Every swing-trading checklist opens with the same gate: only trade when the market itself is in an uptrend. The usual definition is the S&P 500 above its 200-day moving average. Above it, momentum is "high", breakouts are supposed to follow through, and setups are supposed to work. Below it, momentum is "low", and the advice is to trade small or not at all.
That is a testable claim, so we tested it. Across 22 US-listed stocks and sector ETFs, over the window the free backtester covers, we ran three common entry signals, held each trade ten sessions, and split every trade by which regime the index was in on the entry day: 2,630 signal trades, judged against 2,024 baseline trades from the regime rule alone.
The folklore came out backwards, inside the window the gate can see. Both halves of that sentence matter, and the second one first.
What the gate can and cannot see
A 200-day average needs 200 sessions of index history before it exists, and the backtester's window opens in December 2021. So the first day either regime can be classified is 2022-09-16, and the 199 sessions before it, which contain the entire January to September 2022 decline, are warm-up: no trade in this study, signal or baseline, comes from them.
That is not a footnote, because Jan-Sep 2022 is the one stretch in this window where the market fell and kept falling, which is exactly the market the folklore warns about. It was in the data, and the method could not see it. Every low-regime figure below therefore describes below-average days from mid-September 2022 onward, a sample that begins 19 sessions before the 2022 bear market bottomed. Keep that in front of you while reading the rest.
Being long below the 200-day average beat being long above it
Before asking whether any signal works, you need the base rate for each regime: be long whenever the regime is in force, re-entering every ten sessions, no signal at all. That is the number a signal has to beat. (These are systematic ten-day cycles anchored to each regime stretch, not randomly timed entries.)
| Ten-day holds, entered while the S&P 500 was | Trades | Finished green | Median drawdown |
|---|---|---|---|
| above its 200-day average ("high momentum") | 1,672 | 56.6% | -28.8% |
| below its 200-day average ("low momentum") | 352 | 64.2% | -16.8% |
The drawdown gap comes with an exposure gap: the high-regime baseline was invested about 66% of the window, the low-regime one 14%.
The "hard" regime had the higher win rate, by nearly eight points. The reason is what the classifiable below-average stretches were: late September 2022 to January 2023, the SVB weeks in March 2023, October to November 2023, the spring 2025 selloff, and a stretch in spring 2026. Every one of them resolved upward, so ten-day holds entered on below-average days kept catching rebounds. And the sample is thinner than 352 suggests: those trades land on just 16 distinct entry dates, each one all 22 names entering together on the same synchronized ten-day cycle.
The honest limitation stays attached to every low-regime number on this page: a sample of below-average days that starts two and a half weeks before a major bottom, in a window whose every visible dip recovered, guarantees this result. In a sustained decline like the one hidden in this study's own warm-up, the same arithmetic runs in reverse. What the test rules out is only the folklore's confident version, that short holds entered below the 200-day average systematically fail. In everything this window could classify, they did the opposite.
The three signals, split by regime
Same universe, same window, same ten-session hold, same costs as the rest of this series. The regime gate is the backtester's own "S&P 500 vs its moving average" rule, so every row is reproducible in the tool.
| Signal | Regime | Trades | Win rate | Time in market | Median drawdown |
|---|---|---|---|---|---|
| 20-day breakout | high | 811 | 59.6% | 31.4% | -16.4% |
| 20-day breakout | low | 80 | 71.3% | 3.5% | -7.2% |
| 20>50 EMA cross | high | 166 | 57.8% | 6.5% | -11.2% |
| 20>50 EMA cross | low | 32 | 81.2% | 1.7% | -2.2% |
| 1-2 bar pullback | high | 1,263 | 57.7% | 49.0% | -21.8% |
| 1-2 bar pullback | low | 278 | 66.2% | 11.4% | -15.4% |
Median whole-window returns, at the exposure shown in each row: the breakout made +3.3% in the high regime and +6.7% in the low, the EMA cross -1.1% and +2.4%, the pullback +19.7% and +11.8%. Buy-and-hold made +57.3% over the same span, invested every day; at single-digit exposures these are different products, not competing scores.
Two things are true in every row. The low-regime version of each signal won more often than its high-regime version. And the gaps that look dramatic sit on small, correlated samples: the 81.2% trend-cross figure is 32 trades on 29 distinct days, and the 80 low-regime breakouts include 12 names firing together on 2025-05-02, one market call counted twelve times.
No signal clearly beat the regime it was in
The right comparison for each row is not 50%, and not the whole-window base rate either. An up-regime day already carries the market's tailwind, so a signal only earns credit for what it adds on top of being long in that regime at all. Against their own regime's base rate:
| Signal vs its regime's base rate | Edge | p-value |
|---|---|---|
| Breakout in high momentum | +3.0 points | 0.16 |
| Breakout in low momentum | +7.1 points | 0.23 |
| Trend cross in high momentum | +1.2 points | 0.77 |
| Trend cross in low momentum | +17.0 points | 0.05 |
| Pullback in high momentum | +1.1 points | 0.55 |
| Pullback in low momentum | +2.0 points | 0.60 |
Every edge is positive and none of them is clearly distinguishable from zero, and the p-values assume independent trades, which these are not: signals cluster on the same days across correlated names, so the true uncertainty is wider than printed.
In the high regime, where the samples are large, the signals added one to three points over just being long, and could not prove even that. In the low regime the gaps run bigger, up to seventeen points for the trend cross, but on 32 to 278 trades squeezed into a handful of bounce episodes, which is why even the +17 does not clear the bar. The comparison that survives all of it: the regime split moved win rates by around eight points; no signal demonstrably moved them at all. Which condition the market was in mattered more than which trigger got you in.
What the regime actually governs
If the win rates refuse to support "only trade above the 200-day average", the exposure and drawdown columns explain why the advice survives anyway.
A regime gate barely changes how often you win. What it changes is how much of the time you are in the market, and what happens to you while you are. The gated breakout spent 3.5% of the window invested in the low regime and its median drawdown was 7.2%; its high-regime twin spent nine times longer invested and drew down twice as deep. The ungated pullback rule, in the market 70% of the time across both regimes, took a 33% median drawdown, close to buy-and-hold's 33.8% over the same span.
So the 200-day gate is real risk management wearing the costume of a win-rate edge. It decides your exposure, and in a falling market that keeps falling (the case this study's warm-up hid from it), exposure is what kills accounts. That is the defensible version of the checklist rule: trade the low regime rarely and small, not because your entries will fail more often, but because the tail risk while you hold is the one thing these tables cannot price.
How to tell which regime you are in
You do not need a feeling for this; it is one comparison. The index against its own 200-day simple moving average, checked in any charting tool, or as the "S&P 500 vs its moving average" rule under Market regime in our free backtester. Screeners like Finviz show the same thing at a glance, and TC2000 users typically keep it as a standing condition in every scan; Qullamaggie's TC2000 settings are the best-known worked example. If you want single names ranked by momentum rather than the index's state, the momentum screener recomputes the whole large-cap universe nightly.
Test it yourself
The base-rate rows are one rule in the free backtester: open Chart setup, add S&P 500 vs its moving average, set it to over or under a 200-day SMA, and run a ten-bar hold. The signal rows add a second rule on top: Donchian channel break, Moving average cross, or Corrective pullback. Flip the over/under chip and run it again. The interesting experiment is the same one this page ran: compare each signal not to 50% but to the regime rule alone, which is the base rate the signal has to clear. The tool's window rolls forward daily, so your numbers will drift from these.
Frequently asked questions
Is a high momentum market easier to trade? Not by win rate, in the part of our window the gate could classify (September 2022 onward). Ten-day holds entered above the 200-day average won 56.6% of the time; the same holds entered below it won 64.2%, because every classifiable below-average stretch resolved upward. What the high regime offers is more time invested and more signals, not a better hit rate per signal.
Should I stop trading when the S&P 500 is below its 200-day moving average? The win rates on this page do not support stopping. The drawdown and exposure numbers support caution: the gate's real effect was keeping positions rare and small in the regime where a market that keeps falling would do the most damage. The one such market in our window sat inside the indicator's own warm-up period, so the gate's worst-case benefit is exactly the thing this test could not measure.
What is the best indicator for market momentum? The index against its 200-day average is the standard, and it is hard to beat for one number. In this study the regime split separated win rates by about eight points, while none of the three entry signals produced an edge over its own regime that the sample could confirm.
Why do breakouts fail in choppy markets? The premise did not hold in our data. Low-regime breakouts won 71.3% of their ten-day holds, against 59.6% in the high regime. They were also ten times rarer, and twelve of the eighty fired on a single day, so "breakouts in weak markets" mostly meant "the first thrust off a low", which is a different and better trade than the folklore imagines.
Before you trade this
These figures come from one run over 22 US-listed large-cap stocks and sector ETFs, December 2021 to mid-2026. The regime gate needs 200 sessions of warm-up, so the split only covers September 2022 onward, and the window's one sustained decline (January to September 2022) is invisible to it; every below-average stretch the gate could see ended in a recovery, and a sustained bear market would flip the low-regime numbers by an amount no run over this period can estimate. Signals cluster on the same days across correlated names, so every sample is effectively smaller than its trade count, most severely the low-regime base rate's 16 distinct entry days. Win rates are before costs; the return and drawdown medians include commission and slippage. The backtester's window rolls forward daily, so re-running these rules will not reproduce these exact figures. The run behind every table is committed in the repository (backend/scripts/research/regime_study.py) so it can be repeated rather than trusted. Past prices are not a forecast, and nothing here is investment advice.