# Thomas Kwa

Karma: 458

Student at Caltech. Trying to level up.

Submit anonymous feedback here (try to be nice): https://​​www.admonymous.co/​​tkwa

• My fairly low-effort attempt:

My goal here was roughly to minimize the amount of manual analysis. I used Python to graph the mean revenue by species and age, and also estimate the distribution along which other bidders bid.

Everything seems roughly linear, with perhaps some nonlinearity in winter wolf bids. Given that the sample mean revenue should be an unbiased estimator of revenue, and the probability of winning for a given bid should also be an unbiased estimator, we can use linear regression for each species to estimate revenue, and logistic regression on (age, bid) to estimate the probability of winning a given auction with a given bid.

Using logistic regression implicitly assumes the distribution is logistic, which might be problematic but seems okay. Since we want to maximize expected profit = p_win(bid) * (revenue—bid), we can throw this into the scipy black-box optimizer to find the bid price that maximizes profit. This gave us the following bids:

One could probably do better by actually looking at the distribution of non-carver winning bids, but I don’t know the best way to account for selection (using the distribution directly would create a biased distribution, since they’re always higher than the carver bid).

• Why would you? The point of the box spread trick is to reduce interest on margin loans, they don’t increase your buying power. Selling box spreads without a margin loan just makes you pay small amounts of interest for no reason.

• Let’s define Levin complexity as , where l is description length and t is execution time

Should this be l + f(t)?

• I’m confused too. Say you sold 10 32003400 box spread contracts for $196k a while ago, for a “par value” of$20k per box or $200k total. Since the exact box spread you sold is rarely traded, the best bid in the order book will be the sum of the 4 individual leg bids, and likewise the best ask will be the sum of the 4 leg asks. This makes the spread very wide, maybe$15k bid and $25k ask, and I’m not sure how IBKR values your position, maybe it just takes the midpoint. Let’s say it’s normally -$197k at some point in time.

Now say someone buys a 32003400 box for $25k. I’m guessing that IBKR’s algorithm could see the last traded price, decide your position is worth$-250k now, and liquidate you. If you keep an open order to buy back the box for say $16k each, nothing changes and in fact the midpoint price would increase. Maybe the commenter meant to say you should keep an open order to sell more box spreads for slightly above par value? • I agree that closed physical systems aren’t optimizing systems. It seems like the first patch given by the author works when worded more carefully: “We could stipulate that some [low-entropy] power source [and some entropy sink] is provided externally to each system we analyze, and then perform our analysis conditional on the existence of that power source.” Then an optimizing system with X bits of “optimization power” (which is log(target states /​ basin of attraction size) or something) has to sink at least X bits, and this seems like it works. Maybe it gets hard to rigorously define the exact form of the power source and entropy sink though? Disclaimer: I don’t know statistical mechanics. • I haven’t let one expire yet, but that looks right. I don’t anticipate any major problems because none of the sources I’ve seen mention any, and besides all legs expire on the same day. Note that the loan value already has to be within margin borrowing power throughout the entire duration of the loan, since your broker takes into account the box spread as a negative-value asset when calculating margin. Box spreads can’t get you more borrowing power (it’s box spread financing, not box spread borrowing). • My understanding is that each leg of the box spread is marked-to-market, resulting in a small overall capital loss equal to the interest you’re paying to the holder of the long box. For example, the market is up this year, so the bullish legs gained value and the bearish legs lost slightly more value. This is how the Interactive Brokers tax forms work; on the section-1256 section of the 1099 form, the line “Unrealized profit or (loss) on open contracts − 12/​31/​2020” has a loss of about$2000. If interest rates rise so much during a year that the interest is negative, I would expect there to be a small 6040 capital gain. I could be missing something though.

• (For example, imagine a u-shaped craft with a low center of gravity and helicopter-style rotors on each tip. Add a third, smaller propeller on a turret somewhere for steering.)

Extremely minor nitpick: the low center of gravity wouldn’t stabilize the craft. Helicopters are unstable regardless of where the rotors are relative to the center of gravity, due to the pendulum rocket fallacy.

• Depends whether you’re holding it as SPY/​VOO or as a mutual fund (assuming you don’t mean SPX futures, and SPX itself isn’t tradeable). IBKR has somewhat higher margin requirements than the minimums set by the OCC; I think they require 25% maintenance margin for mutual funds and 10-15% for broad-based stock indices.

• This article says executions should occur at 0.29% above treasury yields. Is anyone able to consistently get executions at that rate?

A few weeks ago I was filled at about 0.51% above treasury yield, which is definitely worse than the 0.29-0.35% I previously got. Both were using IBKR smart routing and there weren’t any obvious differences, so I’m guessing box spread rates are driven by supply and demand, and there’s just relatively more supply now than in mid-2020.

• Does c3.ai believe in the scaling hypothesis? If gwern is right, this could be the largest factor in their success.

• Thanks. I tried a couple different switches on my setup (3.5mm through USB-C hub), and the computer didn’t disconnect upon opening the switch, so I’m giving up on this until I change hardware.

• Is it possible to make an hourglass that measures different amounts of time in one direction than the other? Say, 25 minutes right-side up, and 5 minutes upside down, for pomodoros. Moving parts are okay (flaps that close by gravity or something) but it should not take additional effort to flip.

• It has been noted that Uighurs are Muslims, and the reeducation camps (or whatever they are) are largely aimed at eradicating religious extremism (along with ethnic separatism), and yet the countries that are complaining about their situation are the post-Christian countries of the American bloc, not the Muslim countries of the OIC.

I agree that OP’s plan is unlikely to do much good, but I strongly disagree with both the direct meaning and implications of the above sentence. My understanding is that “eradicating religious extremism” is simply the CCP party line and nobody really believes it. Also, whatever the aims of the CCP, we have nearly incontrovertible evidence that their actions include severe human rights violations on a large scale. I’ve also heard that the fact that the OIC supports China’s actions in Xinjiang is response to Chinese bullying, not a reflection that they think such actions are good.

I’m downvoting this comment because it either (a) shows a lack of caring about human rights, or more likely (b) is needlessly unhelpful because it doesn’t start the brainstorm of ways to do good that realistically mesh with the CCP’s strategic concerns.

• Is there software that would let me automatically switch between microphones on my computer when I put on my headset?

I imagine this might work as a piece of software that integrates all microphones connected to my computer into a single input device, then transmits the audio stream from the best-quality source.

A partial solution would be something that automatically switches to the headset microphone when I switch to the headset speakers.

• I’m planning to write a sequence on all positive and negative effects the practice of rationality has had on my life, and I already have one post on pitfalls. Future posts will probably be about things like

• forming realistic expectations

• a list of all the low-hanging fruit I’ve found

• a list of interventions I’ve tried

• weird effects from learning some rationality techniques but not others

• the benefits of diversifying one’s sources of knowledge

• re-learning how to interact with non-rationalist society

These will take a while to gestate and write up, but I expect to have a significant amount of content out before this time next year. Of course, this doesn’t replace a community survey, but I think a case study with emphasis on the practical will contribute significantly.

• Thank you for writing this post. It can be hard to realize that one is in a persistent rut and harder to write it up publicly.

As someone similar to you (math competitions during high school, CS major, ambition often outruns my capacity, trying to avoid reverting to the mean during my adult years), I believe that I’m susceptible to failure modes like this. DM me if you want to video call to exchange advice or something; this would be valuable enough to me that I’m willing to pay a fair rate if you need funding.