►
From YouTube: Governance and Risk Meeting: Ep. 2
Description
In this Governance and Risk meeting, Steven Becker, our Head of Risk at MakerDAO, will continue guiding us through his Governance Risk Framework document.
This one of many governance-specific meetings we will be holding in preparation for the public launch of Decentralised Governance.
Governance and Risk Framework - Part 1: https://medium.com/makerdao/makerdao-governance-risk-framework-38625f514101
Foundation Proposal: https://medium.com/makerdao/foundatio...
Website: https://makerdao.com
Twitter: https://twitter.com/makerdao
Reddit: https://www.reddit.com/r/MakerDAO/
Chat: https://chat.makerdao.com/home
Email: info@makerdao.com
A
Continuing
series
of
meetings
dedicated
specifically
to
scientific
governance
and
risk
those
are
two
and
normally
complex
topics,
but
happily
we
have
experts
with
us
that
help
walk
us
through
some
of
the
deeper
implications
of
those
things
and
what
the
mechanics
will
eventually
look
like
today's
meeting.
Of
course
we
have
Stephen
Becker,
our
rock
star
head
of
risk
and
we
have
a
special
guest
storm.
Peter
Soren
is
our
head
after
you
CERN.
Would
you
like
to
introduce
yourself
so
I?
Don't
miss,
take
care
sure.
B
A
Title
is
in
a
product
which
really
means
bridging
the
gap
between
all
the
technical
stuff
going
on
and
and
what
really
needs
to
be
done
in
in
the
business
sense,
and
then
I
also
have
an
another
job
about
figuring
out
how
we
establish
ourselves
in
Denmark,
but
I.
Think
that's
enough.
For
now.
Yeah,
that's
great,
actually
songless
mo
is
one
of
our
unsung
heroes
at
maker
down
he's
one
of
the
few
people
that
make
sure
that
things
actually
happen
and
happen.
The
right
way,
which
is
valuable
skills,
so
I'm
happy
to
have
him
here,
Steven.
A
B
So
obviously,
I'm
Steven
Becker
I
am
head
of
risk
at
maker,
and
what
we're
trying
to
do
with
these
community
calls
is
a
little
bit
of
an
education
and
a
little
bit
of
a
dialogue
with
respect
to
what
the
risk
function
will
look
like
it
maker,
as
well
as
how
that
incorporates
into
the
governance
aspect.
Now.
B
The
the
first
thing
I'd
like
to
do
is
do
a
bit
of
a
recap
from
the
previous
call
and
then
touch
on
another
question
at
hand.
What
is
scientific
governance
I'd
like
to
follow
that
up
with
a
little
bit
about
how
scientific
governance
is
implemented,
with
a
decentralized
risk
function?
Move
on
to
you
know
what
a
risk
construct
is
and
and
who
creates
the
first
one
and
then
ultimately,
or
rather
finally,
at
least
if
we
can
reach
it
in
the
session,
is
look
at
the
outline
of
the
governance
process.
B
You
know,
especially
as
it
pertains
to
to
risk
management,
so
I
think
a
good
place
to
start
would
be
another
recap
from
the
the
previous
meeting.
We
saw
that
the
risk
management
safeguards
the
the
integrity
of
the
system
and
then,
by
extension,
the
stability
of
the
coin
die
now.
The
the
way
that
you
categorize
risk
in
the
normal
traditional
setting,
I've
sort
of
dropped
to
the
side
and
I've
decided
to
rather
focus
on
the
three
layers
of
risk
that
no
we
a
face
at
make
it
out.
B
The
one
is
obviously
the
economic
risk
which
comes
from
the
you
know:
the
dynamics
in
place,
the
financial,
the
financial
risk
which
comes
from
the
exposures
that
we
have
to
take
with
respect
to
collateral
and
then
obviously
there's
the
operational
risk
which
comes
from
the
sort
of
technical
and
business
structures
and
timings
that
we
need
to
consider
and
have
in
place.
Now.
What
we
saw
is
that
risk
management
is
contextual,
but
for
the
sake
of
make
it
Dow.
B
It
basically
is
the
process
of
looking
at
these
risks
sort
of
looking
at
these
layers,
identifying
the
risks
and
then
creating
a
construct
that
can
manage
them
appropriately
and
in
doing
so,
we
we
also
went
through.
You
know
the
types
of
stable
coins
that
are
out
there
and
you
know
where
we
effectively
place
our
trust
with
these
types
of
stable
coin
and
more
importantly,
we
came
to
the
conclusion
that,
given
your
definition
of
what
the
blockchain
is
to
you
from
an
economic
point
of
view,
you
know
is
it?
B
Is
it
a
a
network
where
we
are
transferring
existing
economies
from
the
traditional
space
onto
it?
I'll
be
creating
an
Augmented
value,
add
to
the
traditional
economy
with
more
granularity
and
permutations,
or
is
it
a
bit
of
both
now
given
that
if
you
have
a
sense
of
how
your
trust
is
defined
in
terms
of
economic
growth
of
the
blockchain
and
on
the
blockchain,
it
becomes
kind
of
natural
to
see
that
a
collateralized,
stable
coin
is
a
better
proposition
in
facilitating
that
economic
growth
going
forward.
B
We
also
touched
on
what
decentralized
risk
management
will
look
like
as
a
function.
You
know
the
breadth
of
it
and
then
the
actual
verticals
in
between,
but
I'll
get
back
to
that
again
today,
but
I
think
before
I
go
on
to
what
is
scientific
governance?
I
shall
pause
there
for
a
second
just
to
see
if
everyone
is
caught
up
and
happy
with
that
recap.
I.
A
B
It
is
extraordinarily
important
to
dissect
this
phrase
and
get
a
sense
of
what
it
means
from
the
make
it
out
point
of
view,
and
you
must
excuse
me
I
like
to
revert
to
history,
two
to
lean
on
a
few
items
of
interest.
So
there's
gonna
be
a
couple
of
dates
thrown
at
you,
but
I
think
it's
gonna,
be
you
know
pretty
interesting.
You
know,
for
me,
scientific
governance
within
maker
is
governing
using
the
science
of
risk
management.
As
a
basis
of
argument.
Now
that's
extrordinary
important
to
to
think
about
that.
B
Clearly,
I
use
the
word
science
here
to
reference
risk
management
as
a
deeply
studied,
organized
body
of
knowledge,
and
the
thing
that
I've
got
to
do
now
is
justify
their
claim
and
then
also
try
and
justify
why
we're
gonna
rely
on
it.
So
much
so
you
know
this
is
where
the
sub
history
lesson
starts
in
terms
of
risk.
Why
is
risk
so
profoundly
useful
in
terms
of
governance
and
in
terms
of
scientific
debate?
B
Well,
the
the
first
part
of
it
is
that
modern
concept
of
risk
actually
starts
at
about
eight
hundred
years
ago
with
the
hindu-arabic
numbering
system
reaching
the
west.
But
the
the
more
profound
aspects
of
risk
started
during
the
Renaissance
the
Renaissance
period
and
funny
enough
it
was
a
French
nobleman
that
went
to
Blaise
Pascal
and
said:
listen.
How
do
you
divide
the
stakes
of
an
unfinished
game
of
chance
between
two
players
when
one
of
them
is
ahead
and
Pascal
thought
about
for
a
while?
B
And
you
know
he
was
the
de-facto
mathematician
of
the
day
and
he
couldn't
quite
figure
out,
so
he
went
to
his
friend
a
lawyer,
but
maybe
you
might
know
this
lawyer's
name.
His
name
was
firma
and
between
Pascal
and
FEMA
they
basically,
according
to
you,
know,
Peter
Bernstein
in
his
book
against
the
guards.
They
discovered
probability
theory.
Now
that
is
the
heart
of
risk.
So
right
from
the
get-go,
you
get
a
sense
that
the
science
of
risk
management
does
have
deep
roots,
with
respect
to
quantitative
and
qualitative
nuances
and
ideas
about
probability.
B
So,
the
next
seventy
years
there
was
a
whole
bunch
of
really
cool
manifestations.
You
had
property
insurance
coming
out
as
a
result
of
this.
This
probability
theory
after
the
Great
Fire
of
1666
in
London
marine
insurance,
which
basically
was
the
precursor
to
business
insurance,
got
started
in
1680
by
this
really
odd
chap
named
Edward
Lloyd
in
a
coffee
shop-
and
you
know
some
folks
might
not
know
him
as
Lloyd's
of
London.
B
The
first
live
tables,
and
this
one
was
was
particularly
interesting
to
me
way
back
when,
when
I
saw,
it
was
actually
devised
by
Edmund,
hey
in
1693.
Why
is
this
important?
Well,
firstly,
Edmund
Halley
is
the
guy
about
Haley's
comment,
but
more
important
is
how
he
encapsulated
probability
theory
in
the
context
of
mortality,
in
the
context
of
understanding
how
long
people
will
live
now,
I
started
life
out
as
trying
to
be
an
actuary
and
I
can
tell
you
that.
B
Wasn't
the
thing
that
drove
me
to
risk,
but
it
was
something
that
I've
got
to
profoundly
understand
the
the
extent
of
risk
you
move
on.
You
find
that
these
these
expectancy
tables
they
drove
the
financial
landscape
of
England
during
the
1720s
all
the
way
through
to
the
1800s.
They
they
financed
themselves
through
life.
Annuities
de
Maire
in
1730
came
out
with
a
normal
distribution.
His
bell
curve,
all
on
the
basis
of
this
most
of
you
guys
might
know
Thomas,
Bayes,
Bayes,
theorem,
1750,
an
English
Presbyterian
minister,
basically
thought
well.
B
Let
me
put
this
into
a
decision-making
framework.
How
do
I
better
improve
my
my
decisions,
given
the
fact
that
I
have
certain
intuitive
concepts
about
what
I
think
it
should
be
extraordinarily
important,
especially
for
the
space,
especially
the
way
we
see
how
we're
going
to
manage
the
risk,
basically
fast
forward
to
1875
Francis
Goldson.
He
basically
devises
linear
regression,
1952
Markowitz.
He
put
some
real
math
to
the
notion
of
diversification.
B
But
what
is
really
interesting
is
that
risk
management
has
this
deep
and
profound
roots,
all
the
way
back
to
this
100-year
period
between
1650
and
1760
fast-forward
to
the
late
1980s,
and
this
is
a
little
bit
anecdotal,
but
it's
actually
quite
true.
It's
all
about
Value
at
Risk,
so
Bankers
Trust
were
really
the
first
company
to
think
about
Value
at
Risk
and
what
didn't
mean
from
a
financial
risk
point
of
view.
B
They
had
to
redefine
a
lot
of
their
questions
and
from
there
the
valued
risk
and
risk
management,
as
we
know
from
the
very
sort
of
contemporary
history,
that's
where
it
grew
now.
Why
did
I
bring
all
this
up?
Well,
I
wanted
to
show
you
that
this
risk
management
function
that
we're
leaning
on
so
heavily
has
got
very
deep
roots
within
probability.
Theory
within
mathematical
development
within
organizational
and
institutional.
B
Organizational
theory
as
well
everything
from
chaos,
theory
to
game
theory
to
quantum
statistics-
all
came
from
this
very
simple
period.
Now
the
reason
I
try
and
bring
this
up
and
lean
on
it.
So
heavily
is
because
we
got
to
look
at
this
and
say
well:
how
are
we
going
to
govern
assistant
based
on
risk
management?
Well,
we
have
this
wonderful
construct
of
risk
management,
literature
which
folks
can
lean
on
and
bring
to
bear
on
what
we
see
in
the
maker
universe.
B
We
will
be
able
to
find
different
points
of
view
on
how
to
assess
risk,
we'll
find
different
constructs
that
can
compete
against
each
other
and
that's
something
I'll
bring
up
again
when
we
talk
about
decentralized
risk,
but
in
the
essence
your
arguments
are
made
through
these
collections
of
risk
models,
known
as
risk
constructs
and
those
risk
models
and
and
by
extension,
risk
constructs
will
lean
heavily
on
what
has
already
been
created.
What
has
really
been
studied
and
really
is
the
case
of
saying:
well,
how
do
we
apply
them?
B
B
There
is
a
sense
of
organization
in
risk
management
that
can
actually
bear
great
fruits
if
used
appropriately
within
a
governance
framework
within
the
ability
to
actually
argue
our
points
and
come
to
resolutions
about
constructs
and
their
applicability,
not
necessarily
the
finer
points
of
like
actual
nominal
outputs,
but
after
that
little
bit
of
a
of
a
long
speech.
I'm
going
to
stop
there
any
questions,
any
comments.
A
Well,
yeah
I
do
and
I'm
afraid
I
might
have
too
many
questions.
So
stop
me
if
I'm
jumping
too
far
ahead,
I'm
always
fascinated
when
I
discover
or
I'm
reminded
of
an
entire
school
of
thought.
That
I
was,
you
know
only
tangentially
aware
of
previously,
but
now
I'm
fascinated
so
risk
management
sounds
fascinating,
but
I'm,
a
complete
outsider,
so
I'm
wondering
whether
other
compete
competing
schools
of
thought
in
risk
management.
Do
you
pick
a
way
of
looking
at
things
or
in.
B
Risk
no
in
risk
management
is
the
umbrella
for
all,
but
within
it,
you've
got
your
statistical
points
of
view
being
a
frequentist
invasion
by
the
with
respect
to
to
risk
as
well.
What
is
the
the
nature
of
the
risk
that
you
face?
I
mean
it
diversifies
across
you
know
I.
So
many
I
just
lost
my
train
of
thought
that
it
diversifies
crossed
so
many
applications
that
to
find
and
abstract
out
different
frames
of
reference
and
different
fields
of
thought.
With
respect
to
risk
management,
you
always
seem
to
go
back
to
probability.
Theory.
B
You
always
go
back
to
the
sense
of
how
do
you
go
about
decision
making?
That's
where
Bayes
came
in
so
prominently,
you
have
to
look
and
say
what
are
my
initial
thoughts
about
including
collateral
ABC
into
our
portfolio.
What
do
I
know
about
them?
The
only
way
I
can
actually
come
up
with
an
intuitive
a
priori.
B
Look
at
this
particular
collateral
type
is
by
using
a
little
bit
of
quantitative
methods
and
by
using
qualitative
assessments,
but
as
we
include
this
particular
particular
collateral
type
into
our
portfolio,
we
have
to
watch
it
carefully
and
diagnose
because
that
new
information
is
going
to
give
us
a
better
sense
of
what
we
need
to
do
to
address.
Not
only
from
that
particular
types,
point
of
view,
but
from
the
collateral
portfolio
point
of
view
as
well
and
then,
obviously,
by
extension,
to
the
system,
integrity
and
the
stability.
Ok,.
B
You
can
come
from.
You
can
come
from
two
institutional
frameworks.
I
give
you
an
example
here.
So
if
you're
sitting
at
a
life
insurer
and
you're
an
actuary,
you
probably
look
at
this
and
say:
oh,
this
is
just
extreme
value
theory.
So
we're
going
to
apply
extreme
value
theory
to
this.
It's
pretty
obvious,
we'll
do
the
necessary
calculations,
maybe
use
Pareto
distributions
for
the
tails
and
come
in
and
create
a
construct.
Then
you
might
find
someone
from
the
the
financial
engineering
side
look
at
this
and
go
oh.
That
just
looks
like
option
premium
calculations.
B
We
can,
you
know,
apply
a
stochastic
differential
equations
to
this.
We
can
maybe
resolve
them
into
the
PDS
construct
by
constructed
by
a
black
and
Scholes,
maybe
even
go
further
and
have
a
look
at
other
processes
as
well
and
and
figure
out
how
to
to
solve
this
issue.
There
are
actually
many
hubs
that
can
attack
this
problem
from
many
different
angles,
so
it
wouldn't
be
a
really
case
of
school
of
thought,
because
that's
a
little
bit
more
profound
and
abstract
stew,
you
know
probability
theory,
but
the
the
perspective
of
application.
B
You
can
do
it
from
insurance,
you
can
do
it
from
a
traders.
Point
of
view
a
financial
engineers
point
of
view.
You
can
even
do
it
from
you
know:
risk
management
where
folks
consider
things
called
real
options
like
actually
being
in
a
corporation
where
you
faced
with
tangible
up
tangible
risks
that
you
need
to
manage.
They
may
even
have
an
idea
of
how
to
approach
this
well.
A
B
This
basically
will
also
talk
to
the
the
one
aspect
that
I
wanted
to
say:
maybe,
if
I
just
go
into,
how
are
we
going
to
implement
this
scientific
government
through
decentralized
risk?
It
then
talks
to
the
the
fact
of
well.
What
is
the
risk
construct
and
there's
who
is
the
first
risk
construct,
and
how
are
we
going
to
be
a
template
for
for
folks
to
copy
and
or
improve
upon.
A
B
I
mean
there
is
going
to
be
a
repository
of
papers
that
we
are
going
to
publish
with
respect
to
our
own
constructs
and
I
hope
that
those
papers
are
heartedly
challenged
by
folks
at
the
outside
that
are
more
profoundly
well-versed
with
respect
to
specific
topics,
so
they
can
actually
take
those
constructs
and
deconstruct
them
down
to
their
basics
and
then
improve
on
them
going
to
recreate
them
in
a
most
profound
but
foundational
way.
But
yes,
the
there
is
going
to
be
along
with
these
videos.
There's
gonna
be
papers
and
eventually
a
couple
of
online
toys.
A
B
B
It's
either
going
to
be
a
case
of
having
it
very
clear
and
open
with
someone
inputs
that
particular
code
draws
down
the
information
from
either
an
external
source
like
windmark
market
cap,
and
they
can
see
the
internals
of
the
calculation
and
they
can
see
how
the
derived
liquidation
ratio
is
outdated
and,
at
the
same
time,
assumptions
what
we
talking
about
in
terms
of
liquidity.
What
are
we
talking
about
in
terms
of
wash
trading?
What
are
the
proxies?
B
We
need
to
consider
in
order
to
to
include
this
in
our
calculations
with
respect
to
the
quotation
ratio,
and
that
also
extends
to
debt
ceiling.
It
extends
to
the
stability
fee
as
well.
So
those
are
the
the
primary
aspects
that
are
going
to
be
available
in
terms
of
some
sort
of
online
models
for
folks
to
play
with.
B
We're
was
ok,
so
the
the
next
thing
is
the
scientific
governance.
How
do
we
actually
implement
this
using
decentralized
risk
function,
and
this
takes
us
back
to
you
know
recalling
what
the
risk
function
is
in
terms
of
how
it
covers
the
breadth
of
models
required
to
make
this
whole
thing
function.
So
if
we
call
that
breadth
the
horizontal
and
then
each
step
in
that
process,
we
obviously
consider
as
a
vertical.
We
can
see
that
there's
two
ways
that
we
can
create
a
better
governing
system
through
scientific
debates.
B
The
one
is,
you
may
have
competing
risk
constructs
from
competing
teams
and
if
their
risk
constructs
are
similar,
then
it's
really
a
sort
of
binary
choice.
Do
we
choose
one
above
the
other?
The
other
aspect,
and
this
talks
to
rich
what
you
said
about
different
schools
of
in
terms
of
risk
management
is,
if
you
produce
a
risk
construct
that
is
orthogonal
in
its
application
to
one
that
is
ready
used
in
a
system.
You
might
actually
weigh
that
one,
a
lot
higher,
because
it
gives
you
a
different
point
of
view.
B
Now
the
more
orthogonal
constructs
you
include
into
your
overall
system,
the
more
you
diversify
out
that
that
chance
of
not
effectively
measuring
a
risk
appropriately
or
not
identifying
the
risk
appropriately.
You
know
trying
to
move
away
from
more
of
the
same
into
something
completely
different
and
there
are
all
kinds
of
angles
out
there.
It
just
really
needs
to
be
explained
and
folks
will
then
have
a
very
good
sense
of
saying.
Okay,
well
I
like
to
approach
it
from
this
point
of
view
or
maybe
I
have
a
completely
different
point
of
view.
B
That's
a
hybrid
of
of
many
methods,
but
either
way
the
risk
constructs
always
put
forward
as
representative
arguments
of
how
the
risk
function
should
work.
So
you
have
competing
constructs
and
then
you
have
collaborative
constructs
where
you
may
have
someone
produce
a
model
that
only
fits
into
one
or
two
verticals.
It
may
require
inputs
from
other
models
within
the
sort
of
horizontal
horizontal
aspect
of
the
the
risk
function.
So
where
does
the
the
governance
play
a
crucial
role?
B
B
The
the
more
risks
actually
come
out
of
the
woodwork
that
you
know
don't
initially
seem
obvious
at
first,
which
also
talks
to
the
reason
why
I
talk
about
risk
in
terms
of
layers,
as
opposed
to
some
sort
of
categorization,
that
most
people
are
used
to
in
a
market,
risk
credit
risk,
liquidity
risk
and
so
on
and
so
forth.
But
that
pretty
much
is
the
aspect
of
how
to
at
least
from
an
outline
point
of
view,
how
you
can
actually
implement
scientific
governance
through
a
decentralized
risk
function.
A
All
right,
if
nobody
else
is
gonna
pipe
up,
I'm
going
to
ask
I'm,
not
sure
I'm,
trying
to
still
build
out
a
mental
model
of
how
scientific
cleverness
will
actually
work
and
I
know
it's
very
early
days,
and
it's
probably
premature.
But
when,
when
we're
I
mean
the
maker,
voters
gather
to
determine
what
proposal
or
what
asset
class
they
want
to
add
to
the
system.
They're
they're
kind
of
voting
on
two
different
things,
potentially
that
they're
voting
on
the
quality
of
the
model
and
the
results
of
that
model
produced
is
that
is
that
correct.
B
Take
a
step
back,
what
was
first
gonna
happen.
Is
the
governor's
are
going
to
decide
on
the
inclusion
of
a
risk
team?
Sure
that's
the
first
thing:
that's
going
to
happen,
regardless
of
the
the
collateral
types.
The
arguments
they
put
forward
have
to
be
convincing.
Now,
there's
a
bit
of
salesmanship
that
needs
to
be
incorporated,
unfortunately,
just
because
you're
the
most
profoundly
smart
person
on
the
face
of
the
planet.
It
means
nothing
if
you
can't
communicate
it
appropriately.
B
So
that
is
the
aspect
that
is
really
important
is
to
find
the
intersection
between
risk
teams
that
can
actually
produce
something
that
is
practical
and
is
well
understood,
because
governor's
will
rely
on
their
input
as
to
what
certain
collateral
risk
profiles
are
going
to
look
like.
So
if
you
have
a
a
risk,
construct
and
I
have
a
risk
construct
and
they're
slightly
different,
but
we
included
in
the
overall
risk
function
of
maker
and
we
decide
we
have
a
look
at
token
ABC.
B
B
What's
a
risk
profile,
it's
the
risk
parameters
that
will
be
implemented
in
the
system
on
the
next
executive
vote,
which
is
something
I'll
also
touch
on
a
bit
later
in
this
in
this
call,
but
so
the
the
maker
token
holders
point
of
view
is
twofold:
one:
the
trust
that
they
put
into
the
risk
teams
representing
the
risk
profile,
calculation
and
then
two
the
choice
on
actually
going
forward
with
accepting
the
collateral
type
with
those
risk
profiles.
So
the
risk
function
is
more
advisory,
it
doesn't
ultimately
say:
listen
make
its
open
holders.
B
A
Okay,
I
think
that
clears
it
up
a
bit
and
it
leads
a
bit
to
my
next
question:
the
traditional
sort
of
scientific
paper
and
there's
data
models,
and
then
people
assign
a
confidence
level.
You
can
kind
of
figure
out
how
confident
they
are
and
the
results
that
they've
they've
predicted
or
promoted
I'm
wondering
if
there's
the
same
sort
of
confidence
level
that
exists
in
in
the
risk
management.
Oh.
B
Yeah
I
mean
the
whole
thing
is
based
on
probability,
theory,
so
sampling
and
you
know
doing
the
proper.
The
appropriate
hypothesis
testing,
which
actually
brings
me
to
another
point
as
well,
is
that
you
know
the
the
one
thing
I
didn't
touch
on
with
scientific
governance.
Is
the
scientific
method
and
I've
intentionally
not
touched
on
it,
because
in
this
day
and
age,
he's
trying
to
see
the
deep
learning
he's
trying
to
take
a
place
and
a
very
functional
place
in
most
of
the
industry,
and
that
generally
doesn't
have
a
scientific
method
behind.
B
It
is
very
empirical-
and
it
literally
is
a
case
of
saying.
Well,
here's
a
whole
bunch
of
data.
You
know
mr.
deep
deeply
at
Network
new
network.
Please
go
figure
it
out.
For
me,
it's
very
black
box,
but
the
point
being
is
not
to
disregard
it,
because
it's
something
you
don't
understand.
We
have
to
include
it,
but
just
know
very
well
that
the
scientific
method
is
not
necessarily
applied
within
these.
These
constructs.
B
So
that's
why
I
didn't
use
it
as
a
origin
or
basis
of
argument,
but
in
terms
of
confidence
in
a
particular
system.
There's
two
ways:
there's
a
back
testing
formality
where
you
say:
well,
you
are
proposing
a
certain
construct.
How
has
it
worked
in
the
past?
How
has
it
worked
over
certain
regimes?
These
are
the
types
of
arguments
that
you'll
find
coming
out
of
the
debate
amongst
risk
teams.
I
have
a
wonderful
liquidity,
valued
liquidity,
adjusted
value
at
risk
model.
It
works
wonderfully.
That's
awesome.
How
did
how
did
to
work
over
the
2008
period?
B
There's
the
debate
and
folks
will
come
out
and
say:
oh
well,
yeah
I
didn't
really
pick
it
up
that
well
or
it
looks
like
you
may
be
overfitting.
That
dates
are
a
bit
too
much.
I
think
your
model
might
be.
You
know
a
little
bit
biased
with
respect
to
those
sort
of
regimes.
That's
where
the
argument
comes
from
moving
forward.
Now
we're
talking
about
out-of-sample
data.
It's
almost
like
you're
doing
a
cross-validation
moving
forward
this
one's
a
little
bit
more
long-term.
B
Metrics
that
allow
you
to
separate
a
model
that
is
working
from
a
model
that
is
failing,
but
unfortunately
there
isn't
a
well-established
predictive
construct
that
will
allow
you
to
get
out
of
a
model
in
time.
You
generally
find
that
when
things
go
wrong,
it
really
is
a
case
of
how
far
in
two
things
going
wrong.
You
are
before
you
start
changing
things
up
well,.
A
That
raises
another
fascinating
kind
of
subject
for
me:
I
think
so
what
happens?
If
somebody
proposes
a
model
of
a
framework
construct,
they
assign
a
value
to
a
rating
to
a
particular
asset
class.
The
asset
class
gets
voted
in
by
the
maker
holders
and
it
turns
out
that
that
model
was
incorrect
and
that
asset
class
behaves
far
differently
than
predicted.
What
would
be
the
next
step
after
that.
B
Well,
that
would
be
awaiting
against
that
risk
construct
and
at
risk
team
right
now.
It
also
depends
on
the
severity
of
it.
I
mean
if
this
is
a
cataclysmic
thing,
then
it's
it's
pretty
easy
to
say
well,
either
you
didn't
spot
the
obvious,
maybe
jump
diffusion
construct
that
is
unlined
this
this
kind
of
token
or
it's
just
simply,
that
is
completely
out
of
the
purview
of
everyone's
domain
space.
There
is,
and
this
this
also
talks
to
something
called
economic
capital.
B
You
can
apply
wonderful
risk
management
techniques,
but
they
only
encapsulate
as
much
as
you
can
statistically
understand.
They
are
going
to
be
events
that
are
full
way
out
of
your
statistical
understanding,
of
your
distributional
analysis
and,
consequently,
you're
going
to
find
that
your
you
are
facing
something
that
is
the
defect
of
Black
Swan,
and
you
hope
that
your
distributional
understanding
and
your
construction
of
your
collateral
portfolio
will
be
such
that
it
will
buffer
you
against
this.
B
A
The
model
is,
we
find
it's
off
and
it's
impacting
potentially
the
long-term
value
of
maker,
and
that's
it's
something
that
needs
to
be
addressed
and
sometime
in
the
next
month
or
two.
Will
there
be
sort
of
reuse
of
that
same
model
or
will
we
need
a
new
model
or
will
we
just
artificially
begin
tweaking
parts
of
that
risk
profile
to
sort.
B
Of
well
system
the
nice
thing
about
risk
management,
especially
financial
risk
management
as
a
whole.
The
the
taxonomy
of
models
is
pretty
clear.
How
you
apply
those
models
is
obviously
another
thing
entirely
so
to
be
able
to
assess
to
be
able
to
assess
the
models.
Applicability
has
been
well-defined
if
you
want
to
know
if
you've
missed
a
specific
spike
in
risk
or
a
specific
spike
in
movement
that,
as
that,
can
be
done
and
can
be
defined
appropriately,
but
there
isn't
going
to
be
I
think
my
my
comments
are
trying
to
make
here
very
badly.
A
I
was
muted,
I
was
just
pointing
out
that
I
can
I
can
feel
new
paradigms
arriving
as
you
talk,
so
I
need
to
digest
some
of
this
stuff.
Jordan
has
a
great
question.
I'm
going
to
read
it
out
with
the
weighted
risk
constructs.
Obviously
it
will
be
good
to
have
that
process
as
transparent
as
possible,
but
I
assume
that
the
models
will
not
actually
live
on
chain
for
the
executive
votes
will
make.
Our
holders
have
access
to
all
the
models
that
produce
the
outcomes
that
are
voted
on.
Yes,.
B
A
B
Well,
there's
two
points
of
reference,
yet
the
first
one
is
that
models
should
be
explicit.
In
other
words,
there
should
be
completely
transparent,
for
the
obvious
reason
that
you
and
I
say
this.
You
know
with
with
the
tone,
for
the
obvious
reason
that,
in
order
to
really
get
to
evaluate
a
model,
you
really
need
to
see
the
guts
of
it,
but
you
are
going
to
find
that
there
are
risk
teams
that
don't
want
to
show
you
their
models
and
those
are
going
to
be.
B
You
know
more
implicit,
unfortunately,
because
of
that
there
waiting's
are
going
to
come
down,
so
you
might
find
that
as
an
example,
a
well-respected
company
out
there,
let's
say
or
in
certain
young,
they
have
a
risk
team
and
their
risk
team
is
contributing,
but
they
think
that
their
IP
is
extraordinarily
important.
They
don't
want
to
divulge
what
they
what
they
have,
but
the
back-testing
and
the
other
sample
testing
show
it
shows
that
they
are
particularly
good
at
doing
this,
which
is
great,
you
can
include
them,
but
you're
gonna
have
to
wait
them
a
little
lower.
B
Maybe
a
lot
lower
than
the
other
constructs
that
you
have
on
board.
I
go
by
the
thesis
that
it
is
better
to
know
that
which
is
working
on
average.
Then
you
know
not
to
know
something:
that's
working
fabulously.
It
goes
to
that
black
box
thinking
and
especially
during
this
sort
of
nascent
period
of
our
organization.
I
think
I'm
more
open
to
the
fact
of
saying.
Please,
if
you
want
to,
if
you
want
to
contribute
to
construct,
make
sure
that
it's
open
enough
for
folks
to
be
able
to
understand,
share,
compete
and
collaborate.
A
Yeah,
it's
interesting.
We
need
to
build
up
a
body
of
knowledge,
that's
that's
useful
for
our
ecosystems,
so
open
this.
More
importantly,
teaching
me
the
second
part
of
Jordan's
question:
do
you
envision
the
models
living
on
chain
and
having
the
authority
to
make
iterative
changes
to
the
core
system
or
are
the
quarterly
executive
votes
intended
to
be
the
long-term
strategy
for.
B
Now
I
would
say
that
the
executive
boats
all
the
way
the
way
forward.
It's
not
you
know
you,
you
need
a
standardized
protocol
for
risk
contracts
before
you
can
even
think
about
that,
and
you
know
contributing
towards
that
will
be
the
the
first
risk
construct
from
the
internal
risk
team.
You
know
creating
the
template
may
be
a
standardized.
A
protocol
will
be
devised
off
the
back
of
it
by
folks.
You
know
much
smarter
than
I,
but
at
the
moment
in
time
it
will
be
through
the
executive
votes,
cool.
A
I
can
add
something
here
to
Jordan
the
the
core
components
of
the
maker
Dao
system
are
essentially
they're
designed
to
be
untouchable
and
less.
There
is
some
kind
of
existential
threat
to
the
system,
so
the
core
components
of
know
the
model
always
stay
the
same
and
things
like
the
risk
parameters
or
what
we
can
change
and
those
are
the
things
I
believe
that'll
be
affected
primarily
by
the
new
asset
classes
and
that
risk
models
that
the
risk
team
will
come
up
with
IV.
If
you
had
some
questions,
if
you
want
to
jump
in
and.
C
So
this
may
sound,
really
stupid.
Cut
me
off
if
I'm
going
to
too
long
here
stevia.
Is
it
fair
to
say
that
there
are
only
three
or
I
guess
four
things
to
consider
the
governance
feed
the
debt
ceiling,
the
percentage
of
collateral
ratio
and
then,
lastly,
which
which
particular
coin
will
be
listed
as
collateral?
Those
are
the
four
things
that
make
our
holders
can
vote
on.
Is
that
right?
Well,.
B
Three
there's
the
debt
ceiling,
liquidation
ratio
and
stability
fee.
Those
are
the
three
ok,
those
the
primary
risk
parameters
that
you
need
to
consider
when
you're
voting
on
a
collateral
type.
So
what
do
you
need?
Two
votes
on.
You
need
two
votes
on
the
wrist
team
and
the
risk
construct
and
that
risk
teams
construct
will
produce
the
values
for
those
three
parameters
I
just
mentioned,
so
you
have
to
vote
on
the
wrist
team.
B
C
Wonderful
and
just
gonna
kind
of
go
off
that
so
you
know
when
I,
when
I
look
at
like,
for
example,
etherium,
if
you
could
just
walk
me
through
the
layers
of
risk.
So
if
I'm
just
tell
me
if
I'm
kind
of
in
the
right
ballpark
here,
you
know
we
have
risks.
Obviously
you
know
the
market
tanks
the
price
goes
down,
but
but
that
could
stem
from
anything.
C
So
one
risk
is
that
the
government
says
ether
is
bad
or
whatever,
and
that
will
go
a
little
bit,
but
a
deeper
risk
is
something
like
Vitalik
gets
in
an
accident
and
dies
right.
So
that's
an
interesting,
lower
level
risk
than
the
higher
level
stuff,
which
is
like
a
newspaper
article.
We
also
have
like
a
smart
contract
bug.
You
know
some
quantum
computer
hack,
so
we
have
these
different
sort
of
layers
here.
C
So
you
know
with
something
like:
let's
say:
digits
gold
I,
don't
see
how
the
price
of
a
digits
token
could
go
down
any
more
than
let's
say
5
or
10%
in
any
given
week,
with
the
only
one
risk
exception
which
is
ether
or
digits.
Gold
team
gets
attacked
in
a
way
where
it
would
probably
go
much
much
lower
in
a
kind
of
bit
connect
style.
You
know
collapse
like
98
percent
in
a
couple
hours.
Is
that
right
to
say?
Yes.
B
What
you,
what
you
would
need
to
do
is
actually
formulate
your
risks
that
you
have
identified
into.
There
is
categories
for
yourself,
I
would
suggest,
there's
the
the
natural
taxonomy
of
risk
categories
market
risk
very
closely
associated
with
that
is
liquidity
risk.
Something
on
the
side
would
be
obviously
credit
risk,
but
the
one
that
you
focusing
on
right
now
is
operational
risk.
B
Now,
there's
the
one
thing
that
the
global
association
of
risk
professionals,
which
is
an
association
of
financial
risk
managers
around
the
world,
they've
studied
this
intensely,
and
they
found
that
the
most
prominent
and
the
most
impactful
risk
to
any
institution
is
operational
risk.
By
far
now,
that's
where
a
qualitative
assessment
of
a
collateral
type
is
so
necessary.
B
Tech
oriented
is
that
if
you
have
a
look
at
these
forking
processes,
it's
all
very
good
and
well,
but
ultimately
the
results
in
fork.
You
will
actually
have
a
higher
exposure
to
operational
risk
at
that
point
in
time,
but
it's
it's
incredibly
hard
to
quantify
operational
risk.
They
are
guys
in
the
traditional
space
that
have
opened
up
offices
to
SRA
businesses,
to
try
and
encapsulate
that
sort
of
risk.
But
it's
intensely
hard.
It's
called
high
impact.
Excuse
me
high
impact,
low
frequency
or
low
impact
high
frequency.
B
These
two
considerations
are
the
cornerstones
to
operational
risk,
but
they
are
very
specific
and
sorry.
They
really
are
unique
to
each
institution,
they're
not
easily
transferable
in
terms
of
impact
and
risk
to
other
institutions,
even
though
folks
that
they
are
trying
to
collect
databases
of
all
these
risks
get
some
sort
of
a
lease
term
modeling
construct
out
of
it.
C
C
C
C
So
they're
very
uncorrelated,
furthermore,
they're
also
uncorrelated
in
the
tight
because
aetherium
is
there
are
people
behind
aetherium
and
there's
like
a
team
and
they
could
get
hurt
or
they
could
change
their
minds
and
influence,
but
with
with
gold
there's
not
a
gold
emperor
in
the
world.
That
could
do
something
and
say
something
really
atrocious.
The
only
thing
they
share
is
the
operational
risk
of
the
etherium
network.
C
So
when
did
you
scold
tokens?
I
would
say
that
and
again
correct
me
if
I'm
wrong
here,
but
a
low
collateralization
ratio
makes
a
lot
of
sense,
because
gold
is
not
very
volatile
compared
to
theorem,
maybe
a
hundredfold
less
volatile,
and
you
know
it
has
almost
no
correlation
with
aetherium,
but
I
would
also
suggest
a
low
debt
ceiling,
because
if
the
digits
token
went
down
it
would
go
down
99%
it
wouldn't
go
down.
You
know
20%.
It
would
because
the
only
way
that
it
would
have
a
crisis
is,
it
would
be.
C
B
You
are
talking
directly
to
the
second
part
of
the
article
that
I'm
going
to
release
and
with
respect
to
debt
ceilings,
and
this
is
just
a
little
bit
of
an
inside
edge.
There
are
three
aspects
to
it
right,
and
this
is
very
much
related
to
the
liquidity
of
a
particular
token,
because
that
is
the
essence
of
why
we
actually
collateralizing
these
tokens.
B
The
one
is
the
sort
of
theoretical
debt
ceiling
you
you
taking
to
mine
this
consideration
of
creating
an
investment
portfolio
where
you
want
to
allocate
appropriately
to
create
the
best
diversification
benefit
you
can
to
improve
your
risk,
adjust
risk,
adjusted
rates
of
return.
It
comes
straight
from
the
traditional
space
and
that
allocation
will
effectively
be
a
debt
ceiling.
So
if
we
don't
look
at
digits
as
a
token
on
a
theorem,
but
we
look
at
it
as
gold,
we
would
probably
give
it
a
massive
debt
ceiling.
B
But
when
you
have
a
look
at
where
the
rubber
hits
the
road,
when
you
have
a
look
at
the
liquidity,
the
market
support
the
ability
to
actually
get
out
of
the
position.
You
start
reducing
that
debt
ceiling
immensely
to
a
level
that
cannot
be
outside
of
a
certain
percentage
of
their
issuance,
so
think
for
a
second
here
now,
if
it
did,
ryx
is
total.
Issuance
e
is
five
million,
and
five
million
was
used
as
collateral
in
our
system.
B
That
would
be
an
awful
place
to
be
because
you're
just
going
to
create
your
own
liquidity
liquidation
crisis,
because
if
something
had
to
happen-
and
you
had
to
liquidate
you'd
have
to
liquidate
the
entire
issue
to
nobody,
that's
an
extreme
example,
but
basically
that's
one
way
of
getting
to
zero
very
quickly.
So
you
need
to
constrain
yourself
to
a
certain
percentage
of
the
issuance
now.
What
is
that
a
certain
percentage
that
certain
percentage
is
actually
a
function
of
how
you
can
get
out
of
your
position
of
a
given
period
of
time?
B
It's
called
a
project
and
execution
trajectory
optimization
problem.
It
sounds
like
a
whole
bunch
of
really
silly
ways
to
put
together,
but
it
really
just
boils
down
to
how
does
the
quiddity
adjusted
value
working
to
your
risk
assessments?
It's
all
very
well
to
say:
I
am
super.
Rich
I've
got
a
hundred
million
dollars
worth
of
token
ABC,
but
if
you
had
to
try
and
get
rid
of
those
tokens
tomorrow,
you
and
Anna
put
maybe
seven
million
right.
B
B
If
you
don't
have
the
ability
to
use
the
die
in
a
very
granular
and
refined
way
on
the
other
side
of
your
system,
so
you're
going
to
have
an
inflection
in
your
operational
risk
and
that's
what
I
spoke
about
in
the
beginning.
With
respect
to
those
three
layers,
the
operational
risk
is
extrordinary
important.
If
one
is
ahead
of
the
other
in
terms
of
the
development
you're
going
to
have
a
fracture
and
that
fracture
is
going
to
expose
you
to
operational
risk,
everyone
has
it.
Everyone
is
exposed
to
it.
It's
just
here.
C
Okay
and
just
one
last
thing
here,
so
is
it
fair
to
say
that
you
know
the
debt
ceiling?
It
seems
that
there
are
two
things
to
consider:
there's
the
debt
ceiling
with
respect
to
the
collateral
type.
So,
for
example,
you
don't
want
to
have
a
debt
ceil,
like
you
said
where
all
of
the
gold
tokens
in
the
entire
digit
system
is
all
locked
up,
because
then
there's
no
liquidity,
but
there's
also
the
debt
ceiling
with
respect
to
other
debt
ceilings
in
the
portfolio
as
a
ratio
drink.
B
Yeah,
that's
where
your
allocation
comes
in,
where
what
you
borrow
from
the
traditional
finance
base
is,
you
know,
effectively
portfolio
construction,
we're
borrowing
a
lot
of
tools
from
there
where
you
say
well,
how
do
you
start
building
a
portfolio?
That's
going
to
generate
a
certain
amount
of
return,
given
a
certain
amount
of
risk,
and
you
can't
have
too
much
exposure
to
one
over
the
other,
because
you
start
messing
around
with
your
risk
adjusted
return.
B
The
same
concept
applies
to
building
the
the
collateral
portfolio,
except
there
are
a
heck
of
a
lot
of
constraints
in
the
crypto
space
that
are
not
in
the
traditional
space.
Well,
on
being
the
disparity
and
liquidity
between
tokens,
the
other
one
in
terms
of
what
I
call
the
quality
vectors,
how
do
you
actually
get
out
of
a
position,
and
what
does
that
mean?
What
does
a
liquidation
event
mean?
Does
it
mean
getting
out
so
can
a
VC
into
token
1
2
3?
C
C
B
So
basically,
you
you've
already
defined
a
couple
of
components
to
what
an
interest
rate
is
I
mean
your
interest
rates
has
to
compensate
for
a
couple
of
things.
The
first
is
inflation,
because
you
want
to
obviously
keep
your
purchasing
power
then
the
second
is
your
credit
risk
like
what
are
you
alternately?
B
What
are
you
painful
now
in
the
maker
Dow
domain?
You
obviously
have
the
dilation
area
aspect
of
the
MKR
token
being
kind
of
a
last
resort.
Now
that
is
a
type
of
credit
risk
that
needs
to
be
appropriately
remunerated
and
comes
by
way
of
a
credit
premium.
So
you'll
have
inflation,
you
have
a
credit
premium
and
then
you
have
the
the
real
interest
rate,
which
covers
the
actual
cost
of
running
an
organization.
So
these
are
the
three
parts
that,
when
added
together
would
become
your
stability
fee.
Now,
where
does
the
collateral
type
work
into
this?
B
Well,
the
collateral
type
works
into
this
in
terms
of
the
premium,
the
credit
premium
component,
because
you
may
have
a
wonderful
liquidity,
adjusted
value
at
risk.
That
tells
you
that
there's
only
1%
chance
of
you
losing
$1,000,000,
okay,
but
that's
just
a
theoretical
idea,
because
it's
based
on
an
empirical
notion
of
what
the
distribution
is
so
there's
this
whole
theory
that
came
out
called
expected
shortfall
which
a
lot
of
well-versed
quantitative
insurers
know
as
extreme
value
theory.
B
You
find
that
if
you
do
go
past
a
certain
threshold
in
risk,
you
may
expose
yourself
to
a
completely
new
distribution
of
risk,
and
that
might
be
a
lot
more
than
you
consider
and
what
we
do
there
is
we
actually
have
to
buffer
it
with
something
called
economic
capital.
So
what
an
investment
bank
does
they
do?
All
this
trading
they've
got
all
this
risk
management,
but
they've
got
this
excess
capital
to
cover
what
they
expect
would
go
wrong
when
things
go
really
wrong.
So
it's
a
really
strange
compound,
a
definition.
B
It's
risk
on
top
of
risk.
Now
that
economic
capital
is
effectively
how
much
MKR
dilution
wood-based,
how
much
MKR
token
holders
would
be
exposed
to
in
terms
of
dilution
of
per
collateral
type
and
that
will
attract
a
credit
premium.
It
won't
be
very
high.
It'll
be
like
half
a
percent
or
60
basis
points
or
whatever
the
case
may
be,
but
it
will
be
included
into
that
stability
fee,
so
the
more
tail
risk.
A
specific
token
has
the
higher
the
stability
theorem
will
be
due
to
the
credit
premium
component.
A
You
have
some
AV
problems,
no
problem
I'll
go
ahead
and
read
out
your
question
for
you
Ashley.
Will
there
be
a
situation
where
there
is
one
asset
that
may
have
two
or
more
risk
profiles,
I
hire
collateralization
ratio
and
allure
stability
fee
or
a
lower
clatter
ization
ratio
and
a
higher
stability
fee
I
would.
B
Think
so
I
would
think
that
you
may
have
different
risk
constructs
proposing
a
risk
profile
that
will
be
slightly
different
and,
as
a
consequence,
you,
you
use
the
the
collective
ensemble
to
maybe
come
out
with
your
ultimate
effective
risk
profile.
So
if
you
do
have
one
with
construct
telling
you
that
debt
ceiling
should
be
no
150
million
another
one
says
it
should
be
70
the
first
one
says:
liquidity
ratio
should
be
120
percent
and
the
second
one
says
150
percent
there's
a
weighting
that
you
have
to
give
to
each
of
these
constructs.
B
That
will
then
result
in
an
overall
calculation
representative
calculation.
For
that
particular
token,
and
then
it
is
up
to
the
MPR
token
holders
to
look
at
that
overall
number
or
if
they
wish
just
to
rely
on
one
risk
construct,
above
the
other.
It
is
really
up
to
them.
This
whole
thing
at
the
end
of
the
day,
is
an
advisory
/,
consulting
idea,
all
right.
This
fascinators.
B
A
Well,
that's
interesting
I
want
to
revisit
that
at
the
next
meeting,
actually
we're
at
one
minute
past
the
hour,
so
we
need
to
shut
it.
Fancy.
You
have
a
great
question.
Jordan
you
had
another
great
question:
I've
added
those
to
the
list
for
the
next
meeting
and
I
also
want
to
revisit
what
Ashley
just
raised
as
well
and
I
have
some
questions
of
my
own
that
we
didn't
get
to
you
so
Thank
You
Steven.
That
was
those
amazing
I
appreciate
that.
Thank.