#tbtf newton | Explore Tumblr Posts and Blogs

palmettocapital · 4 years

Text

A Man For All Markets (2017)

Recommended by: @Dutch_Book (who gave me one of his personal copies) and @JerryCap

Super interesting childhood -- didn’t speak a word until he was two and then began speaking in full sentences; reading at a 10-yr-old level when he was 5. Details a lot of “science play” including a number of fairly advanced amateur chemistry and physics experiments he conducted as a kid

Compares ham radio to the first Internet, allowing amateurs with simple setup to communicate with people all over the world. Had to pass a licensing exam on Morse code to operate one. Used a self-devised system to teach himself Morse that was 4x faster than the method used to teach army trainees

Method for permanent retention was to fall asleep reviewing material he’d taught himself

Martingale system -- betting scheme that requires “doubling up” after each loss in a game like roulette, consistently betting (eg, always red), will eventually recoup your loss but subject to needing a very large bankroll and bets may get so big the casino disallows them

Initially attracted to beating roulette because he believed an equation properly factoring in friction and gravity could predict how the ball would fall, though conventional wisdom was that any noise would be enough to ruin the prediction. Later proved this to be the case.

Basic blackjack/21 dates to at least 1601 (mentioned by Cervantes), referred to by 18th Century French as vingt-et-un. Conventional wisdom was that it was mathematically proven no system of varied bets could beat the house’s edge, but Thorp decided to test it when he learned of a system that lowered house edge to 0.62% (best of the casino games on offer).

His core insight on blackjack was that previous calculations had assumed the odds of any card being played stayed constant throughout the game, whereas Thorp realized the real odds varied wildly and depended on which cards were left in the deck --> means there are periods where player has an edge vs. house. So player just has to keep track of the shifting odds and vary his bets accordingly. This is the beginning of card counting. Simply, player’s edge for a deck in progress depends mainly on fraction/percents of each type of card present.

Modern computing (IBM 704) and Thorp’s access to it as an junior professor at MIT made his advances possible. Basically used a brute-force approach that wouldn’t have been possible by hand to test the impact of removing various card values from the deck on the player’s edge. Taking out small cards (2s-6s, and particularly 5s) helped player, removing 10s and Aces hurt player.

Devised his “ultimate strategy” which assigned a point value to every card, but was too hard to keep track of mentally. Decent compromise for usability was assigning 2s-6s a value of +1, 7s-9s a neutral 0 and 10/J/Q/K/A a -1. As small cards (+1s) are seen, implies a higher proportion of large cards remaining, so count rises and deck becomes more favorable.

5s are the most impactful card, but he decided to use a measure of “ten-richness” for casino play because there are 4x as many 10-value cards in a deck. Deck starts with 36 non-Tens and 16 Tens or a ratio of 2.25. Below 2.25 is “ten-rich”, at 2.0 the player has an edge of 1% and it rises as the ratio drops.

Decided to publish his results quickly because he believe in science it’s common for the time to be right for a discovery, with multiple researchers discovering something at the same time (eg, Newton & Leibniz with calculus; Darwin & Wallace with evolution). This was clearly the case with blackjack owing to improving computing power.

Initial approach was to bet 2x his minimum when edge was 1%, 4x when it was 2%, up to 10x when his edge was 5%.

Spent a lot of time practicing his play as well as working his way up in terms of what he was emotionally comfortable betting in order to be able to play his system calmly/accurately.

Why 16 is such a bad hand in blackjack - draw and you bust with high probability, stand and dealer likely to beat you with 17+.

Odds swing around a lot more near the end of a single deck, so some players use “end play”. Casinos began to combat this with early reshuffling to keep returning deck to its initial state. Early reshuffling slows play and frustrates casual players, which hurt profitability -- eventually card-shuffling machines made this must faster, but casinos paid a per-shuffle fee to those vendors, so it was still a hit to profitability.

Discusses how there are some systems that rely on tracking cards following a shuffle, since one shuffle does not fully randomize a deck (apparently it’s been proven it takes 7 thorough shuffles to effectively randomize a deck). He built mathematical models and empirical studies to study shuffling.

“Third base” (left-most seat) best one for the card counter because he’s last to act and has seen more cards because play begins to dealer’s left.

Tested his mental/emotional state before playing by “counting a deck” -- if he could do this in 20-25 seconds, he was ready to play at speed.

Never play at a table for a 2-card blackjack has been changed from original 3:2 to something lower.

Their roulette strategy applied Kelly criterion thinking, ie that it’s worth forgoing some expected gain for a large reduction in risk. For example, their betting system on roulette yielded average profit at 44% of bet for one pocket, but a 43% advantage if they reduced risk by spreading the bet across 5 numbers (expected number + 2 on each side). Called “voisinage” betting in French.

Along with Claude Shannon (MIT prof & “father” of information theory), created the first ever wearable computer for beating roulette -- believes this is perhaps the first instance of a computer that could outplay any human at a game (later occurred in checkers, chess, Go, Jeopardy)

Key to beating baccarat was that while it was impossible to win on Banker/Player bets (due to card-counting being less effected with 8 decks in a shoe at once), side bets (around natural 8/9) had exploitable characteristics. These were later eliminated.

Baccarat is a high-rolled game and was massively profitable for casinos -- accounted for 50% as much profit as blackjack with 2% as many tables, ie 25x on a per-table basis.

His partner sourced investors for their first fund by going to the courthouse, pulling other funds’ filings, and cold-calling their LPs.

His approach to arbing warrants was to use computers to draw a fair value curve for the security and then buy or short the common/warrant based on which was rich/cheap to the curve, using the slope of the curve at that point as the hedge ratio.

Converted Bachlier’s options pricing theory into a usable trading strategy by subbing in the risk free rate (UST bill with expiry closest to warrant expiry) for growth rate & discount rate. This was in 1967 vs. Black & Scholes publishing their formula in 1972.

Due to inability to assess cultural fit in interviews, his approach was to hire everyone for his fund on a provisional basis for 6 months

CBOT introduction of CBOE in 1973 was a game-changer for increased liquidity and lower costs to trading options (put options came in 1974)

His fund’s risk management approach rejected conventional notions like VaR in favor of more comprehensive/intuitive questions around what major risks could befall market or counterparties.

“Junk bonds” -- Milken/Drexel approach was a game-changer because it provided steady capital to less established companies that had not had financing needs met. This new spigot of capital increased competition for more established companies and led to hostile takeovers and restructurings, and apparently a structural rise in valuations -- he argues that blue chip companies with low returns on capital were vulnerable to Drexel-finance takeovers that often made them more valuable through more aggressive investment.

Interesting discussion of the arbitrage opportunity his son discovered in 1990 to participate in de-mutualization of small S&Ls across the US by becoming a depositor and subscribing pro-rata to rights offerings as they raised equity. Notes that management incentives made it clear this would happen over time as directors had priority subscription rights before depositors and could therefore capture outsize value.

He declined to be an LP at LTCM because he had heard John Meriwether had a history of taking outsize risk at Salomon, and he though Merton & Scholes lacked street smarts/investing experience.

His modeling suggests 2% rule is preferable to “4% rule” typically used in retirement calculations.

Key to using leverage responsibly is factoring in the worst imaginable outcome and whether you can tolerate it. If not, reduce borrowing.

Optimal bet under Kelly = edge/odds. Causes you to bet larger in situations with minimal downside or big edge. Because Kelly method leads to large swings, most users go <optimal, like half-Kelly or less. Requires long time horizons.

Since 1520, 85 institutions have remained continuously in existence -- 70 of these are universities.

Believes it’s easy to spot bubbles once well under way. Hard to profit, but key to avoid it.

Believes in scaling down TBTF institutions as a general rule.

Studies pre- and post-GFC show earnings and stock performance for a given company correlate negatively to the % of corporate profit paid to top-5 executive.

Believes there is a need for more probability/statistics education at the K-12 level -- aligns with Annie Duke there.

Economists have found one factor explains a nation’s future economic growth & prosperity above any other -- output of scientists & engineers.

Takeaways

Thorp is clearly a remarkable person -- a genius autodidact and a classic example of what Taleb would call a “skeptical empiricist”, somewhat obsessive about testing conventional wisdom using his own calculations, models and experiments, often identifying significant profit opportunities. It’s also a story about how the ability to stay abreast of technological breakthroughs (modern computing, new programming languages, new financial instruments, etc) over a long career can be extremely profitable, especially if you consistently get to them before others.

There are lots of remarkable autobiographical features I skipped over in my notes that make the book particularly worth reading. Thorp is an amazing idea of “American dream” type mobility -- born to a working class family during the Great Depression, attending public schools, winning scholarships and eventually working his way up to great wealth. The book makes him sound like the financial world’s Forrest Gump, as it covers almost every major issue of the last 50 years and Thorp/s involvement -- for example, he invents the card-counting method; “beats” blackjack, roulette and baccarat; meets Buffett as he’s winding down the Buffett Partnership; discovers Black-Scholes before Black/Scholes/Merton; participates in the early days of convertible arbitrage and evolves it into statistical arb; determines Madoff is running a fraud in 1991; pioneers quant and factor investing and the market-neutral approach to hedge funds, helps Ken Griffin & Frank Meyer get Citadel off the ground and signs on as their first LP; etc.

In many ways, the book is written for an investor with little technical knowledge and does a good job explaining much of what he worked on. The later ~1/3 of the book focuses on a number of his business ventures, opinions on politics/wealth/charity, and high-profile anecdotes that are interesting without being particularly technical.

#gambling #risk #investing #investor profile

0 notes