Day 5 #RWRI 14
1/n
Started with a Q&A, and then moved to optionality, convexity and model error. @DrCirillo talked about EVT and @financequant was able to join us again.
Personal highlights
- Optionality: You have the right but not the obligation. Notice the asymmetry here.
2/n #RWRI 14
- Optionality gives you convex payoffs (very rarely optionality <=> convexity doesn't hold)
- Example of optionality: trial and error (T&E)
- How to find hidden optionality?
A. Not easy. You need to train for it. Domain dependent, but T&E gives you optionality
3/n #RWRI 14
- Optionality implies an asymmetry. Can you have optionality without a sucker on one side?
A. Yes. Trial and error
-Low diversification makes you fragile. If you don't have a tail hedge, you'll eventually blow up.
4/n #RWRI 14
- Green lumber fallacy: What you need to know isn't necessary what you're supposed to know. See
en.wikipedia.org/wiki/Antifragi…
- Model error: 3 classes
1. Linear (practically imposible to miss)
2. Missing significant variable
3. Missing second order effects
5/n #RWRI 14
- Model error: How to stress test your model?
Use the (1+a) rule. For x, compute x(1+a) and x(1-a) and see what happens.
- Beware of BS predictions like "x has probability 0". It can't be 0 because it misses error on the left
E.g @NateSilver538's 2016 predictions
6/n #RWRI 14
- Raphael after this:
- Extremes are not the same as outliers
Extremes: An event within the range of variation. Basically, unusual max/min (eg someone who is 2.1m tall)
Outlier: An event out of the natural range of variation (eg. 3m tall)
7/n #RWRI 14
- How to correct outliers?
A. Directly (just change the value), with the avg of nearby values, or by removing them.
- How to correct extremes?
A. Don't. You'll miss important information about the distribution. Don't even think about it.
8/n #RWRI 14
- Pasquale's fat tail's taxonomy:
- On predictions: Is not about *when* something is going to happen, but whether or not that something is possible. Remember what we have been saying since day 1. It's about the impact.
9/n #RWRI 14
Some papers:
arxiv.org/pdf/1505.04722…
https://t.co/1fAPabIMOb
Nassim joined the optional third session of the day by phone because he is driving to NC. Nothing stops #RWRI
10/n #RWRI 14
Forgot to add
- When to use EVT (Extreme Value Theory)?
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.