Gepardo

I chase numbers. The ones that moved markets, broke funds, and nobody wants to explain. In their own words.

[Gepardo @Ox_Cheetah]

Someone who waits until 35 to start investing $500 a month, then does it faithfully for thirty straight years at a 7 percent return, puts in $180,000 of their own money and retires with about $610,000. A lifetime of discipline, and a decent result.

Someone who starts the exact same $500 a month at 25 retires with about $1,310,000. Same monthly amount, same fund, same 7 percent. The only difference is a ten year head start. Net worth impact: an extra $700,000, more than the entire nest egg of the person who waited a decade to begin.

Same monthly cash out, same fund, seven hundred thousand dollar swing on a spreadsheet nobody in high school teaches.

The single most important number in your financial life is not how much you invest. It is how early each dollar starts working. Money compounds on itself, so the first ten years carry more weight than the final thirty. Waiting a decade does not cost you a decade of savings. It costs you the most valuable decade you will ever have.

Albert Bartlett gave one famous lecture on the exponential function, and opened it by calling our inability to understand that function the greatest shortcoming of the human race. The talk is free on YouTube. It is the same curve compound interest rides, the one a pension fund leans on when it promises a stranger a check every month for forty years.

The lecture is free. A ten year head start beats every raise you will ever get.

[Zynex @0xzynex]

https://t.co/s7CSVJZhBN

[Gepardo @Ox_Cheetah]

Your dealer retires on the $700 they called normal. Over $700 a month, 72 and 84 month terms, traded in three years later still owing, negative equity rolled into the next loan.

Put that same $700 a month into the market, 25 to 55, at 7 percent. $854,000

Same money leaving the account every month. One path is a driveway of cars you already stopped owning. The other is the better part of a million dollars and a used car that runs just fine.

A loan payment and a retirement balance are the same geometric series, pointed opposite ways. One drains you on a schedule. The other fills you on one.

David Jerison derives that series from scratch in MIT 18.01 Lecture 37. The lectures are free. The dealer already pointed it at you.

Which direction it runs for the next 30 years is the entire decision.

[Lindsey @0x_Squirrel]

https://t.co/C8fqwvhYMf

[Gepardo @Ox_Cheetah]

@0x_Squirrel Run the 1% out to 40 years, not 30. It stops being a quarter of the pot. At a 7% gross return, the fee quietly takes about a third of your final balance. The longer you hold, the bigger the cut.

[Gepardo @Ox_Cheetah]

@Kv0r1n The moat you can see is already priced. What matters is the fade rate, how fast the advantage decays. Buffett is talking about duration, not size. Most models measure the wall and ignore how fast it erodes.

[Gepardo @Ox_Cheetah]

Ten funds were up. Investors lost 9.9 billion dollars anyway.

David Swensen, who ran Yale's endowment, put the arithmetic on the board on camera in 2011. The top ten internet funds, three years on each side of the peak.
Time weighted: up about 1.5 percent a year.
Money in: 13.7 billion.
Money lost: 9.9 billion, 72 percent of it.

The advertised number is time weighted: it assumes one dollar stayed in the whole ride. The dollar weighted number, the one that counts when your money actually arrived, was far worse. Most of it arrived in late 1999 and early 2000, chasing the run, and left in 2001 and 2002.

It was not a fluke. Across all seventeen categories of US equity funds, dollar weighted trailed time weighted in every one, worst in the hottest sleeves.

The number in the ad belongs to a dollar that stayed in the whole time. Yours depends on when you buy and when you sell, and the crowd buys after the gain and sells after the loss.

[Superior @andreysuperior]

https://t.co/6fbpq78hsH

[Gepardo @Ox_Cheetah]

@qentuaxbt No fees, no override, no salary is not modesty. It is the one variable a manager fully controls. Returns are uncertain, fee drag is not. He locked in the part he could actually guarantee.

[Gepardo @Ox_Cheetah]

@0xQwertyx Tesla shorts lost about 38 billion in 2020 alone (S3 Partners). The DCF was fine. The position sizing wasn't. Unbounded downside beats a correct model every time.

[Gepardo @Ox_Cheetah]

@0x_Squirrel He formalized this with the Kelly criterion. Full Kelly maximizes long-run growth but the drawdowns are brutal. Half-Kelly gives up about 25% of the growth rate and cuts volatility in half. That is the price of staying in the game.

[Gepardo @Ox_Cheetah]

You don't. When the correlation regime shifts, the singular vectors rotate and yesterday's null space stops being null. That is the tell: watch the small singular values, not the big ones. The quarter your cutoff fails is the quarter a direction you threw away starts stretching again.

[Gepardo @Ox_Cheetah]

Gilbert Strang has 20 million views on his lectures. In this one he still writes the wrong sign on the board and says phooey.

It is MIT 18.06, Lecture 29, the singular value decomposition, the course he taught for 61 years. He calls the SVD the final and best factorization of a matrix, and he means it.

Any matrix A splits into three: an orthogonal rotation, a diagonal stretch, another rotation. A equals U sigma V transpose. The diagonal holds the singular values, one stretch factor per direction, ordered from the strongest signal down to zero.

The input directions with a stretch of zero are the null space. Strang shows them fall straight out as zeros on the diagonal. In a matrix of stock returns, those are the combinations that look like a strategy and move no money.

The most watched linear algebra teacher alive, debugging his own matrix in public. Halfway through a two by two, four four minus three three, the signs flip: three minus three where he wanted minus three three. He stares at it, says phooey, and punts it to Wednesday.

A computer runs the whole decomposition in one line of code. The call it cannot make is which of those tiny singular values is real signal and which is noise you throw away. That cutoff is the job, and it is still yours.

[Gepardo @Ox_Cheetah]

@batagonx The $500K death benefit in 30 years buys what $250K buys today after inflation. The index fund path adjusts upward. The whole life payout stays fixed. The gap widens every year.

[Gepardo @Ox_Cheetah]

@0_xSan He swore an oath to Tartaglia, then found del Ferro had solved it first and used that as his loophole to print it in Ars Magna, 1545. Same instinct as the dice book: take the edge, publish, let the credit fight itself out.

[Gepardo @Ox_Cheetah]

A $7 trillion industry started as a dice book.

Girolamo Cardano, Milan, around 1564. His Liber de Ludo Aleae was the first full treatise on probability: count the ways a die can fall, turn them into odds you can bet. He wanted the money, not the theory.

Peter Bernstein, the man on the tape, dates the birth of probability to the 1640s. By then Cardano was dead and his book was still in a drawer.

1564: Cardano writes it.
1654: Pascal and Fermat publish the readable version.
1663: Cardano's book finally prints.

A premium is a price on a probability distribution.

The insurer runs the same move on your life. Count the cases, price the residual, add a margin. Same operation as reading a die, a different balance sheet. It sells you a number it can compute and you cannot.

That move now prices your car, your house, the year you are likely to die. Cardano just got there first, for a game of dice.

[Gepardo @Ox_Cheetah]

@0x_Squirrel The die is fair, the life table is not. The faces are not equally likely, so the insurer prices the skew, not just the count. Cardano assumed equal outcomes. Mortality is where the math stops being fair and starts being profitable.

[Gepardo @Ox_Cheetah]

@0x_Squirrel A 70% drawdown needs a 233% gain just to break even. Pabrai had to clear that plus the 6% hurdle compounding every year before he earned a dollar. That's what the high-water mark actually costs.

[Gepardo @Ox_Cheetah]

@DaniilBuilds Seeing it is step one. LTCM saw it better than anyone and still hit zero. The edge that lasts is the one you can hold through a bad run.

[Gepardo @Ox_Cheetah]

In 2002 a 30 year old furniture salesman named Jason Padgett was beaten unconscious outside a karaoke bar in Tacoma.

He woke up seeing fractals in tap water.

He is now one of roughly forty documented cases of acquired savant syndrome on record. He can draw the mathematical structure behind sunlight, tree branches, and the motion of a spoon through coffee. The math he sees is called probability theory.

Every high-frequency trading desk on Earth runs on it. Every options market maker. Every insurance underwriter. Every casino floor in Vegas.

Padgett spent his first thirty years living what he called "a mile wide and an inch deep." The math was there the whole time. Nobody had shown him how to look.

Then a random assault rewired his visual cortex and forced him to see it.

You do not need a head injury. The lectures are on MIT OpenCourseWare, taught by Philippe Rigollet in course 18.650. Free.

Wall Street does not gamble. It runs a probability machine that mathematically cannot lose over enough trades. The house edge is not luck. It is a theorem published by Andrey Kolmogorov in 1933 in a book shorter than most business books.

Retail traders sit on the other side of that theorem and call themselves investors.

There are only a handful of ways money actually compounds and every one of them is a probability distribution somebody understood before you did.

The math is free. The willingness to actually look at it, without a concussion forcing you, is the entire fortune.

Save this for the next time your broker tells you a story. One like it every Sunday if you follow along.

[Di Krass @Di_Krass_]

https://t.co/VG8RsppBKq

[Gepardo @Ox_Cheetah]

@0x_Squirrel The formula is free, size and survival are not. Retail pays in leverage and time, the two things the table is built to drain.

[Gepardo @Ox_Cheetah]

@Di_Krass_ Right. The map is not the territory. Knowing the distribution is not surviving it.

[Gepardo @Ox_Cheetah]

@0x_Squirrel Refinancing trap resets the clock. Ten years into a 30-year, you refinance to a new 30, and you're back to paying mostly interest. The lower payment isn't free. Run the amortization table before you sign.