The Physics of Wall Street: a brief History of Predicting the Unpredictable

“Cervantes . . . wrote stories in which his characters became proficient . . .”: See 
“rinconete and cortadillo” in cervantes (1881).
88 “. . . and then offered to submit Thorp’s paper . . .”: the paper was accepted and 
published as thorp (1961).
89 “His name was Manny Kimmel”: Kimmel’s life, including the story of how 
his parking lot business was transformed into the time Warner empire, is described in 
Poundstone (2005) and, especially, Bruck (1994). the back story told here is based on 
those sources; the story of Kimmel’s trip to vegas with thorp is based on thorp (1966).
91 “. . . in a book, Beat the dealer . . .”: See thorp (1966).
92 “. . . a paper written by one of Shannon’s colleagues . . .”: the paper was Kelly 
(1956). for more on the Kelly criterion, see thorp (2006), MacLean et al. (2011), and 
thorp (1984, Pt. 4).
92
“Kelly was a pistol-loving, chain-smoking, party-going wild man . . .”: this ac-
count of Kelly is based on Poundstone (2005).
93 “Imagine you’re in Las Vegas, betting on the Belmont Stakes . . .”: for the pur-
poses of illustration, I am intentionally ignoring federal laws on wire-based gambling 
that were in effect in the 1960s.
94 “. . . you’re guaranteed to win one of them . . .”: Suppose you have $100 to start 
with, and you put $17 on epitaph and $83 on valentine. If valentine wins, you get back 
your original $83, plus 5/9 more, for a total of $129. But you also paid $17 for your (lost) 
bet on epitaph. So your total profit is $12. Meanwhile, if epitaph wins, you get your $17 
plus 7 times more, for a total of $136, less the $83 you bet on valentine (and lost). So in 
this case your profit is $53. In either case you win.
96 “This would initialize the device . . .”: these details about the computer are 
based on thorp (1998).
96
“. . . the calculations for that level of precision were far too complicated”: In fact, 
the calculations for where a ball will fall given just the standard roulette setup — a ball 
rolling around a rotating wheel — were not too complicated for the computer to solve. 
However, roulette wheels are designed with small bumps on the wheel that act as ran-
domizers, so that if the ball hits one of these bumps, it will bounce around and change its 
trajectory. the computer was not able to predict precisely how these randomizers would 
affect where the ball would land, thus introducing additional uncertainty.
98 “. . . essays that featured papers by Bachelier, Osborne, and Mandelbrot”: Im-
portantly for the central argument of this book, thorp confirmed in an interview that 

---

240 
•
t h e p h y s i c s o f wa l l s t r e e t
he read the paper by Mandelbrot included in the cootner volume, in addition to the 
papers by Bachelier and osborne. though he saw how the effects of fat tails could affect 
a model based on the log-normal distribution, he nonetheless decided to use osborne’s 
simpler model for returns in constructing his options pricing formula. He took tail ef-
fects into account by behaving cautiously in the applications of his pricing formula, that 
is, by remaining cognizant of the fact that it would fail under certain circumstances.
102 “In 1967, he wrote a book . . .”: the book was thorp and Kassouf (1967).
103 “And Princeton-Newport’s demise was particularly dramatic”: this story, that 
Princeton-newport was moving positions off its books to avoid tax losses, is based on 
the accounts in Poundstone (2005) and Stewart (1992), and in contemporary news sto-
ries such as eichenwald (1989a, b). In an interview, thorp emphasized another aspect 
of the allegations, which amounted to stock parking in the other direction: a trader at 
Milken’s firm, Bruce newberg, was using Princeton-newport to move positions off of 
his books to avoid federal reporting laws and drexel’s trading rules. regan, newberg, 
and the other defendants were initially found guilty of both varieties of charges, though 
the convictions were overturned on appeal and the defendants were cleared of wrongdo-
ing.
104