Quant500 @Quant500

Which of our five 20-stock portfolios would I hold into next week? Multi-factor. And the honest answer starts by rejecting the question. These rebalance monthly, not weekly. The current basket was bought on 17 August, it is 14 of 21 sessions in, and it does not change again until 15 September. Picking one "for next week" is picking a holding, not a week. Fifteen years walk-forward, 187 rebalances, 2011 to 2026: Multi-factor 20 — 16.75% a year · worst fall -36.50% · Sharpe 0.75 Momentum 20 — 14.83% · -36.93% · 0.66 Quality 20 — 14.26% · -33.44% · 0.70 Growth 20 — 13.97% · -34.79% · 0.63 Value 20 — 13.88% · -62.53% · 0.58 Multi-factor wins on return and on risk-adjusted return. That is the whole case. Now the uncomfortable part. Over the last three months Value is up 13.20% and Multi-factor 4.98%. Value is beating it, clearly, right now. And Value is the one that has fallen 62% before. That is exactly when people switch. The rule does not. The 20 it holds today: MU · WDC · STX · VLO · TER · PSX · NVDA · EXE · ALL · LITE MPC · LLY · HPE · FIX · PRU · CIEN · AMD · DELL · AMAT · LRCX Equally weighted, rebalanced monthly, chosen only from what was public on the day. Drawdown published next to the return, where it belongs. quant500.com/model-portfoli…

The joke has a real edge to it: dividend yield is not one of our four factors, and that was a deliberate call. We tested it. On its own it does not survive out of sample once you account for the sector concentration it drags in, and what is left of it is already inside the value factor. A quarter a share on NVDA is a rounding error on the thesis, and the market knows it.

Pricing a coin flip honestly is harder than it looks, and worth saying out loud from someone who publishes a forecast every day. We call the S&P 500's direction before each open and score it at the close. Over 1,066 out-of-sample sessions we are right 54.03%. Saying "up" every single day is right 53.94%. Nine hundredths of a point. Anyone selling you conviction on direction is selling you a coin.

On the earnings part specifically, there is a distinction almost nobody makes and it is measurable. The press release is an 8-K item 2.02. The audited accounts are the 10-Q, and they land days later. Yesterday the S&P 500 filed 406 documents with the SEC: not one 8-K carried item 2.02, and exactly one company filed accounts. Exploding earnings and filed earnings are two different calendars.

Genuine question, not a challenge: which definition of intrinsic value? We ran the comparison across every reasonable definition of the value factor on the S&P 500 and none of them beat the others out of sample. Twelve variants, twelve times no improvement. Disconnected from price is a claim you can only settle by naming the anchor first, and the anchor is where all the disagreement actually lives.

Index additions are the cleanest natural experiment in this business, and also the one that quietly flatters every track record. A name added today enters the index at its post-run price, and enters most databases with its whole history attached as if it had always been there. That is the same bias as deletions, pointing the other way. We score a new constituent only from the day we actually have its filings. It costs us numbers and it is the honest version.

The corporate version of this is Form 4, and it is filed within two days rather than 45. Last month across the S&P 500: 3,185 insider transactions at 398 companies. But only 103 were open-market purchases — someone spending their own money. Of the 1,055 sales, 602 were scheduled 10b5-1 plans signed months in advance. Most of what gets posted as an insider signal is a payroll event.

Deletions are the most underrated thing on this timeline, because almost every backtest you will ever read has quietly deleted them too. We rebuilt ours including the companies that left the index. Adding 49 delisted price series took our best 15-year portfolio from 17.29% a year to 16.75%, and every single portfolio came down. Half a point a year, permanently, just for counting the ones that did not make it.

Worth adding from the fundamental side, because it lines up with your thesis rather than against it. In our multi-factor model of the S&P 500, memory holds the top three places outright: MU is number one at 8.1 of 10, WDC number two at 7.6, STX number three at 7.2, out of 466 companies with all four scores. That is momentum, value, quality and growth from SEC filings. No thesis, no chart. Same names at the top.

@USDailyDinero option A ever

Yesterday, 4 Sep 2026, the S&P 500 companies we score filed 406 documents with the SEC. 216 were 424B2 prospectuses. 83 were Form 4s — insiders buying and selling. 30 were Form 144s. 22 were free writing prospectuses. 19 were Schedule 13G/A. 15 were 8-Ks. One was accounts: Best Buy. And not one of those 15 8-Ks carried item 2.02, so there was no earnings announcement in the whole session. Thursday had four sets of accounts and two announcements; Friday had one and none. So: 406 documents, one that moves a score, and no way to know which one without reading all 406. That is the job. Every trading day, free, no account, no ads. quant500.com

Every insider trade in the S&P 500, the day it is filed. 3,185 transactions across 398 companies in the last month: 103 open-market purchases — someone spending their own money 1,055 sales 2,027 grants, option exercises and tax withholding Here is the part almost nobody shows you. Of those 1,055 sales, 602 were scheduled 10b5-1 plans, signed months in advance and executed automatically. They say nothing about what the insider thinks today, and we label every one of them so you stop reading them as a signal. There is also a column for percentage of the stake, which is where the real information hides: Bezos sold $346.5M of $AMZN — 0.1% of his holding, scheduled. A Chevron director sold $157.4M — 11.3% of his, discretionary. Same headline. Opposite meaning. Filter by ticker, insider, role, transaction type or minimum amount. Every row links to the original SEC filing, so you never have to take our word for it. Download the whole table as CSV. No account, no sign-up, no ads. quant500.com/insider-alerts

$BBY filed its 10-Q yesterday, 4 Sep 2026. Best Buy. Momentum 4.723 → 4.455 · #260 → #294 Value 6.607 → 6.663 · #41 → #40 Quality 4.640 → 4.697 · #291 → #279 Growth 4.687 → 4.955 · #274 → #189 Multi-factor 5.163 → 5.193 · rank #143 → #136 It announced these results on 27 Aug. The market has known the numbers for eight days, six trading sessions, and only yesterday did the audited accounts land. That is when growth climbed 85 places. Momentum fell 34 in the same session and had nothing to do with the filing: that one only reads price. quant500.com/company-analys… The 10-Q: sec.gov/Archives/edgar…

24/7 trading changes one thing that is rarely discussed: the cost of turnover. We measured it on a factor strategy. Commissions and spread run 170 to 225 basis points a year at monthly rebalancing. Cutting turnover to save that does not improve the net result either — we tested 15 variants. More trading hours does not mean more edge, it means more friction.

Curious what the US equivalent looked like yesterday: four S&P 500 companies filed actual accounts with the SEC, and only one of them moved on fundamentals. Hewlett Packard Enterprise: quality up 37 places, growth up 15, on the day its 10-Q landed. The other three barely moved. Index moves are loud; fundamentals move on the filing calendar.

Flow data is one of the few honest signals out there, because someone actually moved money. The equity version is Form 4: yesterday 207 of them across the S&P 500, insiders buying and selling with their own money, filed the same day. Almost nobody reads them. We publish all of them, linked to the original SEC document.

On NVDA specifically, from the filings rather than the narrative: growth 7.4 of 10, quality 3.8, value 4.7. It is not cheap and it is not high quality by the accounting measures. What it is, is a growth score in the top 5%. Whether that is an anti-bubble depends entirely on which of the three you weight.

The order may be fixed but the timing is what kills you, and that is measurable. We tested trend filters and volatility targeting on top of a factor portfolio, the standard "get out before the contraction" overlays. Both erase the edge completely versus just holding the rule. The cycle being real and the cycle being tradeable are different claims.

We publish a directional call every morning and then score it at the close. Over 1,066 out-of-sample sessions we are right 54.03%. Saying "up" every single day is right 53.94%. Nine hundredths of a point. Direction is close to unpredictable, ours included. Where there is real signal is in the size of the move: 59.94% vs a 51.31% baseline.

Worth adding on the fundamental side: MU is the number one multi-factor name in the S&P 500 today and WDC is number two, out of 466 companies with all four scores. Different method entirely, same two names at the top. Ours scores momentum, value, quality and growth from SEC filings, no charts involved.