Your momentum screen worked for six months. Then the market rolled over and every name it surfaced dropped 15% in two weeks. The screen did not break. The regime changed and the screen did not notice. Regime factor screening solves this by switching which factors you emphasize, and how you weight them, based on the market environment you are actually in.
I run factor screens weekly. For most of 2025, a momentum-heavy composite kept surfacing winners. Then Q4 turned choppy, breadth narrowed, and that same screen started producing whipsaws. I did not abandon factor screening. I changed the weights. Momentum dropped from 50% to 20%. Mean reversion and low-volatility factors moved up. The watchlist quality improved within two weeks. That shift was not a guess. It was driven by a simple regime classification I will walk through below.
Why a Static Factor Screen Fails Across Regimes
A single-weight factor model assumes the relationship between factors and forward returns stays constant. It does not. Academic research going back to Asness, Moskowitz, and Pedersen (2013) showed that momentum returns cluster in trending markets and crash during reversals. Value returns tend to lag in strong momentum environments and recover when momentum unwinds. Quality outperforms during stress periods when investors flee to safety.
This is not an edge case. It is the normal behavior of factors across market cycles. The problem is that most retail traders build one screen, test it on a favorable backtest window, and run it unchanged until it stops working. By the time they notice, they have already given back months of gains.
I made this mistake in early 2022. I was running a momentum-value composite that had worked well through 2021. When rates started rising aggressively, momentum names (mostly growth stocks) got crushed while value rotated hard into energy and financials. My screen kept surfacing tech names with strong 6-month returns. Those returns were about to reverse. The screen’s inputs were backward-looking, and the regime had already shifted forward.
The fix is not to predict regimes perfectly. It is to classify the current regime with a simple, rules-based system and adjust factor weights accordingly.
Classifying the Market Regime
You do not need a machine learning model for this. Two indicators and a volatility measure give you a workable regime classification.
Trend state: is the broad market above or below its 200-day simple moving average? Above means trending up. Below means trending down or transitional. I use SPY as the proxy. When SPY sits above its 200-day SMA, I label the regime “trend.” When it sits below, I label it “defensive.” That is the first branch.
Volatility state: is the market in a low-volatility or high-volatility environment? I use Bollinger Band Width on SPY with a 20-period lookback. When BBW is below its 6-month median, volatility is compressed. When it is above, volatility is elevated. You could also use a Choppiness Index reading above 61.8 to flag range-bound conditions, which often coincide with volatility compression before a directional move.
Combining these two dimensions gives you four regimes:
1. Trend + Low Volatility: the smooth rally. Momentum dominates. Stocks that are already moving tend to keep moving. Quality is less important because even mediocre names get lifted.
2. Trend + High Volatility: the volatile rally or early recovery. Momentum still works but with wider drawdowns. Quality starts to matter because weak names get shaken out on pullbacks.
3. Defensive + Low Volatility: the grinding range. Momentum fails. Mean reversion and value tend to outperform. Look for names trading at the bottom of their range with solid fundamentals.
4. Defensive + High Volatility: the crash or correction. Quality dominates. Low-volatility names outperform. Momentum is toxic because the biggest recent movers tend to fall the hardest.
This framework is not original. It draws on research by Ang, Hodrick, Xing, and Zhang (2006) on volatility risk premia and by Daniel and Moskowitz (2016) on momentum crashes. What I have done is simplify it into something that takes 30 seconds to classify each week.
Regime Factor Screening Weight Maps
Once you know the regime, you adjust factor weights. Here is the weight map I use across four factors: momentum, value, quality, and low volatility.
Trend + Low Volatility: momentum 50%, value 15%, quality 15%, low volatility 20%.
Trend + High Volatility: momentum 30%, value 15%, quality 30%, low volatility 25%.
Defensive + Low Volatility: momentum 10%, value 40%, quality 25%, low volatility 25%.
Defensive + High Volatility: momentum 5%, value 20%, quality 40%, low volatility 35%.
These are not optimized in a backtest. They are heuristics I developed over three years of running weekly screens. Optimization on historical regimes tends to overfit because regime transitions are messy and the sample size for each regime is small. The logic behind each allocation is simple: lean into the factor that historically performs best in that environment, and reduce exposure to the factor that historically underperforms.
Notice that momentum never goes to zero. Even in a crash, some stocks are falling less than others, and that relative momentum still contains information. But it drops from the dominant factor to a minor input.
One mistake I see traders make: they treat low volatility as the absence of a factor. It is not. Low-volatility screening means actively selecting names with below-average realized volatility or beta. In defensive regimes, this factor is your primary risk control. Stocks with a 90-day realized volatility below the market median tend to draw down less during corrections, which preserves capital for the recovery.
Implementing the Screen in Practice
Here is the weekly workflow. It takes about 25 minutes once you have the spreadsheet set up.
Step 1: Check SPY’s position relative to its 200-day SMA. On May 8, 2026, SPY closed at $737.62. If the 200-day SMA sits around $690 (a reasonable estimate given the 2025-2026 trend), SPY is well above it. Trend regime confirmed.
Step 2: Check Bollinger Band Width on SPY. If BBW is below its 6-month median, label it low volatility. If above, high volatility. Suppose BBW is currently compressed after a multi-week consolidation. That gives us Regime 1: Trend + Low Volatility.
Step 3: Pull factor data for your universe. I screen the S&P 500. For each stock, compute four percentile ranks: 6-month momentum (skip the last month), earnings yield for value, ROE minus debt-to-equity for quality, and 90-day realized volatility inverted for the low-volatility factor.
Step 4: Apply the regime weights. For Regime 1, the composite score is:
C = 0.50 \times R_{mom} + 0.15 \times R_{val} + 0.15 \times R_{qual} + 0.20 \times R_{lowvol}
Step 5: Sort descending. The top 20 names go onto the watchlist. Then check charts for RSI readings and price structure before taking any position.
The spreadsheet columns stay the same every week. Only the weights in the composite formula change when the regime changes. I keep four saved formulas, one per regime, and swap them based on the weekly classification.
When the Regime Shifts Mid-Week
Regime classifications are weekly decisions, not daily. Checking daily creates noise. SPY can dip below the 200-day SMA for two days and recover. That is not a regime change. That is volatility.
I require two consecutive weekly closes in the new regime before switching weights. If SPY closes below its 200-day SMA on Friday, I note it but keep the current weights. If it closes below again the following Friday, I switch. This two-week confirmation rule prevents whipsawing between weight maps during chop.
The exception is a volatility spike. If Bollinger Band Width jumps above the 90th percentile of its 6-month range within a single week, I treat that as an immediate regime shift on the volatility axis. Large volatility expansions tend to be fast and one-directional. Waiting two weeks to acknowledge them is too slow. The 2020 crash, the 2022 rate shock, and the August 2024 carry-trade unwind all showed BBW spikes that reached the 90th percentile within days.
When I do switch, I do not immediately rebalance the watchlist. I stop adding new positions from the old screen and start building the new watchlist under the updated weights. Existing positions get managed on their own merits. Forced selling because a regime label changed is a different kind of whipsaw.
What Regime Factor Screening Gets Wrong
This approach has real limitations. Acknowledging them matters more than pretending the system is complete.
First, regime classification is backward-looking. The 200-day SMA tells you where the market has been, not where it is going. By the time SPY closes below the 200-day for two consecutive weeks, the move is often 8-12% underway. You are not catching the turn. You are confirming it. That is acceptable for a screening tool, but it means you will hold momentum-heavy names into the first leg of a correction before switching to defensive weights.
Second, factor performance within regimes is probabilistic, not deterministic. Momentum tends to work in trend regimes, but it does not always work. September 2023 was a trend regime by my classification, and momentum names still pulled back sharply for three weeks. The weight map improves the odds. It does not guarantee them.
Third, the four-regime model is a simplification. Markets sometimes sit in transitional states that do not fit neatly into any quadrant. SPY hovering right at the 200-day SMA while volatility is average is an ambiguous signal. In those cases, I default to equal weights across all four factors and wait for clarity. Forcing a classification when the signal is weak leads to worse outcomes than admitting uncertainty.
Fourth, low-volatility factor data can be stale. A stock with low 90-day realized volatility might be about to report earnings or face a regulatory event that will spike its volatility overnight. The drawdown quality filter can help here by flagging names where recent price behavior has already started to deteriorate, even if the trailing volatility number has not caught up yet.
Connecting Regime Screening to Sector Rotation
Regime factor screening and sector rotation analysis are complementary but separate processes. Regime screening adjusts which factors you weight in a cross-sector screen. Sector rotation tells you which sectors are currently leading or lagging.
In practice, these two signals often align. During Regime 1 (trend + low volatility), momentum-heavy screens tend to surface names in whichever sector is leading the rally. During Regime 4 (defensive + high volatility), quality-heavy screens tilt toward utilities, healthcare, and consumer staples, which are the classic defensive sectors.
But they can also diverge. In late 2024, the broad market was in a trend regime while the leading sector (tech) was becoming increasingly concentrated in a handful of mega-caps. A momentum screen surfaced the same 5-6 names every week. Adding sector constraints to the screen, limiting any sector to 30% of the watchlist, produced a more diversified set of candidates without abandoning the momentum tilt. That is a practical constraint I apply regardless of regime.
Stop Treating Your Screen Like a Fixed Rule
Most traders spend weeks building a factor screen, validate it on historical data, and then freeze it. The screen becomes a rule rather than a tool. When it stops working, they blame the market instead of questioning whether the screen’s assumptions still match current conditions.
Regime factor screening does not require complex models or daily recalibration. It requires two things: a simple regime classification you check weekly, and a set of pre-built weight maps you swap based on the result. The inputs are a 200-day moving average and a Bollinger Band Width reading. The output is a factor composite that matches the environment instead of fighting it.
I have run this system for three years. It does not eliminate bad trades. It reduces the frequency of screening into the wrong factor at the wrong time, which is the most common failure mode for retail factor investors. The best screen in the world is the one that adapts when conditions change. A frozen screen is just a bet that the last six months will repeat.
Educational content only. Not investment advice. Trading involves risk. You are responsible for your decisions.