, because ten observations can't distinguish a real edge from pure chance.

Alpha-beta doesn't get a better score, it gets the same correct answer with a fraction of the work, by ignoring branches whose outcome is already known.

That's the real skill in chess engines and quant desks alike. Not searching everything, but knowing precisely what to ignore.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF

[Furiokao @furiokao]

@BlueDeltaHQ this is such a wild way to look at market data

[Furiokao @furiokao]

Eugene Fama figured out decades ago why almost nobody in the asset management industry actually wants proof that their strategy works, and he doesn't dress it up.

"They don't want to do this because it's going to kill their fees. Why would you want to go from 1% down to fractions of several basis points on your management fee? They're not going to volunteer to do that."

He'd learned it the hard way. Paul Samuelson kept bringing him in to explain his research to TIAA-CREF, hoping it would change how they invested. It never did. Fama says the exact same thing happened everywhere he tried, until he stopped bothering to talk to practitioners at all.

That's the fork the article above is really about, just from the other side. A firm that gets paid its fee whether or not the edge is real has no reason to go looking for the truth.

A firm that only survives if the edge is real, a trading desk, a quant fund, has no choice but to run expected value, base rates, sample size on itself constantly, because the market will find the lie for free if they don't.

None of this sits behind a paywall. What nobody hands you is a business model that actually punishes you for being wrong, which turns out to be the only thing that reliably makes anyone install these five models for real.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF

[Furiokao @furiokao]

@Dep2chic it is actually wild how much emotions dictate the market cycles

[Furiokao @furiokao]

Andrew Lo has spent his entire career studying finance, and this is the blunt conclusion he keeps coming back to.

"It's amazing how bad we are as homo sapiens in managing our finances."

His OpenCourseWare lectures have been watched millions of times. He built an entire theory on that line, the adaptive markets hypothesis: efficient markets isn't wrong, it works most of the time, it's just not complete, and the gap is exactly where human behavior overwhelms the beautiful logic of the math.

His fix for that gap is almost embarrassingly simple. Before anything goes wrong, run the scenario now, while you're calm, so panic never writes the decision. In March 2020 the market fell 30% in days.

People who sold to escape the pain missed the rebound weeks later. Anyone who'd pre-committed to sit still wouldn't have noticed the dip.

Same fight as the article above, just off the trading desk. Expected value thinking stops you grading a decision by how it felt after. Lo's scenario analysis is that discipline run before the fact, pricing the drop while no fear is in the room to argue back.

The lectures are free on MIT OpenCourseWare and have been for years. What nobody hands you is the will to write the plan on a calm day for the day you won't feel calm at all.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF

[Furiokao @furiokao]

@viniluer it is so wild how we still struggle with this

[Furiokao @furiokao]

@betraidx good situation

[Furiokao @furiokao]

@monokern It's quite simple if you follow the instructions, thanks

[Furiokao @furiokao]

@monokern I'm glad that it's useful to you

[Furiokao @furiokao]

Patrick Winston justifies one of his teaching habits with actual math, not intuition, and the math is straight out of a probability class.

"If you want to reduce the probability that a particular individual will miss what you're saying to less than 1%, all you need to do is say it three times, and the laws of probability tell you you've done your job."

His reasoning: about 20% of any audience is mentally checked out at any moment. Say something three times, treat each pass as roughly independent, and the odds of missing all three drop to 0.2 cubed, under 1%.

He's not repeating himself because people are slow. He ran a probability calculation and built the lecture around the answer.

That's the whole article above compressed into a teaching trick. Most people would just feel like repeating something is safer and leave it there. Winston does the arithmetic, gets an actual number, and only then decides how many times to say it.

Structure instead of a hunch, exactly the line the article draws between the person who feels a bet is fine and the person who knows the probability it's fine.

The lecture is free and has been online for years. What nobody hands you is the habit of stopping at "that feels safer" and asking what number is actually behind the feeling, whether you're teaching a room or pricing a trade.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF

[Furiokao @furiokao]

Patrick Winston opens MIT's 6.034 by handing students two words that will do more for them than the entire syllabus.

"Once you have a name for something, you get power over it. You can start to talk about it."

He calls it the Rumpelstiltskin Principle. Before you can reason about a bias, a mistake, a pattern, you need a label for it, otherwise it just stays a vague feeling you can't argue with or build on.

Then he warns against the opposite mistake. Calling an idea "trivial" instead of "simple" is a way of dismissing it before you've ever used it.

He tells the room that MIT students miss real opportunities constantly, because they assume an idea can't matter unless it's complicated, when the simplest ideas in the field are usually the most powerful ones.

That's the entire premise of the article above, just said from the other direction. Expected value, base rates, ergodicity, updating, sample size, none of these are hard math, anyone can state them in a sentence.

What actually separates a quant is that they gave each one a name early enough to catch it firing in real time, while everyone else is standing there feeling something with no word for what they're feeling.

The lecture is free and has been online for two decades. What nobody hands you is the discipline to name the simple thing instead of waiting for a complicated one to feel earned.

[Furiokao @furiokao]

https://t.co/8wXADZ29zF