notes-summary

at some point i'd like to summarize some advice that i'd give myself. the following is NOT yet that, it's just notes towards that.

note: i tend to think and write in a left-brained manner; i categorize things precisely and define in a more rigorous than typical (although certainly not rigorous in the absolute sense) manner. There is a danger to this sort of thinking; once you assign precise categorizations and definitions to concepts, you find it hard to 'unsee' these assignments and to recall your earlier, more fluid conceptions. The human mind and language's innate capacity for fluid thinking is extremely powerful and its power is dulled by this. So beware.

One partial (less than 40% effective, i would guess) defence against this: as you read my conceptions of words, pause whenever you have a tension in the back of your mind that I'm oversimplifying or leaving out another sense of a word, and think about that other meaning for a little while.

hierarchy of evaluation:

• benefit
• cost and benefit
• profit: benefit - cost
• ROI: (benefit - cost)/investment
• expected value of ROI ratio: mean(ROI + 1) over the hypothesized probability distribution of outcomes
• Sharpe ratio of returns: mean(ROI + 1)/std(ROI + 1) over the hypothesized probability distribution of outcomes
• geometric expected ROI ratio: exp(mean(log(ROI + 1))) over the hypothesized probability distribution of outcomes, which i think is equal to mean(ROI + 1)/var(ROI + 1) if you assume log-normal returns (in other words, the standard deviation divisive penalty is applied twice instead of once when you are dealing with an investment approaching the size of your entire portfolio) (see http://www.decal.org/file/2945 section 6.4 "Continuous/Gaussian analytic Kelly" http://www.decal.org/file/1071 )

In words: The simplest thing to do is to only consider benefits, leading to doing things that are very rewarding but have even greater costs. Then you consider costs and benefits, and stay away from very costly things. Then you quantify these and subtract cost from benefit, allowing you to choose to do even costly things if the benefit is even greater. Then you consider also reusable but limited resources such as time and capital which must be allocated among various profitable opportunities, dividing by these to get ROI. Then you consider that there are different outcome scenarios with different benefits and costs and calculate the expected value over these. Then you consider that, holding expected return equal, investments with more uncertainty are worse than ones with less uncertainty, because a portfolio with higher variance will have slightly less compounding returns (because on some steps, due to unlucky losses, the amount available to be invested on the next step will be lower). Taking this effect into account, assuming a large portfolio and Gaussian log-returns (i think?), you derive the Sharpe ratio.

This still allows you to make extremely high risk, high reward choices such as playing Russian roulette with other players much richer than yourself, choices which, if repeated arbitrarily often, would tend towards a 100% probability of ruin. Taking the geometric mean instead of the arithmetic mean should be applied to one's total assets rather than to each investment in isolation, and it results in seeking survival instead of profit maximization, as the loss of 100% of your capital is penalized with negative infinity (and for this reason, it can't be used exactly in real-life situations, in which there is always some infintesimal chance of 100% loss). If one is considering each of a portfolio of very many, uncorrelated assets in isolation, one should instead take the Sharpe ratio of each asset, because in this case a large loss in all assets at once is unlikely; more technically, one should use portfolio analysis (i think this requires nonlinear programming) to compose a portfolio with maximal geometric mean. Note that in reality one's estimates of risks and returns may themselves be off; to compensate for these, people often calculate a geometric-mean-maximized portfolio and then reduce the leverage by a constant factor, a strategy known as 'fractional Kelly' (see also http://www.edwardothorp.com/sitebuildercontent/sitebuilderfiles/Good_Bad_Paper.pdf ). A commonly used fraction is 0.5, called "half-Kelly". "If you are off by a factor of two on your risk of loss estimate, a full Kelly bet will reduce your return expectation to zero. But a half Kelly bet will leave you with 2/3 of the return expectation...With the full Kelly bet, you stand a ... 50% chance of losing 50%...at half the Kelly bet you stand .. a 25% chance of losing 50%...Your expected gain with the half Kelly bet is reduced by 25%" -- http://www.distressed-debt-investing.com/2010/04/kelly-formula-and-event-driven.html

(note that Kelly betting, although satisfying the solvency criterion (no Russian roulette), still has close to zero Arrow-Pratt risk aversion, so if you want more risk aversion you should definitely do something like fractional Kelly, at the least)

and also you have to discount for time

Note that above I am speaking the language of finance, but this applies to any sort of 'investment'. Notably, in life you should also consider the profit derived from an investment of time. One difference to consider is that after you've spent some time pursuing some goal, your benefit from further pursuit of that goal increases; this is why it's better to have one job than to try to have a portfolio of 1000 jobs.

Learning is valuable. One should consider optimizing one's rate of learning when formulating strategy. Two tips about learning: (a) a lot of it happens as a slow background process in your brain. If you start learning something long before you need it, and think about it a little bit from time to time, and then review it before you need it, you'll understand it much better by the time you need it just by the passage of time, compared to if you cram right before then. (b) Everything has a lot of details that you don't realize until you do it. So the way to learn many things it to do many instances of them, all the way thru. Just thinking about them isn't nearly as good.

About intellectual creativity: creativity requires time anti-discipline; you have to be able to drop everything and think when inspiration hits you

Good execs CARE about their companies.

qualia: the key to life is to appreciate the qualia of little experiences, such as the way airplanes sound, and the way the clouds look on a beautiful day

people's happiness seems to depend on:

• rewards for their effort
• control over their lives
• meaningfulness of what they are doing
• connection to other humans

people can avoid total unhappiness with their jobs by one or more of:

• having a job that makes them a lot of money
• having a job that they enjoy doing on a day-to-day basis
• working on a project that achieves an altruistic goal that they think is important
• working with people whom they like

It is very dangerous to accept a job with none of those qualities, thinking, well, that's just to support myself, what matters is the rest of my life. You spend a lot of time working at your job, and I know people who were quite unhappy with themself after spending a long time in a job with none of those attributes.

there are some contributors to happiness that one should be wary of, because they are not sustainable. It makes people happy to be better than almost all others, to be seen as better than almost all others, and to take vengeance on others. As the limit of population goes to infinity, there will be an infinite number of people who are better than you and seen as better than you, so these are not sustainable. Taking vengeance has a tendenacy first to stoke negative emotions and arrogance in oneself, and second to cause the target to feel wronged (perhaps unjustly), causing them to want to seek vengeance against you.

ethics is the only thing you control.