jd jdmaturen

On the relative value of startup stock

How do you compare the potential value of different stage companies? And as a company progresses what is the appropriate amount of stock to give new employees?

Using data from Crunchbase and Yahoo Finance we can calculate the average value created per company broken down by how much money they have raised. From this we can then compute relative returns at the different stages, e.g. $1M raised, $10M, $100M, etc.

tl;dr

For this exercise we'll calibrate relative to expected return from a company that's raised $1M.

Heads or Tails Extended

In fact, only 2 out of 2,862 broad domestic stock funds were able to outperform their peers consistently over five years, according to one measure: performance in the top quartile of funds over five consecutive 12-month periods ended in March 2014. That translates to just 0.07 percent of the funds, which means that more than 99.9 percent of funds fell short of that feat.
…
Repeat those double flips five times and you’ll find the probability of a mutual fund ending up in the top quartile five times in a row through chance: 0.098 percent. (We’re flipping the coin twice for each year of mutual fund performance.) That’s a bigger probability than the 0.07 percent scored by the actual funds. This means that if mutual fund managers had just flipped coins, roughly three of them — not two — would have been expected to end up in the top quartile for five years in a row.

Indeed:

&gt;&gt;&gt; 100 * .25 ** 5

Keybase proof

I hereby claim:

I am jdmaturen on github.
I am jdmaturen (https://keybase.io/jdmaturen) on keybase.
I have a public key whose fingerprint is A95A 6A67 8A9C 7701 272D 7CBD 126A 7513 766A 6932

To claim this, I am signing this object:

	6An
	80P
	2lf
	1c6
	7zS
	1U6
	2de
	2yZ
	2a2
	6I8

	burp:~ jd$ curl --referer http://t.co/ -is i.imgur.com/7RsFUmL.jpg
	HTTP/1.1 302 Moved Temporarily
	Server: cloudflare-nginx
	Date: Wed, 14 Aug 2013 18:58:01 GMT
	Content-Type: text/html
	Content-Length: 165
	Connection: keep-alive
	Location: http://imgur.com/7RsFUmL
	CF-RAY: 9dfdfc465cb0707

	"""
	Implementation of the shifted beta geometric (sBG) model from "How to Project Customer Retention" (Fader and Hardie 2006)

	http://www.brucehardie.com/papers/021/sbg_2006-05-30.pdf

	Apache 2 License
	"""

	from math import log

	"""
	Implementation of the beta-geometric/NBD (BG/NBD) model from '"Counting Your Customers" the Easy Way: An Alternative to
	the Pareto/NBD Model' (Fader, Hardie and Lee 2005) http://brucehardie.com/papers/018/fader_et_al_mksc_05.pdf and
	accompanying technical note http://www.brucehardie.com/notes/004/

	Apache 2 License
	"""
	from math import log, exp

	import numpy as np

	"""
	Get a bunch of Crunchbase data, but respect the API limits.

	Author JD Maturen
	Apache 2 License
	"""
	import logging
	from random import random
	import sys

	import math

	import numpy as np
	from scipy.stats import beta

	def nps_beta_dist(sample_size, promoters, detractors, confidence=95):
	"""
	Confidence range of NPS score. NPS score is defined as the percent of promoters
	minus the percent of detractors. See also http://en.wikipedia.org/wiki/Net_Promoter

	burp-2:foo jd$ python probabilistic_strategy.py 1000 16 64
	request_count: 1000 cluster_size: 16 threshold: 64

	probability of being quotad at a given local concurrency:
	1 0.015070684079
	2 0.07629533815
	3 0.202571187171
	4 0.378831226431

	mean observed global concurrency limit: 64.033