David Chau

Saw the estimate that levered ETFs need to sell about $32B into the close today. Biggest one on record apparently. SOXL down around 30% today. Because of my ADHD, I ran a quick study. Since 2020 only. Every SOXL day that closed down 15% or more. That's the kind of day that forces this selling. It has happened 21 times. If you bought the close on those days, the median return 12 months later was +118%. At 3 months the median is only +10%. That's actually worse than buying SOXL on a random day. And for the really ugly days, down 20% or more, the median at 3 months is negative. The money in this trade has always been at month 6 to 12. Never at month 1. The underlying is the cleanest part of the whole study. SOXX was positive 12 months later 17 out of 17 times. SOXL still found a way to be down a year later about 1 in 5 times. July 2024 is the example I keep coming back to. SOXX up 1.6% over the next year. SOXL down 47% over the same year.

I ran a quick study on SPX because I wanted to know one thing. After a red week, what percentage of the time does week 2 close above week 1's high? So I went through every red week since 2022. 104 weeks. Week 2 closed above week 1's high only 24% of the time. The other 76% of the time it either closed inside week 1's range or below week 1's low. And that is with week 2 still closing green about half the time. That was the part that got me. Week 2 can be green and still never really recover week 1. This year it has been even more one sided. 16 red weeks so far in 2026 and week 2 has closed above week 1's high just once.

RUT IC spreads today approaching max profit today too 🙌

Yesterdays IC expired max profit today.

SPX iron condor, entered at the close, expiring the next day. 20-delta shorts, $5 wide wings, collects about $170 against $330 of risk. 1,091 trades, May 2022 through September 2026. Unfiltered, it has no edge. EV is basically zero, 64% win rate. The median trade wins $158, and the tail losses hand it all back. That is before commissions. If you sell these every day on autopilot, you are donating. But one condition changes everything: how much SPX moved yesterday. When SPX moved 1% or more the day before entry, the same condor made +$40 per trade across 278 trades. Positive in all 5 years. At 1.5% or more, +$44. And it is monotonic. Big prior move: +$32. Medium move: -$12. Quiet day: -$21. The calmer yesterday was, the worse the condor does today. The logic is simple once you see it. After a big day, implied vol stays scared while the market usually calms down. You are selling yesterday's fear at tomorrow's prices. On quiet days there is no fear premium to collect, but the tail risk is still there. Same $170 credit, none of the edge. Two more findings. Friday entries were the best day (+$29, weekend decay into the Monday expiry) and Wednesday entries the worst (-$31). And a high VIX minus realized vol spread, which every premium seller is taught to love, was a losing condition here (-$15). In a 1DTE condor, that spread usually means live stress, not rich premium. Best combo: prior-day move of 1% or more on a Thursday or Friday entry. +$65 per condor across 132 trades. Honest caveats: daily closes only, no commissions or slippage modeled, and I sliced the data a lot of ways. The prior-move result is the one I trust because it is monotonic, it worked every year, and the mechanism makes sense. Historically realize vol is always under implied vol, so selling condors when realize vol is above will naturally revert.

You guys asked for crypto, so I ran the %B bottom study on BTC and ETH. The answer is uncomfortable. Same setup as the SPY study. 6.75 years of dailies. 11-day pivot windows. Bottoms confirmed by a 5% bounce within 10 days. %B(20,2) under 0.15 as the washout flag. BTC: 125 confirmed bottoms. The flag caught 44 of them. 35%. ETH: 129 bottoms. Caught 39%. Compare that to 62.5% on SPY. The rule does not port. Look at where bottoms actually print. Median %B at a BTC bottom is 0.26. On SPY it is 0.10. Crypto bottoms do not get as washed out relative to the bands, because the bands themselves are enormous in a volatile market. The 0.15 line was fit to SPY's personality. Two more problems. Gold was worse than crypto: 32% recall, and the flag fired on 23% of ordinary gold days. In a slow grindy market the flag trips constantly and means nothing. And the 5% confirmation that works for SPY is noise for crypto. BTC printed 125 "bottoms" in 6.75 years. Most of those are wiggles, not washouts. You would want a 15-20% confirmation move before the pivots mean anything on BTC. The one thing that survived everywhere: specificity. Across 345 tops in BTC, ETH, GLD, and SLV, the flag fired on 3. It still does not mistake tops for bottoms. It is just much worse at finding bottoms anywhere but SPY. The lesson I keep relearning: indicators have personalities. A rule found on one market is a hypothesis about that market, not a law. Test it on the new market before you trust it there. Sources: price data from Yahoo Finance (GLD, SLV, BTC-USD, ETH-USD daily bars, 2019-2026). SPY reference figures from the same study using Yahoo Finance and Polygon. Pivots defined as the lowest low in an 11-day window, confirmed by a 5% rally within 10 days. %B computed on 20-day, 2 standard deviation Bollinger Bands at the daily close.