Steversteves — History of the U.S. told by SPX (TradingView, 24 Sep 2023)

A published chart idea by steversteves on tradingview: a century of American history narrated through the S&P 500, from the Roaring Twenties to the 2022 bear market, with a regression channel drawn over the whole span and about twenty chart snapshots along the way. Arrived via Telegram.

Tier: T3. Self-published retail commentary, first person, no dataset named and no method stated for the regression it reasons from. Not T4 — the author did the work he describes ("horrendous hours of research"), the moves and durations are specific, and nothing here is assembled from other listicles. Not higher, because every quantitative claim rests on a line drawn on a chart. For the historical facts themselves, better sources exist and this is not one of them; what it is good for is recorded below.

What it says

The narrative runs 1920s → 1930s Depression → the 1932–36 recovery and the 1937 relapse → WWII → the postwar bull run → the corrections of the 1950s–60s → the 1970s "lost decade" → the 1980s, Black Monday → the dotcom bubble → the housing bubble and Iraq → 2008 → COVID → 2022.

The specific numbers it gives, all read off the author's charts rather than from a cited dataset:

• 1932–1936 recovery: +316% over 1,676 days, halted by the 1937 recession.
• WWII: the Dow rose 10% on the first trading day after the invasion of Poland; fell ~2.9% after Pearl Harbor and recovered it within a month; +50% start to end of the war.
• 1953: a ~15% fall from January to September, which he calls the first correction not caused by a catastrophe, "just simply a cumulative effect of various small things."
• Early 1980s: +50% in ~273 days, then a 1981 correction, then +200% over 1,827 days.
• 2003–2008: the index roughly doubles over five years while LMT outpaces it by nearly 100% and RTX grows over 200% — his evidence for "war tends to be looked at positively by the market", which he flags with his own eye-roll.
• 2008: −55% over 518 days. 2020: −35% in days, then +123% in 701 days, which he attributes to quantitative easing and stimulus.

The argument

Two claims, and they are different in kind.

The market always cares about the same five things — interest rates, inflation, geopolitics and economic policy, war, corporate earnings. He arrives at it by noticing that the causes he lists for the 1950s and 60s corrections are the causes he lists for the recent ones: "the same concerns that plagued the market in the 50s, 60s and 70s are still the ones affecting us today." This is unfalsifiable as stated — any market move can be assigned to one of five buckets — and it is the piece's headline.

Every correction has regressed to the mean, and 2022 did not. This one is falsifiable, and it is why the piece is worth a page. He checks it under two different trend specifications — a quadratic fit and a log-linear fit — and reports that both say the same thing: prior corrections returned to the fitted mean, the 2022 bear market did not. If the index kept rising from there, he says, it would be "among the first bear markets in SPX history to not have undergone a regression to the mean... a bear market that accomplished, well, nothing."

Why it earns a page here despite the tier

It is the corpus's clearest specimen of the untested version of a claim the corpus can test. mean-reversion holds Chan's treatment: reversion is a property you check before building anything, with the ADF test, the Hurst exponent, the variance ratio test and the half-life. This piece asserts reversion over a century of index prices and uses none of them. "The mean" is whatever the fitted line says it is, and the author's own presentation shows how free that choice is — he fits it two ways and shows both.

Two things follow that are to his credit and are the reason the page says specimen and not blogspam. He fits it twice instead of once, which is more model-checking than the format usually carries. And his conclusion is a negative result about his own thesis: after arguing that corrections revert, he ends by flagging the case that didn't. Set against ai-trading-bot-platforms-2026, which is what T4 in this spoke actually looks like, the difference is visible.

It also gives market-regime-analysis its longest horizon by far. That page's regimes are months long, labelled by hand in smolinski-h1-2026-aar and fitted with Gaussian mixtures in multithreaded-ml-trading-robot. This one runs a century and labels regimes by narrative — the Depression, the lost decade, the never-ending bull run — which is the same operation at a scale where nobody can backtest it.

The claim has not been checked

The piece was written in September 2023 and its only falsifiable claim is about what happens next. Nothing in this corpus records where the S&P 500 went after that date, so the prediction sits open: three years on, this wiki cannot say whether the 2022 bear market ever did revert to either of his fitted means. That is a gap in the spoke, not in the source — and closing it needs a price series and a stated fitting method, which is exactly what the piece itself lacks.

Related

mean-reversion · market-regime-analysis · backtesting · backtest-overfitting · steversteves · tradingview · smolinski-h1-2026-aar · algorithmic-trading · synthesis