Thread by @mjmauboussin

A couple of weeks ago I mentioned Jeff Friedman's book, War and Chance.
It's about uncertainty in international politics, but has a lot of application to thinking about businesses and investing. Some thoughts:
amazon.com/War-Chance-Ass…

2/Two quotes from early on:
“…analysts avoid assessing uncertainty in a manner that supports sound decision making" and “analysts cannot speak about the future without engaging in some form of probabilistic reasoning.”

3/Common mistakes:
1.Use words of estimative probability instead of probabilities themselves;
2.Discuss relative probability – probability assessment that conveys an analyst’s beliefs with respect to an unspecified baseline;

4/ Common mistakes (con't):
3.Conditioning – occurs when analysts describe the conditions that must hold in order for a statement to be true.

5/Epistemic uncertainty = outcomes are unknown because analyst possesses incomplete information. This type of uncertainty is always subjective.

“To say that a judgment is subjective does not mean that it is careless or uninformative”

6/Probability – reflects analysts’ beliefs about the chances that a statement is true;
Confidence – reflects the extent to which analysts belief they possess a sound basis for assessing uncertainty

7/ Basis for confidence:
1. Reliability of available evidence;
2. Range of reasonable opinion surrounding a judgment;
3. Degree to which analysts believe their judgment could change in response to new information

8/How to find your indifference point:
1. You will $1,000 if a Republican wins next election;
2. You win $1,000 if coin toss comes up heads.
If you prefer bet on Repub, outcome is at least 50%
If you bet on coin, outcome of Repub is no higher than 50%
Toggle until indifferent

9/If you believe range is 40-60% with equal probability, take the average = 50%
If equally credible analyst are between 40-60%, take the average
If analysts are not equally credible, weight their views
Translate views into bets --> eee @AnnieDuke

10/“There is no instance in which foreign policy analysts can coherently offer a qualitative probability estimate but not a qualitative probability estimate.”

11/“…the value of precision in probability assessment reflects a generalizable skill that foreign policy analysts can cultivate through training, effort, and experience.”
See @PTetlock and @dgardner and @superforecaster

12/"…if foreign policy analysts are truly worried about making errors or having their judgments distorted, their best move might be to embrace probabilistic reasoning, rather than avoid it.”

13/ “In the long run, analysts who are evaluated using the Brier score will achieve their best results by reporting their true beliefs.”

14/A decision is worthwhile if pB > C, where p = probability of success, B = benefit, and C = cost. Or worthwhile if p > C/B
You can invert; don’t know p, figure out what C/B would makes sense.

15/Basic Standards for Assessing Uncertainty
1.“describe the uncertainty surrounding any prediction or policy recommendation;”
2.Never use relative probability or conditioning;

16/Basic Standards (con't)
3.“…assessments of uncertainty should be clear enough for readers to reliably understand what they mean. Supplement key judgments with numeric percentages;”
4.“consistently distinguish between assessments of probability and confidence.”
FIN