Will Quant Finance End Up Like Data Science

Summary

In this video, Dimitri discusses the future of quantitative finance and data science, addressing the evolving educational requirements and industry standards. He critiques the degradation of data science into simple analytics, contrasting it with the more rigorous and structured field of quantitative finance. Dimitri predicts a continued demand for master's degrees in quantitative finance due to the complexity of model building and the high stakes involved, while data science may see a bifurcation into less technical roles and those requiring advanced degrees. He also touches on the cultural divide between data science and traditional statistics, advocating for a more integrated approach to utilizing statistical tools.

Takeaways

• 🔍 The future of Quantitative Finance is being shaped by evolving educational requirements, with a shift from needing a master's degree to potentially entering the field with an undergraduate degree.
• 📈 There's a growing debate on the qualifications needed for a Quant role, with some suggesting that the field is becoming more accessible to those without advanced degrees.
• 🎓 Data science is facing its own challenges with credentialing, as the field has seen a dilution in the definition of what constitutes a data scientist, with some roles now being filled by those with basic data analysis skills.
• 📊 The speaker criticizes the data science field for seeking shortcuts and quick solutions, rather than a deep understanding and rigorous approach to model building and analysis.
• 🏢 Companies are redefining job titles to reflect the changing nature of data-related roles, with titles like 'machine learning engineer' or 'AI specialist' becoming more common.
• 📉 The 2007-2008 financial crisis impacted the perception and requirements of Quant roles, leading to a decrease in the perceived need for advanced degrees in the field.
• 🧑‍💼 There's a bifurcation in Quant roles, with junior positions handling more basic data analytics and senior roles being filled by those with PhDs or advanced expertise.
• 💼 The speaker predicts that Quant Finance will maintain a higher educational bar compared to data science, due to the critical nature of the models and the financial risks involved.
• 🔧 The data science community is described as somewhat toxic and resistant to integrating traditional statistical methods, preferring to focus on newer techniques and tools.
• ⏳ The speaker anticipates that it may take 20 years or more for the data science field to mature and fully integrate traditional statistical methods into its practices.

Q & A

What is the main concern raised by the subscriber's question about the future of quantitative finance?

The main concern is about the evolving requirements for entering the field of quantitative finance, particularly whether a master's degree will still be necessary in the future, as the field seems to be opening up to undergraduates.

What does the speaker think about the current state of data science?

The speaker believes that data science has become a 'massive joke', with the field being degraded to the point where basic data analytics and the use of tools like Excel are being equated with data science.

Why does the speaker think that data science has been degraded?

The speaker thinks data science has been degraded because many people are looking for shortcuts and quick solutions, leading to a dilution of the skills and knowledge required in the field.

What is the speaker's view on the role of data science in the job market?

The speaker views data science as a field that is becoming more segmented, with less technical roles not requiring a master's degree and more technical roles starting to require advanced degrees again.

How does the speaker describe the evolution of quantitative finance in terms of job requirements?

The speaker describes quantitative finance as a more mature field that has already gone through a phase of dilution in job requirements but has now stabilized, with a clear distinction between junior and senior quantitative roles.

What is the speaker's opinion on the necessity of a master's degree in quantitative finance?

The speaker believes that a master's degree will continue to be a minimum requirement for rigorous roles in quantitative finance, as the field requires a high level of skill and knowledge.

How does the speaker compare data science to quantitative finance?

The speaker compares data science to quantitative finance as a 'little brother', suggesting that data science is more general and less mature, while quantitative finance is more structured and defined.

What does the speaker think about the culture within the data science and machine learning community?

The speaker views the culture within the data science and machine learning community as toxic, with a tendency to be anti-statistics and to focus on quick solutions rather than robust, reliable models.

What is the speaker's prediction for the future of data science and machine learning?

The speaker predicts that data science and machine learning will eventually mature and merge back with traditional statistics, utilizing all the available tools effectively.

Why does the speaker believe that banks and firms are reluctant to hire undergraduates for quantitative roles?

The speaker believes that banks and firms are reluctant to hire undergraduates because the quality of models produced by less experienced individuals is often not up to the standard required for the high-stakes decisions in quantitative finance.

Outlines

00:00

🔮 The Future of Quantitative Finance

Dimitri discusses the future of quantitative finance in response to a subscriber's question. He highlights the evolution of educational requirements for data scientists and the potential trajectory for quants. Dimitri expresses skepticism about the devaluation of data science as a field, noting the shift from requiring a master's degree to accepting undergraduate degrees. He criticizes the industry's dilution of what constitutes a data scientist, suggesting that the role has been reduced to basic data analysis. Dimitri also touches on the maturity of quantitative finance compared to data science, suggesting that quant finance has already undergone a cycle of devaluation and is now more structured.

05:00

📉 The Dichotomy of Quantitative Roles

In paragraph two, Dimitri delves into the bifurcation of quantitative roles within the finance industry. He describes how firms are separating junior and senior quantitative positions, with juniors handling data analytics and seniors focusing on complex model development. Dimitri argues that while junior roles might be filled by undergraduates, the senior roles increasingly require PhDs due to the complexity and rigor required. He also discusses the reluctance of firms to invest in extensive training programs for undergraduates, preferring to hire those with advanced degrees who are already well-prepared.

10:00

📈 The Role of Data Science in Finance

Paragraph three addresses the role of data science within the broader context of finance and machine learning. Dimitri views data science as a subset of machine learning and AI, which in turn are subsets of statistics. He criticizes the data science community for being insular and resistant to traditional statistical methods. Dimitri also discusses the cultural divide between data scientists and statisticians, suggesting that data scientists often lack a deep understanding of the models they create. He emphasizes the importance of model robustness and reliability in finance, contrasting the data science approach of rapid model development with the more cautious and rigorous approach required in quantitative finance.

15:01

🏦 The Economic Reality of Hiring in Quantitative Finance

In the final paragraph, Dimitri reflects on the economic considerations of hiring within quantitative finance. He notes that banks and firms are often looking to hire undergraduates to save on the costs associated with higher degrees. However, he argues that this approach rarely results in high-quality models due to the complexity of the work. Dimitri suggests that the industry may be moving towards a segmentation where more technical roles require master's degrees, while less technical roles might suffice with undergraduate degrees. He concludes by reiterating the need for the data science community to mature and integrate more traditional statistical methods into its practices.

Keywords

💡Quantitative Finance

Quantitative Finance, often abbreviated as 'Quant', refers to the use of mathematical and statistical techniques to manage financial risks in a firm and to make informed financial decisions. In the video, the speaker discusses the evolution of the field and the changing educational requirements for professionals entering the field, noting a shift from requiring a master's degree to potentially accepting undergraduates.

💡Data Science

Data Science is a multidisciplinary field that uses scientific methods, processes, algorithms, and systems to extract knowledge and insights from structured and unstructured data. The speaker in the video critiques the field for becoming oversaturated and for the degradation of its standards, suggesting that what was once a rigorous discipline has been reduced to basic data analytics in some cases.

💡Master's Degree

A Master's Degree is an academic degree awarded by universities or colleges upon completion of a course of study demonstrating mastery of a specific field of study or area of professional practice. The video discusses the historical requirement for a Master's Degree in Quantitative Finance and the potential shift towards accepting undergraduate degrees.

💡Undergraduate

An undergraduate is a student who is pursuing a bachelor's degree at a college or university. The script mentions the possibility of Quantitative Finance roles becoming accessible to undergraduates, which contrasts with the traditional requirement of a Master's Degree.

💡Machine Learning

Machine Learning is a subset of artificial intelligence that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. The speaker discusses the role of machine learning in data science and its impact on the job market, suggesting that the field is becoming more segmented and specialized.

💡Data Analytics

Data Analytics is the process of examining, cleaning, transforming, and modeling data to extract useful information, draw conclusions, and support decision-making. The video script uses this term to describe the more basic tasks that some data scientists are now performing, which contrasts with the more complex model-building tasks traditionally associated with the field.

💡Excel

Microsoft Excel is a widely used spreadsheet program for data analysis and management. The speaker humorously points out that in some cases, proficiency in Excel is now considered a form of data science, indicating a perceived lowering of the bar for what constitutes expertise in the field.

💡Econometrics

Econometrics is the application of statistical methods to economic data. It helps in determining the relationship between economic variables. The video mentions a disdain for econometrics within some data science circles, suggesting a disconnect between traditional statistical methods and newer data science practices.

💡Model Development

Model Development in the context of the video refers to the creation of mathematical or statistical models for predicting outcomes or understanding complex systems. The speaker emphasizes the importance of rigorous model development in Quantitative Finance, as opposed to the more ad-hoc approach sometimes seen in data science.

💡MLOps

MLOps, a portmanteau of 'Machine Learning' and 'DevOps', refers to a set of practices for managing the entire lifecycle of machine learning systems. The video script touches on the idea that some data scientists might rush to develop models without proper testing or consideration for long-term reliability, leading to a need for MLOps practices to ensure stability.

💡Statistician

A Statistician is a professional who works with the collection, analysis, interpretation, presentation, and organization of data. In the video, the speaker suggests that some of the tools and methods used by data scientists are not new, but rather traditional statistical approaches repackaged, and that a true understanding of these methods requires the training of a statistician.

Highlights

The future of quantitative finance and the evolving requirements for data scientists are discussed.

Historically, a master's degree was the minimum requirement for a career in quantitative finance.

The perception of data science as a shortcut to success and the degradation of its standards are critiqued.

The blurring lines between data analytics and data science, with Excel and GBA now considered part of data science.

The distinction between data science and quantitative finance, with the latter being more mature and structured.

The impact of the 2007-2008 financial crisis on the perception and requirements of quants.

The trend of firms hiring PhDs for complex model building in quantitative finance.

The segmentation of data science into less technical roles requiring only an undergraduate degree.

The critique of the data science community for its resistance to traditional statistics.

The need for data science to mature and integrate with traditional statistics for long-term success.

The potential for data science roles to evolve into more specialized titles like ML engineers.

The challenges faced by banks in hiring undergraduates for quantitative roles due to the complexity of tasks.

The importance of rigorous education and training for those in quantitative finance roles.

The author's perspective on the necessity of a master's degree for quantitative finance roles.

The author's prediction that quantitative finance will not follow the trend of lowering educational requirements.

The call for a more responsible and rigorous approach to data science and machine learning practices.

The author's closing thoughts on the future of data science and quantitative finance.

Transcripts

foreign

hey YouTube It's Dimitri and today we're

going to answer a subscriber's questions

specifically on the future of

quantitative Finance here so I was going

to read this whole thing but I don't

want to bore you too much essentially

what the question is here which I'll put

on the screen

um is they've been talking about you

know what do you think is going to

happen to the future of quantitative

Finance uh you know you speak extremely

difficult to become a data scientist and

there used to be this minimum required

requirement to have a master's degree

and then it's kind of evolved in change

and now it's basically like you can get

in with an undergrad

uh but you know they're kind of

wondering what's going to happen with

Quant finance and essentially like

they've seen you know kind of mixed

perspectives on you know do you need a

masters do you not need a master's in

Quant Finance in the future where do I

think this is kind of going to go

um and they understand you know that you

know for data scientists a lot of

companies don't know what they really

are and so that kind of muddies the

water a little bit

um so basically what are the minimum

requirements here of a Quant in the

future

these videos you guys ask the tough

questions uh honestly it's because data

science I think is a massive joke

um I'm gonna find so many people

watching this but that's okay

um the longer I study data science the

more data scientists I work with uh the

more I realize it's like everybody just

wants a shortcut and everybody wants the

get rich quick scheme and I don't blame

actual true data scientists per se

I'm going to put a little pin in that

for now because there's really yeah it's

just been degraded down to nothingness

like data analytics is the same as data

science now uh using Excel and GBA

apparently is now data science and

everybody I talked to as a data

scientist like even people that have

degrees that aren't even related like oh

I've taken a day of the science course

as if like they're gonna like you know

stamp their resume or stamp their

graduate degree like their data

scientist because they took a course

um I mean it's laughable it's like me

taking you know English 401 or something

and then being like I'm an English

expert I'm going to write a bunch of

books and novels and you know stories

and fairy tales and I'm excellent of

course I'm not I'm terrible at these

sorts of things

um but no data science and machine

learning as a whole is one of those

weird Fields now or if you look at a lot

of the job descriptions so I've been in

this industry a bit here in Quant

Finance

um I sit on fintech now it's where I'm

working currently I advise and discuss

and chat with people in D5 fintech

Finance traditional Finance it's Quant

Finance Banks investment firms like I

rub shoulders a lot of different people

even in industries that are analytical

driven that have full staffed data

science teams that are highly rigorous

but the reality is is that now people

have kind of gotten away from data

science because when it started being a

data scientist with somebody who could

actually build models and they took it

seriously as an actual science

unfortunately though what's ended up

happening is that

it's degraded down into like I mentioned

people using Excel and you know coding

things now in Python and R so simply

that they have no idea what's even going

on they can't even address the issues

themselves and so uh all these firms now

all these big Tech firms those other

companies have started now just changing

titles like you're now a machine

learning engineer or you're I don't know

a machine learning scientist or I work

in AI uh again they're technically

different things but the industry is now

coping with the struggle of you've

degraded data science what it was

originally down to like lowly data work

and there's nothing wrong with that but

to do General analytical data

calculations and things like taking

averages

or just haphazardly fitting something

like a model to you know a bunch of dots

on the screen it doesn't really matter

you don't need a rocket scientist or a

Quant to do this or someone super

rigorously educated and trained with a

master's degree from top university or

even a PhD and so you have to kind of

weed through this this is the one piece

of this this is the

you know this is what's happening to

data science as a whole uh Quant Finance

has got already gone through this so

Quant Finance is much more mature uh in

the life cycle of career development

education and things and it is

struggling as well here which I perhaps

make another video on in the future

um but it went through that phase where

it was like being a Quant was like top

paying amazing stellar and then you know

people couldn't pay you enough people

couldn't find you if you had a master's

in Quant Finance uh before like 2007

2008 it was the dream time to be a Quant

so what ended up happening here is that

financial markets blew up 2007 2008 uh

way too many Master's programs have been

created over the years and again the

quality and standard of quants has been

degraded down to very very minimalist

portions but I think more importantly

here we have had many many banks and

firms so quantitative finances much more

structured and defined uh so Banks and

firms that need people to build models

to to do things where you're either

right or wrong and if you're wrong you

lose a lot more money it's easier to see

when you're right and wrong these sort

of firms they've sampled and tried doing

the undergrad path and there are still

some which I won't name here firms and

banks that are doing it and most firms

that have attempted and tried this have

failed absolutely miserably the ones

that are still continuing to do it here

in 2023 they've actually split these

into two different jobs and you have

like a junior quantitative person in a

senior quantitative person and

realistically what's happening is Junior

quantitative people are doing like the

data scientists are doing they're just

doing data analytics and simple

things uh and then what's ended up

happening is they have the actual quants

which now they're bringing in a lot of

them are just phds and they're having

the phds actually do the model fitting

and the theoretical part of actually

putting all the pieces together making

sure it's correct and it's going to work

um so they can say and I've seen many of

them standing on their uh you know Ivory

Towers saying oh we're equal opportunity

employers and we're trying to bring in

undergrads here but Quant Finance is one

of those areas where it's like they just

can't boil it down you have to have so

many skills to do it correctly that

there's no way I don't think to get

under the Masters minimum here now you

could create massive training programs

and I've seen firms do this where

they've brought in undergrads and then

they have all this mandatory Education

and Training and it's like hundreds or

thousands of hours of training to get

these undergrads to that point but again

it's an investment piece here do firms

really want to do this some do some

don't 99.9 of all Quant Firearms don't

want to do that because to hire a bunch

of training and educational people which

is just expensive and often it's not

worth the effort when you can just go

out and find Master's students now data

science as a whole let's just break this

down more specifically it's kind of like

a weird I don't know it's like the

little brother of Quant Finance in many

ways but a lot more General so iview

data science as just like the whole

picture of your model Developers for a

wide area but unfortunately you

realistically only use tools in the

machine learning space uh that's kind of

what's happened here because if you

start backing out the reality of this

when firms hire such as myself in the

quantitative Finance space I view data

science as just a subset in machine

learning and AI as a subset

AI is kind of on the border because you

can do automation with that but the

model development portions of these as a

subset of Statistics so and then data

scientists get up in arms and machine

learning people and they're yelling and

screaming oh you don't know it's

completely different uh you're still

using logistic regression you're still

using OLS as much as you cry and

complain and hate linearity and you know

oh what about these non-linear cases

here you know you can do all that

actually with linear regression again

I'm not going to go into that doing data

Transformations and variable

Transformations going to the models but

what ends up happening is that they're

so specialized into only one area it's

like you kind of have a specialty but

you only can use a couple tools because

if you use all the tools you're not

really a data scientist or machine

learning expert you're just a

statistician or a Quant in the finance

space but more or less I figure like

data science was going to take a more

well-rounded role which they have not

the community itself I think is quite

um quite toxic to be quite honest with

you like people I've met and ran into it

ends up in this weird weird nuanced

space where it's like they don't want to

use any tools except for those in their

space and there's this like I'm on

LinkedIn the last few weeks you're

scrolling and it's like there's so many

just garbage pieces of people

complaining about how horrible

statistics are how horrible econometrics

are uh there's even people in Quant

Finance like this which I don't have

much respect for in this aspect though

they have other great contributions as

well the industry

um but it's like why would you do

anything with half the tools like I

wouldn't go fix my car and say I'm only

going to use this half the toolbox

because the other half isn't good and

this applies you into the stats for them

as well there are many people that are

pro and anti machine learning on the

stat side I think though it is starkly

different that data science machine

learning is very toxic as a culture in a

community it's very anti-stats where on

the stats I think a lot of us are just

more or less like

we're leery like why are you like you're

doing this new approach and often I put

it in air quotes new approach which is

typically a traditional approach been

relabeled and then it just takes us time

to figure out what you're trying to do

uh and then a lot of it's just

nonsensical so the data science approach

I am a hundred percent against for most

problem solving there are cases where

you could use it but the data science

approach being I have data I need

maximum accuracy Do or Die let's get

maximum accuracy and that's what ends up

happening and they even have these so

someone who actually managed teams and

runs people and worries about

profitability and things that have

nothing to do with the model development

process

um on the management side of this right

I don't want to have to have models blow

up because in Quant Finance again here

it's going to act my finance background

the issue with this is I can't afford to

have a model just blow up and just not

have a model

like this might be okay in the investing

side because again in investing you have

so many dollars and if you have a model

blow up and you just want to close the

position so you're like ah the model is

not really working anymore we didn't

really lose too much but it's not

working you can just close that and just

hold on to cash now on the banking in

the sell side of this uh we have to make

loans to people that's how we make money

and that's how we employ thousands and

thousands and thousands of people at

these massive Banks and even fintech uh

D5 and crypto firms as well a lot of

these that are not focused on the buy

side the investing piece of it but are

on the sell side you have to have a

fraud detection model to detect fraud

you have to have all these operational

models to determine optimizations of

different sorts of problems like

portfolio positions for example perhaps

more on the investing side and when

these things blow up you have to have

something there the problem with the

data science approach is that you know

you just

slap something together to hurry to get

to a solution and you don't care if it

blows up or not because you're just

going to redevelop a new model and

there's always this idea that keeps

getting pushed so it's not actually

implemented in practice by many firms

which is that you're going to automate

this whole process completely so now you

take machine learning and Ai and you put

them together and what people oddly

don't understand is you can take

statistics and Ai and automate

statistics as well that's what stepwise

regression or stepwise uh variable

selection is so stepwise forward and

backward selection you could literally

just automate it and have it go out and

magically pick variables throw them in

find the best fit and then just keep

generating model after model after model

and when the models blow up it just

automatically generates new models now

the problem with this is in practice I

mentioned it's just when they blow up

data scientists just go

wasn't me I don't really care it was

just a model and like

I just want to strangle people to death

often because I see this everywhere and

it's not even like it's not even in

firms I'm at it's like you see people on

LinkedIn posting this I look in forums

and communities I talk to friends of

mine that are running large teams I talk

to friends of mine that are on the data

science side and I'm like

I respect you and you're an expert and

there are good data scientists out there

so do not take this as they're all bad I

have friends that work in the data

science Community top-notch

um again master's degrees and they are

excellent and I bring these up I'm like

doesn't this just drive you absolutely

nuts like they fit do you have this

issue and I'll explain like this person

fit a model there's no consideration for

the actual usage of it it was all just

slapped together and yes the fit was

Stellar nobody understood the model

nobody could figure out how Dependable

the model is going to be and nobody

could tell me how robust the model was

going to be and there was almost no

testing because again why the hell would

you test anything when you can just slap

it into python or into R and it will

magically shoot out a model and it just

gives you magical operational you know

execution of it so ml Ops as we like to

call it

um but who cares how it really works

like it just and I asked you don't

doesn't this drive you nuts do you not

look at the mathematical equations do

you not need to understand how these

things are freaking working and my

friends like yeah yeah Dimitri it does

it does drive me nuts but you get over

it and I think that's going to be the

difference between so going back to the

point of this video uh that's going to

be the difference here between I think

machine learning data science and Quant

Finance I think that as we're seeing now

data science is starting to like segment

into

um

not very technical roles like simple

analytic roles where you can do an

undergrad I think the more technical

roles are starting to require Masters

again it's still there so I think in

many ways if you're on the job search

one easy way to sort them is to look at

does it require a master's degree it's

probably pretty rigorous if it does not

require a master's degree it's probably

going to be data analytics because again

they're all labeled data science but

again I think the problem with machine

learning data science as a whole is it

has to merge back in with traditional

statistics at some point and

just the way that it's set up and

operates

um I think you're going to continue to

see this weird segmentation inside of

machine learning and I think

unfortunately it's going to take

probably at least 20 years or more for

the machine Learning Community to

realize like we need to know what we're

doing before just slapping things in or

like trying to hurry and get a solution

and trying to reinvent the wheel every

single day which is just tiring beyond

belief to deal with and so I think that

piece of it once it finally becomes more

mature I think you'll finally get maybe

some new titles where you have uh like

we're having data scientists now it's

kind of viewed as like a not real

skilled position and now we have like ml

Engineers is a more skilled technical

position I think you'll continue to see

that split

um where eventually ml will hopefully

mature enough as a community and as a

field of study that it will come full

circle and actually utilize all the

tools just like stats is trying to do in

many cases and I don't know how we're

going to merge these two things back

together

um because I mean the terminology is

different current and yet it's the exact

same thing in many situations so that's

going to be a little bit of a challenge

here but I think you will start to see

that again those that actually need a

master's degree will continue to need it

and those it can do with an undergrad

continue to do with an undergrad I don't

think Quantum Finance though is going to

deviate down that

I have seen so many banks so it's a side

note wrap up here in a story I have seen

so many banks pushing for this because

they do not want to pay the price tag

that Master and PhD quants cost firms do

not want to pay it they are desperately

looking for solutions to hire undergrads

unfortunately though it just it never

results in good quality models because

there's so much effort work that goes

into this and even as someone who hires

and trains training someone with a

masters in a PhD is still a ton of work

even when you hire the best of the

brightest from the best programs in the

country in the world it is still a ton

of training and a ton of cost and

because of that it's just easier to hire

Master students who already have that

rigor who have that you know that drive

to actually get that graduate degree and

have all that additional education that

a university actually did for them and

they paid you know 70 to 100 000 for so

anyways thanks for listening thanks for

watching and as always until next time

[Music]

thank you