Markets and economy in the age of thinking machines. PhD macroeconomics. Angel investor.
Joined January 2013
- Tweets 9,247
- Following 91
- Followers 162,918
- Likes 4,870
Why so many recommending "The Lessons of History" on here?
This book is basically what happens when your grandpa recently took a pattern recognition class and proceeds to explain entire human civilization with the confidence and rigor of a horoscope reader.
On large gap-down days, market becomes difficult to trade in either direction because of all the institutional cross-currents creating extreme chop. Just avoid those days all together. Your EV is likely negative. Not worth the stress.
Large gap up days don't have same problem.
Seems true to any field—
Outsider sees magic.
Insider sees crude patcher-upper.
Marketing sells dream.
Reality is a mixed bag of all above.
If you are actually good, then being “promoted” to management is like the most benign looking death trap ever.
This, of course, assumes you have a viable strategy that you have carefully thought through in the first place. If that's the case, then it is more optimal to:
1. Complete the strategy validation through extensive paper trading at full size
2. Use paper trading to identify and resolve any technical or operational issues
3. Once you're ready to trade the complete strategy, deploy it with full size live.
By deploying at full size immediately, you face the complete challenge upfront, allowing you to focus entirely on adapting to the strategy's true characteristics rather than managing an artificial scaling process. This approach, while potentially more stressful initially, provides clearer feedback and a faster path to strategic competence.
Scaled deployment extends the period during which you're not collecting meaningful performance data. Since the strategy's actual risk-adjusted returns can only be properly evaluated at full size, starting small delays your ability to validate the strategy in live trading.
Trading at reduced size can also create false confidence. Early success with smaller positions might not accurately represent how you'll react to full-size trades, potentially leading to overconfidence before the real challenge of managing full positions begins.
The compounding effect of multiple scaling adjustments can obscure the strategy's true performance characteristics. Each size increase introduces a new variable, making it harder to distinguish between strategy performance issues and normal statistical variance.
Scaling up gradually doesn't eliminate this challenge; it merely fragments it into multiple smaller but prolonged adjustment periods, potentially extending the total psychological stress over a longer timeframe.
A strategy's expected performance metrics are based on specific position sizing rules. Trading at reduced size means operating outside these tested parameters, potentially leading to premature strategy adjustments based on statistically insignificant results.
Position sizing is an integral component of any trading strategy, as carefully developed and integrated with entry and exit rules. Artificially constraining position size means you're not actually trading the strategy you tested, but rather a different system with modified risk parameters.
The psychological challenge of watching larger P&L swings needs to be addressed eventually.
Starting small and gradually scaling up creates multiple psychological adjustment periods. Each time you increase position size, you must readjust to new profit and loss magnitudes, essentially forcing yourself to adapt repeatedly to what is fundamentally the same trading setup.
Why "starting small and working your way up" may not be optimal when deploying new trading strategy.
The temptation to scale up gradually feels safer. The fear of large, immediate losses and the need to "get comfortable" make this approach seem responsible.
However...
The problem with “build things people want” is— if you are too good at it, you may just forget why you got started in the first place. And without that reason there’s no point building anything at all.
Had problem in code. A tricky one. Gave it to Claude. Couldn’t solve after multiple tries. Gave it to o1 pro. Did it after 3 attempts.
May keep paying the $200 after all.
Sure, Shakespeare is just some arrangements of 26 letters. Very trivial indeed.
People have too inflated sense of what it means to "ask an AI" about something. The AI are language models trained basically by imitation on data from human labelers. Instead of the mysticism of "asking an AI", think of it more as "asking the average data labeler" on the internet.
Few caveats apply because e.g. in many domains (e.g. code, math, creative writing) the companies hire skilled data labelers (so think of it as asking them instead), and this is not 100% true when reinforcement learning is involved, though I have an earlier rant on how RLHF is just barely RL, and "actual RL" is still too early and/or constrained to domains that offer easy reward functions (math etc.).
But roughly speaking (and today), you're not asking some magical AI. You're asking a human data labeler. Whose average essence was lossily distilled into statistical token tumblers that are LLMs. This can still be super useful ofc ourse. Post triggered by someone suggesting we ask an AI how to run the government etc. TLDR you're not asking an AI, you're asking some mashup spirit of its average data labeler.