SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
ikicker
Topic Author
Posts: 10
Joined: June 8th, 2011, 12:19 am

Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 15th, 2019, 8:09 pm

I am curious as to what you think the applicability of multi-state-models (MSM's) are to venture capital finance, as opposed to utilizing less technical methods or real options valuation and perhaps we can come up with some examples of next generation phasing within venture capital. I'm curious as to what those critical points are, the probabilities of success, etc.

In his most recent interview, Elon Musk described his thinking of innovation revolving around Neuralink: "So you're saying there's a chance..." - Elon Musk describing the motto for any engineering team that takes on difficult challenges, in this case Neuralink aligning BCI & AGI. 1-in-1000000 becomes 1-in-1000 becomes 1-in-10 becomes a reality. Full conversation: https://www.youtube.com/watch?v=smK9dgdTl40

Essentially, Elon Musk thinks about innovation as a MSM. MSM's are commonly utilized in the medical profession and the hard sciences. This is a very, very unique idea. Here is a recent paper summarizing MSM's:

Inference for outcome probabilities in multi-state models (Per Kragh Andersen and Maja Pohar Perme) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2735091/
----
Undergraduate: accounting, finance, information systems; Graduate: MBA/finance; Graduate certificates: data science, applied statistics, advanced valuation; PhD: data science
 
User avatar
Paul
Posts: 9769
Joined: July 20th, 2001, 3:28 pm

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 15th, 2019, 8:59 pm

Would you say that MSM are between "less technical methods" and "real options"?

Although I've written about real options I've never used them in practice. But when it comes to any new business I tend to start off by saying it has a 5% chance of success, and then this could happen or this or this. So I see different states, but I don't usually quantify them. But I do like things that have several different routes to a payoff, or even better would be routes to several payoffs/spin offs.
 
User avatar
ikicker
Topic Author
Posts: 10
Joined: June 8th, 2011, 12:19 am

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 16th, 2019, 10:34 pm

Would you say that MSM are between "less technical methods" and "real options"?

Although I've written about real options I've never used them in practice. But when it comes to any new business I tend to start off by saying it has a 5% chance of success, and then this could happen or this or this. So I see different states, but I don't usually quantify them. But I do like things that have several different routes to a payoff, or even better would be routes to several payoffs/spin offs.
I think this is a more concise explanation: https://www.maths.lancs.ac.uk/~titman/leeds_seminar.pdf

Now that I consider it, I think MSM's give you a better figure to put into a real options model. The issue with computing real options is the standard deviation usually. There is a way to compute standard deviation at some future date based off of a payoff table. The MSM gives you a more accurate perception of those probabilities. If you put a chart together of potential outcomes in Visio, you could break down something actually very complex into a relatively simple problem.

When people don't understand finance that well or come to understand it on their own, I've found that they come up interesting ways of modeling problems mathematically. In this instance, I think that is why Elon Musk came up with a MSM model. I ran into another really successful guy who was doing something with capital structure and valuation multiples. The way that Elon Musk describes it, he thinks about the relationship between the multiple states as being a function of your rate of breakthroughs in success. There is also the popular terminology surrounding burn rates, etc. He makes no mention of this, but it is sort of inherently implied. For example, you receive some capital, and there is some future date at which you must complete your innovation otherwise you will have "burned through" all of your capital, at which point you are ready to go to market. What is the probability of hitting that target spot on or exceeding it at any point in time? In theory there is a distribution of potential outcomes revolving around that target date.

I've had to use real options before when computing parts of contracts, granted I had my own company at that point so I was able to do everything at the level I would like to for analysis. There was one acquirer who frequently liked to have put options on the sale inserted in the contracts (the right to sell remaining shares, up until a future date, at a specified multiple of EBITDA). I used a forecast of scenario's to determine the payoff table, and the payoff table to determine the standard deviation (if you have any payoff table of any kind, you can compute expected variance in addition to expected value).
  • There is also a manner in which you can compute standard deviation using a binomial method as well at any point in time. You can either do it the same way as with a multivariate scenario outcome, or slightly differently, but in theory you should come up with similar values.
There is also likely some sort of use when it comes to default probabilities, utilizing a technique like the Merton model for firm credit risk: https://courses.edx.org/c4x/DelftX/TW3421x/asset/Week5_PD2_2.pdf

For example, when examining the credit default spread with fixed income securities (a Venture-Debt scenario):
  • Equity risk factor - the credit (or default) spread can also be thought of as a put option. There is an inherent put option on any bond for default because they can default and give up their assets.
  • Bankruptcy is just the exercise of this put option
  • It’s value is = to the Max (Liabilities – Assets,0)
  • This gives you your option adjusted spread
----
Undergraduate: accounting, finance, information systems; Graduate: MBA/finance; Graduate certificates: data science, applied statistics, advanced valuation; PhD: data science
 
User avatar
katastrofa
Posts: 8527
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 16th, 2019, 11:11 pm

Multistate Markov models are indeed powerful in domains like medicine, because they make use of longitudinal data typically gathered in clinical trials (with all their features, such as attrition, censoring, etc.). I wouldn't say there is anything "very, very unique" about them. They can be thought of as an extension of hidden Markov models to a situation in which transitions happening at arbitrary times. I and my colleagues in social sciences used to design and implement such models on daily basis.
If you want a true advantage, you hardcode the problem in a quantum annealer (D-Wave or Sycamore) ;-)
 
User avatar
Cuchulainn
Posts: 60518
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 11:31 am

Multistate Markov models are indeed powerful in domains like medicine, because they make use of longitudinal data typically gathered in clinical trials (with all their features, such as attrition, censoring, etc.). I wouldn't say there is anything "very, very unique" about them. They can be thought of as an extension of hidden Markov models to a situation in which transitions happening at arbitrary times. I and my colleagues in social sciences used to design and implement such models on daily basis.
If you want a true advantage, you hardcode the problem in a quantum annealer (D-Wave or Sycamore) ;-)
Since when have social scientists a $10,000,000 budget for a D-wave computer??
Clients are Google. NASA, Lockheed...
Is there a Community version  :mrgreen:
http://www.datasimfinancial.com
http://www.datasim.nl

Approach your problem from the right end and begin with the answers. Then one day, perhaps you will find the final question..
R. van Gulik
 
User avatar
Cuchulainn
Posts: 60518
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 12:57 pm

I am curious as to what you think the applicability of multi-state-models (MSM's) are to venture capital finance, as opposed to utilizing less technical methods or real options valuation and perhaps we can come up with some examples of next generation phasing within venture capital. I'm curious as to what those critical points are, the probabilities of success, etc.

In his most recent interview, Elon Musk described his thinking of innovation revolving around Neuralink: "So you're saying there's a chance..." - Elon Musk describing the motto for any engineering team that takes on difficult challenges, in this case Neuralink aligning BCI & AGI. 1-in-1000000 becomes 1-in-1000 becomes 1-in-10 becomes a reality. Full conversation: https://www.youtube.com/watch?v=smK9dgdTl40

Essentially, Elon Musk thinks about innovation as a MSM. MSM's are commonly utilized in the medical profession and the hard sciences. This is a very, very unique idea. Here is a recent paper summarizing MSM's:

Inference for outcome probabilities in multi-state models (Per Kragh Andersen and Maja Pohar Perme) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2735091/
As kats states, MSM looks awful like Markov Model (new wine in old wineskins).
Why all these new fancy names 
Last edited by Cuchulainn on November 17th, 2019, 1:12 pm, edited 1 time in total.
http://www.datasimfinancial.com
http://www.datasim.nl

Approach your problem from the right end and begin with the answers. Then one day, perhaps you will find the final question..
R. van Gulik
 
User avatar
trackstar
Posts: 26820
Joined: August 28th, 2008, 1:53 pm

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 1:09 pm


Since when have social scientists a $10,000,000 budget for a D-wave computer??
Clients are Google. NASA, Lockheed...
Is there a Community version  :mrgreen:
Yes, actually - they are trying to stimulate interest and increase adoption in the community and so are offering free access and training.

See, for example D-Wave's LEAP program here: Take the Leap - D-Wave Systems

From the page:
  • Free access: free, immediate access to a D-Wave 2000Q quantum computer
  • Familiar software: Ocean software development kit (SDK) with suite of open source Python tools and templates
  • Hands-on coding: interactive examples and Jupyter notebooks with live code, equations, visualizations, and narrative text
  • Learning resources: comprehensive live demos and educational resources
  • Community support: community and technical forums for developer collaboration
I have been to a number of conferences where QC was on the agenda and this has been mentioned in the talk.
 
User avatar
trackstar
Posts: 26820
Joined: August 28th, 2008, 1:53 pm

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 1:23 pm

To give IBM and Sycamore (Google) equal air time:

IBM Q Experience

IBM Q Network

"Easily program with Qiskit software in the cloud

Program with Qiskit notebooks powered by Jupyter technology integrated into our platform. Qiskit is a full stack quantum software framework that provides all the quantum development tools you need."

They have communities on Slack, Stack Exchange, and Git Hub, among others.

On Google's side, see this, for example:

First Look at 'SycamoreGoogle's Quantum Computer - Gizmodo October 23, 2019

Hello quantum world! Google publishes landmark quantum supremacy claim - Nature News October 23, 2019
For Google's Sycamore paper see, Nature 574, 461-462 (2019)

The competition between the two firms has appeared in a number of stories, including this:

IBM and Google clash over milestone quantum computing experiment - Quanta Magazine October 23, 2019
Last edited by trackstar on November 17th, 2019, 5:05 pm, edited 4 times in total.
 
User avatar
Cuchulainn
Posts: 60518
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 1:53 pm

wow, thankx

fixed..
Last edited by Cuchulainn on November 17th, 2019, 3:06 pm, edited 1 time in total.
http://www.datasimfinancial.com
http://www.datasim.nl

Approach your problem from the right end and begin with the answers. Then one day, perhaps you will find the final question..
R. van Gulik
 
User avatar
trackstar
Posts: 26820
Joined: August 28th, 2008, 1:53 pm

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 2:04 pm

Sure - one further resource for HPC more generally - the big annual Supercomputing conference starts tomorrow and runs until the 22nd.

Great website - lots of information, even for those who cannot fly (or use quantum teleportation!) to arrive in Denver today. : )

SC19

Attendance is in the 1,000s - lots of exhibitors, many tracks, and workshops. Sponsored by ACM and IEEE.

**

Ah - on T&C - maybe it sees that you are in EU - perhaps call customer service tomorrow and see what they can do to hook you up?
 
User avatar
katastrofa
Posts: 8527
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 2:48 pm

Is the US or Australia health care country budget enough, Cuchulainn?
 
User avatar
Cuchulainn
Posts: 60518
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 3:05 pm

Is the US or Australia health care country budget enough, Cuchulainn?
Not if Trump can help it; in fairness, you mentioned social sciences, so I had to start somewhere.
I would prefer more beds in hospitals.
http://www.datasimfinancial.com
http://www.datasim.nl

Approach your problem from the right end and begin with the answers. Then one day, perhaps you will find the final question..
R. van Gulik
 
User avatar
Cuchulainn
Posts: 60518
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 3:16 pm

Ah - on T&C - maybe it sees that you are in EU - perhaps call customer service tomorrow and see what they can do to hook you up?

fixed. I tried another PC. 

They call me the wanderer

Yeah, the wanderer
I roam around, around, around
http://www.datasimfinancial.com
http://www.datasim.nl

Approach your problem from the right end and begin with the answers. Then one day, perhaps you will find the final question..
R. van Gulik
 
User avatar
trackstar
Posts: 26820
Joined: August 28th, 2008, 1:53 pm

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 3:45 pm

Ah - on T&C - maybe it sees that you are in EU - perhaps call customer service tomorrow and see what they can do to hook you up?

fixed. I tried another PC. 

They call me the wanderer

Yeah, the wanderer
I roam around, around, around
Excellent - and now you have an interesting new sandbox to wander around in!
 
User avatar
Cuchulainn
Posts: 60518
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Multi-state probabilities/models (MSM) and regression applied to venture capital and innovation

November 17th, 2019, 3:55 pm

It's moving, but only 1 minute QPU time/month.
I'll need a bigger boat to do a social network analysis of this site, down the years.
dwave-ocean-sdk OK 
Remark: It would be nice to run on PC before burning my monthly qbit allowance!
Last edited by Cuchulainn on November 17th, 2019, 4:05 pm, edited 1 time in total.
http://www.datasimfinancial.com
http://www.datasim.nl

Approach your problem from the right end and begin with the answers. Then one day, perhaps you will find the final question..
R. van Gulik
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On