In the last two weeks, I've gotten really into quantitative analysis and wow, your content just keeps pushing my interest up. I'm hoping this isn't just a phase, I'm actually enjoying learning about all of this! Thank you for this great content, Dimitri! :D
@@DimitriBianco That's neat. I hope to get to the point where I can understand what forward propagation means LOL. Besides the informational videos, like this one. I enjoyed your reflection/journey on becoming a quant video. Mainly since I'm not coming from a financial background, it's good to know that someone made it to be a quant without the path being laid fully out to them. I'm a CS major at a uni in NY funny enough and trying to self-teach myself to become a Quant hopefully so that video just resonated with me more and gave me some hope that I can maybe do it.
I remember from one of the wizards in Market Wizards who said that the Black-Scholes model is not only bad for real life use but a lot of the users tweak it just to make their funds and institutions look more advanced
You should make a video on how to price an option by by relaxing one of the main BSM assumptions and show how to use numerical methods to price it in R or Python.
Hey Dimitri, I really enjoyed this video! I'm a CS student interested in quant finance and have been learning on the side for about a week now. Can you upload more of these informational videos. Much appreciated.
Basicall you have to be good enough in math so that you can derive your own model based on your data and assumption you use. You can use BSM model for your starting point. And then you can make your own and test it out.
Black scholes is a easy to understand, easy to use model that can be higly usefull for valuing short term options, say having life of2-3 months, as the assumptions of constant volatility and constant risk free rate can hold good for short durations. This is more so for traders ,who donot have access to more accurate volatility prediction models using time series or other techniques.
What nobody is telling retail investors: Nobody has ever modeled long indexes. Big wealth manager analysists and associates only do long analysis with industry trends and never use index data; the best mostly do dividend returns. There is no math that is going to model predictive analysis of long term index behavior using only index back-data because driven by pure demand and speculation...
Hey Dimitri, great explanation there! I wonder if financial engineers utilize time series models in practice, like ARIMA, ARCH, GARCH etc. to model stock returns and volatility. Thanks!
Would you agree with the caveat that they're often being used by different people rather than a person directly using both consistently? ( e.g. Buy side/Sell side ) Have you come across many roles that routinely deal with Q and P space?
I covered most of this in my undergrad... does it get much more difficult? We covered the formal proof for the black Scholes way; and how to get to the same conclusion through normal assumptions, equivalent portfolio, binomial trees etc. And using alternative portfolio theory assumptions to use binomial trees for estimates of exotic options. I'm considering going back to do a masters, and wondering how challenging it becomes. And how applicable, as I'm starting to hold options in my portfolio but mostly focus on what the greeks are telling me and comp vol to my assumptions in the current market. Aside from learning to code R does the theory get much much deeper? I did an honors BMath in Act Sci with a finance specialization, so more math focused. But we mostly covered the proofs and would get asked to apply with a small change in assumptions (eg. An exotic where interest rate changes being asked to find the modeled change in price ). Any recommendations on good programs would be appreciated as well.
Yes, it gets much deeper. The Black Scholes is an amazing equation due to the closed form solution however it has too many assumptions which results in inaccurate pricing. Good masters programs will focus numerical methods and program optimization to find better solutions however they are not closed form. This results in every firm having their own proprietary formula for a trade-off between run time and accuracy. I like Jim Gatheral at Baruch. He's a sharp guy and his research with students is very valuable. NYU's program Peter Carr is also note worthy due to the number of practitioners teaching there.
But Heston model is not good to catch a good volatility smile. But it can be fixed by changing initial condition of stochastic volatility (which is not random - it given by simplicity) to some random variable with gamma distribution (as example)
Is the main idea that all the distributions converge almost surely to the model? For example the probability of arbitrage is centered at 0, but not necessarily equal to 0
Distributions play a role in stochastic calculus and there are assumptions about normal distributions with a mean of zero (random walk) but the real critical point is risk neutral probabilities.
Hey dimitri awesome and very direct video as per usual. Question: I'm currently an undergraduate studying towards Software Engineering(Similar to CS however where we do more System Architecture and application they do some discrete maths other than that the course is exactly the same. ) Im currently in my 2/4th year at uni about to go into 3rd. I plan on doing a MFE at a US university as that's where most of the jobs are & best programs are. My question is, in your day to day job when you are implementing code, based on models, in your opinion what kind of understanding would you need to be a tremendous Quant developer? Any help is appreciated
Typically the guys implementing models are computer science majors not quants. Implementation needs to be very good at programming so they can optimize run-time. A lot of these CS guys are making great money but they just do different things. Quants focus on math and stats for model and theory development.
@@DimitriBianco Hey Dimitri, Thanks for the response very interesting, sorry just a follow up question. With a strong emphasis on CS and programming, what kind of Quants exist where I could utilise my Software Engineering degree, any information is seriously greatly appreciated. Wish you had a website of your experiences and opinions, would be seriously awesome.
@@TheXenaFilms I would go for a job focused around algo trading (either model developer or implementation). They use a lot more intense programming than other types of firms because they have to have low latency. Watch the video I linked below if you haven't seen it. The part of the video with Haim Bodek will give you more insight around quants who code a lot. ua-cam.com/video/kFQJNeQDDHA/v-deo.html
Why is every one so obsessed with degrees in this field? Probably most good mathematicians / computer scientists / engineers could grasp something like black-scholes in a day... Its rather interesting how inefficiencies in the market can be uncovered by quantitative methods in reality. But there seem to be only a few players applying this successfully and they obviously have no incentive to give away their secret sauce. But I wonder how many private players there are applying e.g. supervised or reinforcement learning to make a buck. In your video you say its mostly regulations that prevents machine learning being leveraged, but this would not apply to an individual, so in this case the individual could very well be in advantage to the big players. Have you ever tried to apply your quantitative knowledge to outperform the market with your personal investments?
To clarify, the regulations I mention apply to banks. Hedge funds are not considered banks and are using machine learning and AI. There is a lot of hype around this topic however from what I have seen most hedge funds are just playing with these ideas. By that I mean investing small amounts and testing. Most of the quant funds that have been around a long time are still relying on traditional statistics and mathematics. I have used a few basic quant concepts in personal investing however to really do things right you need a lot of time and data which is hard to come by when you have another full-time job. Now I just invest in long-term strategies and take advantage of market shocks from bad news. Also as a side note, big banks are using machine learning however it is supervised. In my other video I mention there are issues with regulation however regulators are starting to get used to supervised versions of machine learning. Most banks have been using it for fraud detection for many years but they are hesitant to use it for regulatory models.
@@DimitriBianco Thanks for your great answer, I really appreciate it and for me it just today became clear the huge difference between the fields referred to as "quantitative". I always had the point of view that everything in finance is some deviation of the situation: "Here you have some money, make the best out of it". Nowadays this seems to mainly apply to (hedge?) funds. Although I think also the normal business of a bank in a free market should boil down to this, but as we have tons of regulation (and tradition?) there are now certain ways how this needs to be pulled off (models being explainable seems to be a huge restriction of the freedom). Because of this it needs experts to handle these (artificial) discontinuities. But disregarding this, wouldn't the return one (or one system) delivers be the best (unbiased?) estimator (versus some regulation induced heuristic)? This would automatically be risk-adjusted by a survival of the fittest style selection in the longterm, because of inefficient actors losing money/going bankrupt. From the videos I have seen you do not seem too opposed towards regulations, I really wonder why you take this position? In economics you have this beautiful (probably unbiased? though very noisy) signal of the monetary return why can't the banking/regulating world rely more on this? Don't get me wrong, I think you in the US are far ahead in relying on this signalling in comparison to us in the EU. We are rather going backwards step by step right now...
In the last two weeks, I've gotten really into quantitative analysis and wow, your content just keeps pushing my interest up. I'm hoping this isn't just a phase, I'm actually enjoying learning about all of this! Thank you for this great content, Dimitri! :D
Next week is a math example of forward propagation for neural networks. I hope you enjoy it. What has been your favorite video so far?
@@DimitriBianco That's neat. I hope to get to the point where I can understand what forward propagation means LOL. Besides the informational videos, like this one. I enjoyed your reflection/journey on becoming a quant video. Mainly since I'm not coming from a financial background, it's good to know that someone made it to be a quant without the path being laid fully out to them. I'm a CS major at a uni in NY funny enough and trying to self-teach myself to become a Quant hopefully so that video just resonated with me more and gave me some hope that I can maybe do it.
I remember from one of the wizards in Market Wizards who said that the Black-Scholes model is not only bad for real life use but a lot of the users tweak it just to make their funds and institutions look more advanced
You should make a video on how to price an option by by relaxing one of the main BSM assumptions and show how to use numerical methods to price it in R or Python.
Hey Dimitri, I really enjoyed this video! I'm a CS student interested in quant finance and have been learning on the side for about a week now. Can you upload more of these informational videos. Much appreciated.
update ?
Buffet essentially said the same thing as it pertains to short vs long timeframes, in that its only helpfull for very short timeframes.
Basicall you have to be good enough in math so that you can derive your own model based on your data and assumption you use.
You can use BSM model for your starting point. And then you can make your own and test it out.
You are right, Derman mentions The Fruit Salad analogy in his book My Life as a Quant. A really fun read btw.
He's working on a memoir right now.
Black scholes is a easy to understand, easy to use model that can be higly usefull for valuing short term options, say having life of2-3 months, as the assumptions of constant volatility and constant risk free rate can hold good for short durations. This is more so for traders ,who donot have access to more accurate volatility prediction models using time series or other techniques.
What nobody is telling retail investors: Nobody has ever modeled long indexes. Big wealth manager analysists and associates only do long analysis with industry trends and never use index data; the best mostly do dividend returns. There is no math that is going to model predictive analysis of long term index behavior using only index back-data because driven by pure demand and speculation...
I agree with this. I often see people convinced they can predict the future with the past.
Hello, could you elaborate on what this means?
Hey Dimitri, great explanation there! I wonder if financial engineers utilize time series models in practice, like ARIMA, ARCH, GARCH etc. to model stock returns and volatility. Thanks!
Yes, time series is used a lot in quantitative finance for stocks, volatility, credit risk, market risk, and a lot of other areas.
Would you agree with the caveat that they're often being used by different people rather than a person directly using both consistently? ( e.g. Buy side/Sell side ) Have you come across many roles that routinely deal with Q and P space?
You are correct, the roles are separated. There would be a conflict of interest if not.
Can you provide the name of those textbooks? They looked like Springer texts.
They are discrete time and continuous time by Steven Shreve.
How does the time equation effect the market? Mean time-solar time
I covered most of this in my undergrad... does it get much more difficult? We covered the formal proof for the black Scholes way; and how to get to the same conclusion through normal assumptions, equivalent portfolio, binomial trees etc. And using alternative portfolio theory assumptions to use binomial trees for estimates of exotic options.
I'm considering going back to do a masters, and wondering how challenging it becomes. And how applicable, as I'm starting to hold options in my portfolio but mostly focus on what the greeks are telling me and comp vol to my assumptions in the current market. Aside from learning to code R does the theory get much much deeper?
I did an honors BMath in Act Sci with a finance specialization, so more math focused. But we mostly covered the proofs and would get asked to apply with a small change in assumptions (eg. An exotic where interest rate changes being asked to find the modeled change in price ).
Any recommendations on good programs would be appreciated as well.
Yes, it gets much deeper. The Black Scholes is an amazing equation due to the closed form solution however it has too many assumptions which results in inaccurate pricing. Good masters programs will focus numerical methods and program optimization to find better solutions however they are not closed form. This results in every firm having their own proprietary formula for a trade-off between run time and accuracy. I like Jim Gatheral at Baruch. He's a sharp guy and his research with students is very valuable. NYU's program Peter Carr is also note worthy due to the number of practitioners teaching there.
@@DimitriBianco I think I saw a lecture online by Peter Carr on some more recent work by Ross. Seemed to be working on some interesting ideas.
@@DimitriBianco RIP Peter Carr, gone too soon.
What do you think about pricing of Bermudan options as they are a mixture of American and European. Which is better. Binomial or black scholes
Thank you Dimitri, I love this video!
Heston stochastic volatility model has also a closed form solution for EU Call options, using characteristic functions!
Yes, that is another closed form example. It usually slips my mind.
But Heston model is not good to catch a good volatility smile. But it can be fixed by changing initial condition of stochastic volatility (which is not random - it given by simplicity) to some random variable with gamma distribution (as example)
Is the main idea that all the distributions converge almost surely to the model? For example the probability of arbitrage is centered at 0, but not necessarily equal to 0
Distributions play a role in stochastic calculus and there are assumptions about normal distributions with a mean of zero (random walk) but the real critical point is risk neutral probabilities.
You should definetly create a discord server, cool vids
@@გრუტი we have one here:
discord.gg/X5dsW25GcX
Wow amazing! Very informative
Hey dimitri awesome and very direct video as per usual. Question: I'm currently an undergraduate studying towards Software Engineering(Similar to CS however where we do more System Architecture and application they do some discrete maths other than that the course is exactly the same. ) Im currently in my 2/4th year at uni about to go into 3rd. I plan on doing a MFE at a US university as that's where most of the jobs are & best programs are. My question is, in your day to day job when you are implementing code, based on models, in your opinion what kind of understanding would you need to be a tremendous Quant developer? Any help is appreciated
Typically the guys implementing models are computer science majors not quants. Implementation needs to be very good at programming so they can optimize run-time. A lot of these CS guys are making great money but they just do different things. Quants focus on math and stats for model and theory development.
@@DimitriBianco Hey Dimitri, Thanks for the response very interesting, sorry just a follow up question. With a strong emphasis on CS and programming, what kind of Quants exist where I could utilise my Software Engineering degree, any information is seriously greatly appreciated. Wish you had a website of your experiences and opinions, would be seriously awesome.
@@TheXenaFilms You might be looking for a quant developer role, but you have to be damn good.
@@TheXenaFilms I would go for a job focused around algo trading (either model developer or implementation). They use a lot more intense programming than other types of firms because they have to have low latency. Watch the video I linked below if you haven't seen it. The part of the video with Haim Bodek will give you more insight around quants who code a lot.
ua-cam.com/video/kFQJNeQDDHA/v-deo.html
Very insightful and interesting video, I really appreciate your content Dimitri!
Thanks!
Finance academic ...only teaches , on buying a stock on monday morning & sell on friday afternoon. Its for Finance Casino.
Of course, for calibration
interesting video.
Why is every one so obsessed with degrees in this field? Probably most good mathematicians / computer scientists / engineers could grasp something like black-scholes in a day... Its rather interesting how inefficiencies in the market can be uncovered by quantitative methods in reality. But there seem to be only a few players applying this successfully and they obviously have no incentive to give away their secret sauce. But I wonder how many private players there are applying e.g. supervised or reinforcement learning to make a buck. In your video you say its mostly regulations that prevents machine learning being leveraged, but this would not apply to an individual, so in this case the individual could very well be in advantage to the big players. Have you ever tried to apply your quantitative knowledge to outperform the market with your personal investments?
To clarify, the regulations I mention apply to banks. Hedge funds are not considered banks and are using machine learning and AI. There is a lot of hype around this topic however from what I have seen most hedge funds are just playing with these ideas. By that I mean investing small amounts and testing. Most of the quant funds that have been around a long time are still relying on traditional statistics and mathematics. I have used a few basic quant concepts in personal investing however to really do things right you need a lot of time and data which is hard to come by when you have another full-time job. Now I just invest in long-term strategies and take advantage of market shocks from bad news.
Also as a side note, big banks are using machine learning however it is supervised. In my other video I mention there are issues with regulation however regulators are starting to get used to supervised versions of machine learning. Most banks have been using it for fraud detection for many years but they are hesitant to use it for regulatory models.
@@DimitriBianco Thanks for your great answer, I really appreciate it and for me it just today became clear the huge difference between the fields referred to as "quantitative". I always had the point of view that everything in finance is some deviation of the situation: "Here you have some money, make the best out of it". Nowadays this seems to mainly apply to (hedge?) funds. Although I think also the normal business of a bank in a free market should boil down to this, but as we have tons of regulation (and tradition?) there are now certain ways how this needs to be pulled off (models being explainable seems to be a huge restriction of the freedom). Because of this it needs experts to handle these (artificial) discontinuities.
But disregarding this, wouldn't the return one (or one system) delivers be the best (unbiased?) estimator (versus some regulation induced heuristic)? This would automatically be risk-adjusted by a survival of the fittest style selection in the longterm, because of inefficient actors losing money/going bankrupt. From the videos I have seen you do not seem too opposed towards regulations, I really wonder why you take this position?
In economics you have this beautiful (probably unbiased? though very noisy) signal of the monetary return why can't the banking/regulating world rely more on this? Don't get me wrong, I think you in the US are far ahead in relying on this signalling in comparison to us in the EU. We are rather going backwards step by step right now...