Skip to content

Momentum Investing

Version: 2026-09-24
Language: English
Purpose: To organize common momentum-investing mathematical models, academic strategies, risks, and ways to convert them into a reproducible MarketLens screening system.
Note: This document is for educational and research purposes only and does not constitute personal investment advice.


  1. What Is Momentum?
  2. Basic Mathematical Tools Before Calculating Momentum
  3. Model 1: 12-1 Price Momentum
  4. Model 2: 6-1 and MSCI Risk-Adjusted Momentum
  5. Model 3: Relative Strength / RS Percentile
  6. Model 4: Time-Series Momentum / Trend
  7. Model 5: 52-Week High Momentum
  8. Model 6: Earnings Momentum / SUE / PEAD
  9. Model 7: Industry / Sector Momentum
  10. Academic Strategy: Winners Minus Losers
  11. Why Can Momentum Work? What Is the Evidence?
  12. The Biggest Risk: Momentum Crash
  13. Why Is Momentum Attractive to Retail Investors?
  14. Risk-Control Mathematics
  15. Recommended MarketLens Architecture
  16. MarketLens Momentum Score: A Reproducible Version
  17. Metrics You Must Calculate in Backtesting
  18. Common Mistakes and Pitfalls
  19. Data Requirements and Implementation Priorities
  20. References and Official Methodologies

The simplest idea behind momentum is:

Assets that have performed relatively strongly in the recent past may have a higher tendency to remain relatively strong over certain short- to medium-term horizons; similarly, recent weak performers may continue to remain relatively weak.

However, “momentum” is not a single formula. It can be divided into several types:

  • Price Momentum: How much the asset itself has risen over a previous period.
  • Relative / Cross-Sectional Momentum: Which stocks are strongest relative to other stocks.
  • Risk-Adjusted Momentum: For similar returns, preference is given to stocks with lower volatility and smoother price trends.
  • Time-Series Momentum: Comparing an asset with its own history to determine whether its trend is rising or falling.
  • 52-Week High Momentum: How close the price is to its 52-week high.
  • Earnings Momentum: Whether price performance continues to drift after an earnings surprise.
  • Industry Momentum: Selecting strong industries first, then selecting strong stocks within those industries.

Therefore, a practical system should not treat “Momentum” as one single number.


2. Basic Mathematical Tools Before Calculating Momentum

Section titled “2. Basic Mathematical Tools Before Calculating Momentum”

If the price rises from (P_0) to (P_1):

R=P1P0−1R = \frac{P_1}{P_0}-1

For example, if the price rises from $100 to $120:

R=120100−1=20%R = \frac{120}{100}-1=20\%

If daily returns are (r_1,r_2,\ldots,r_n):

Rcum=∏i=1n(1+ri)−1R_{\text{cum}} = \prod_{i=1}^{n}(1+r_i)-1

If complete price data are available, this is equivalent to:

Rcum=PnP0−1R_{\text{cum}} = \frac{P_n}{P_0}-1
rtlog⁡=ln⁡(PtPt−1)r_t^{\log} = \ln\left(\frac{P_t}{P_{t-1}}\right)

Log returns are convenient for volatility estimation, statistical models, and time-series analysis, but simple returns are generally sufficient for momentum ranking.


Assume the standard deviation of daily returns is:

σdaily=SD(r1,…,rn)\sigma_{\text{daily}} = SD(r_1,\ldots,r_n)

Annualized volatility is commonly calculated as:

σannual=σdaily252\sigma_{\text{annual}} = \sigma_{\text{daily}}\sqrt{252}

If weekly returns are used:

σannual≈σweekly52\sigma_{\text{annual}} \approx \sigma_{\text{weekly}}\sqrt{52}

Before combining signals measured on different scales, they can be standardized:

Zi=xi−μσZ_i = \frac{x_i-\mu}{\sigma}

where:

  • (x_i): the value for stock (i)
  • (\mu): the mean across the entire universe
  • (\sigma): the standard deviation across the universe

For example, a Z-score of +2 means the value is approximately two standard deviations above the mean.


If MarketLens scans 7,000 stocks, a signal can be ranked and converted to a 0–100 scale:

Percentilei=100×ranki−1N−1Percentile_i = 100 \times \frac{rank_i-1}{N-1}

In implementation, make sure the ranking direction is consistent — i.e., whether a higher signal value always represents a better result.


12-1 Momentum means:

Measure approximately the previous 12 months of performance, but skip the most recent month.

Using monthly data:

MOM12−1=PT−1PT−13−1MOM_{12-1} = \frac{P_{T-1}}{P_{T-13}}-1

An approximate daily-data version is:

MOM12−1≈Pt−21Pt−252−1MOM_{12-1} \approx \frac{P_{t-21}}{P_{t-252}}-1

The actual number of trading days in a year varies slightly, so a program should ideally handle this by trading-day index or date offsets.


One common reason is to reduce the effect of very-short-term reversal and market microstructure noise.

In other words:

  • Medium term: momentum may exist.
  • Very short term: reversal may occur.

This is also why the Kenneth French Momentum factor uses prior (2–12) returns, excluding the most recent month.


12 months ago: $100
1 month ago: $140
Today: $150

MOM12−1=140100−1=40%MOM_{12-1} = \frac{140}{100}-1 = 40\%

The move from $140 to $150 during the most recent month is excluded from the 12-1 signal.


4. Model 2: 6-1 and MSCI Risk-Adjusted Momentum

Section titled “4. Model 2: 6-1 and MSCI Risk-Adjusted Momentum”

MSCI’s published Momentum methodology is a useful reproducible reference.

MOM6=(PT−1PT−7−1)−RfMOM_{6} = \left( \frac{P_{T-1}}{P_{T-7}}-1 \right) - R_f MOM12=(PT−1PT−13−1)−RfMOM_{12} = \left( \frac{P_{T-1}}{P_{T-13}}-1 \right) - R_f

where:

  • (P_{T-1}): price one month before the rebalance date
  • (P_{T-7}): price seven months before the rebalance date
  • (P_{T-13}): price thirteen months before the rebalance date
  • (R_f): local short-term risk-free return over the corresponding period

Both measures exclude the most recent month.


MSCI then divides Price Momentum by the stock’s volatility:

RAMi=MOMiσiRAM_i = \frac{MOM_i}{\sigma_i}

The MSCI methodology uses:

The annualized standard deviation of weekly local price returns over the previous three years.

Therefore:

RAM6,i=MOM6,iσiRAM_{6,i} = \frac{MOM_{6,i}}{\sigma_i} RAM12,i=MOM12,iσiRAM_{12,i} = \frac{MOM_{12,i}}{\sigma_i}

The intuition is:

If two stocks both rise 30%, but Stock A rises steadily while Stock B reaches the same return through large swings, Stock A has higher risk-adjusted momentum.


First convert each measure into a Z-score:

Z6,i=RAM6,i−μ6σ6Z_{6,i} = \frac{RAM_{6,i}-\mu_6}{\sigma_6} Z12,i=RAM12,i−μ12σ12Z_{12,i} = \frac{RAM_{12,i}-\mu_{12}}{\sigma_{12}}

Then combine them with equal weights:

Ci=0.5Z6,i+0.5Z12,iC_i = 0.5 Z_{6,i} + 0.5 Z_{12,i}

MSCI further standardizes the combined score and winsorizes it to approximately (\pm3).

The official Momentum Score transformation is:

Score={1+Z,Z>0(1−Z)−1,Z<0Score = \begin{cases} 1+Z, & Z>0\\ (1-Z)^{-1}, & Z<0 \end{cases}

4.5 Why Is This Model Suitable for MarketLens?

Section titled “4.5 Why Is This Model Suitable for MarketLens?”

Advantages:

  • Price history alone is sufficient.
  • Combining 6-month and 12-month horizons captures both intermediate and medium-term momentum.
  • Volatility adjustment avoids simply chasing the most erratic stocks.
  • Z-scores make cross-stock comparison easier.
  • It is straightforward to backtest.

5. Model 3: Relative Strength / RS Percentile

Section titled “5. Model 3: Relative Strength / RS Percentile”

Relative Strength here does not mean RSI (Relative Strength Index).

RS asks:

How strong is a stock relative to the rest of the market?


5.1 Most Transparent Version: 12-1 Percentile

Section titled “5.1 Most Transparent Version: 12-1 Percentile”

First calculate:

Mi=MOM12−1,iM_i=MOM_{12-1,i}

Then rank all stocks in the universe:

RSi=PercentileRank(Mi)RS_i = PercentileRank(M_i)

For example:

  • 7,000 stocks in the universe
  • ORCL ranks in the top 5% by momentum

This can be expressed approximately as:

RS≈95RS \approx 95

This method is particularly suitable for MarketLens because it is completely transparent.


5.2 Important: IBD RS Rating Is Not a Public Fixed Formula

Section titled “5.2 Important: IBD RS Rating Is Not a Public Fixed Formula”

The IBD Relative Strength Rating is a proprietary rating.

Therefore, a set of weights found online should not be presented as the exact official IBD formula.

If MarketLens implements a similar concept, it is better to use a distinct name such as:

ML Relative Strength Percentile

rather than claiming to reproduce IBD RS.


MarketLens could define a transparent custom model:

RSraw=0.40R3M+0.20R6M+0.20R9M+0.20R12MRS_{\text{raw}} = 0.40R_{3M} + 0.20R_{6M} + 0.20R_{9M} + 0.20R_{12M}

Then:

RSpercentile=PercentileRank(RSraw)RS_{\text{percentile}} = PercentileRank(RS_{\text{raw}})

But note:

This is a custom transparent model, not a claim to reproduce the proprietary IBD rating.


Cross-sectional momentum asks:

“Which is stronger when Stock A is compared with Stocks B, C, and D?”

Time-series momentum asks:

“Compared with its own history, is Stock A currently in an uptrend or a downtrend?”


TSMOMt=sign(Rt−k,t)TSMOM_t = sign(R_{t-k,t})

If:

R12M>0R_{12M}>0

Signal = +1.

If:

R12M<0R_{12M}<0

Signal = -1.

Moskowitz, Ooi & Pedersen (2012) studied multiple futures markets and found time-series momentum characteristics in returns over the previous 1–12 months.


A common quantitative expression is:

Positiont=sign(R12M)×σtargetσassetPosition_t = sign(R_{12M}) \times \frac{\sigma_{\text{target}}} {\sigma_{\text{asset}}}

Meaning:

  • Uptrend → long direction
  • Downtrend → short / avoid
  • Higher volatility → reduce position size
  • Lower volatility → position size may be increased

For a long-only MarketLens system, a negative signal can simply mean “not eligible” rather than requiring a short position.


A practical and intuitive trend filter is:

Pt>SMA200tP_t>SMA200_t

Then add:

SMA200t>SMA200t−20SMA200_t>SMA200_{t-20}

or:

SMA200t>SMA200t−30SMA200_t>SMA200_{t-30}

The second condition helps confirm that the SMA200 itself is rising, rather than the price having only temporarily crossed above the moving average.


63 trading days is approximately three months:

ROC63=PtPt−63−1ROC63 = \frac{P_t}{P_{t-63}}-1

ROC63 is a shorter-term momentum signal.

It is not the same as 12-1 momentum because the two use different time horizons.


George & Hwang (2004) studied the information contained in a stock’s price relative to its 52-week high.

The simplest measure is:

HighRatioi=Pimax⁡(Pi,t−251:t)HighRatio_i = \frac{P_i} {\max(P_{i,t-251:t})}

The range is usually:

0<HighRatio≤10<HighRatio\le1

For example:

  • Current price = $145
  • 52-week high = $150
HighRatio=145150=0.9667HighRatio = \frac{145}{150} = 0.9667

So the current price is 96.67% of the 52-week high.


It can also be expressed as:

Distance52W=PtHigh52W−1Distance52W = \frac{P_t}{High_{52W}}-1

Using the same example:

Distance52W=−3.33%Distance52W = -3.33\%

This means the stock is 3.33% below its 52-week high.


For example:

HighRatio >= 0.90

This means the stock is trading at least at 90% of its 52-week high.

However, this condition should not be used alone as a buy signal. It is better combined with:

  • liquidity
  • trend
  • RS
  • volatility
  • VCP / consolidation
  • market regime

8. Model 6: Earnings Momentum / SUE / PEAD

Section titled “8. Model 6: Earnings Momentum / SUE / PEAD”

This is not pure price momentum. It is a form of fundamental momentum.

The simplest form is:

UEt=EPSt−E(EPSt)UE_t = EPS_t-E(EPS_t)
SUEt=UEtσ(UE)SUE_t = \frac{UE_t} {\sigma(UE)}

where:

  • (EPS_t): actual EPS
  • (E(EPS_t)): expected EPS
  • (\sigma(UE)): historical volatility of earnings surprises

A traditional earnings expectation can be written as:

E(EPSt)≈EPSt−4+δE(EPS_t) \approx EPS_{t-4} + \delta

Therefore:

UEt=EPSt−EPSt−4−δUE_t = EPS_t-EPS_{t-4}-\delta

Then standardize it:

SUEt=EPSt−EPSt−4−δσ(UE)SUE_t = \frac{EPS_t-EPS_{t-4}-\delta} {\sigma(UE)}

PEAD = Post-Earnings-Announcement Drift.

The idea is:

After an earnings surprise, the market price may not immediately incorporate all of the information. The price may continue to drift in the direction of the surprise for a period of time.

If MarketLens later adds earnings and fundamental data, it could include:

Positive SUE
AND price momentum positive
AND RS high

This creates:

Price Momentum + Earnings Momentum


Moskowitz & Grinblatt (1999) found that the industry component is important in explaining stock momentum.


Assume sector (s) contains (N_s) stocks.

Equal-weighted return:

Rsector,s=1Ns∑i=1NsRiR_{sector,s} = \frac{1}{N_s} \sum_{i=1}^{N_s}R_i

It can also be market-cap weighted:

Rsector,s=∑iwiRiR_{sector,s} = \sum_i w_iR_i

where:

∑iwi=1\sum_iw_i=1

Rank all sectors:

SectorRSs=PercentileRank(Rsector,s)SectorRS_s = PercentileRank(R_{sector,s})

Then:

Select Top Sectors
↓
Select Top Stocks within those sectors

That is:

Market→Sector→StockMarket \rightarrow Sector \rightarrow Stock

10. Academic Strategy: Winners Minus Losers

Section titled “10. Academic Strategy: Winners Minus Losers”

The classic momentum concept in Jegadeesh & Titman (1993) is:

  1. Rank stocks by past performance.
  2. Buy the strongest group.
  3. Short the weakest group.
  4. Hold for a short- to medium-term period.

The Kenneth French Data Library momentum factor uses prior 2–12 month returns.

Stocks are classified by:

  • Size: Small / Big
  • Momentum: Loser / Neutral / Winner

This creates six portfolios:

  • SL
  • SN
  • SW
  • BL
  • BN
  • BW

The Momentum factor is:

MOM=12(SW+BW)−12(SL+BL)MOM = \frac{1}{2}(SW+BW) - \frac{1}{2}(SL+BL)

In other words:

Winners Minus Losers.


A retail investor does not have to short losers.

A simplified approach could be:

Universe
↓
Liquidity filter
↓
Momentum rank
↓
Keep only Top 10% / 20%
↓
Trend filter
↓
Risk filter
↓
Hold a basket of the strongest stocks

This is not the same strategy as the academic long-short portfolio, so its backtest results should not be treated as equivalent to the academic momentum premium.


11. Why Can Momentum Work? What Is the Evidence?

Section titled “11. Why Can Momentum Work? What Is the Evidence?”

Several important studies include:

They found that strategies buying past winners and selling past losers generated significant positive returns across 3–12 month holding horizons.

They found that industry momentum can explain a substantial portion of individual-stock momentum.

They found that proximity to the 52-week high contains additional information about future returns.

They found time-series momentum across equity-index, currency, commodity, and bond futures.

They found common structures in value and momentum premia across multiple markets and asset classes.


Research by Daniel & Moskowitz on Momentum Crashes shows that momentum strategies can occasionally experience rare but extremely severe negative returns.

These crashes are particularly likely when:

  1. The market has previously fallen sharply.
  2. Volatility is high.
  3. Past losers have suffered very large declines.
  4. The market suddenly rebounds strongly.

Assume a momentum portfolio is:

Long Winners
Short Losers

After a market panic:

  • Winners: more defensive and have fallen less.
  • Losers: may already have fallen 50–80%.

If the market suddenly rebounds:

  • Winners +3%
  • Losers +30%

The long-short return becomes:

RMOM=3%−30%=−27%R_{MOM} = 3\%-30\% = -27\%

Therefore:

The most dangerous period for momentum is not necessarily the crash itself; it can be the extremely sharp rebound following a major decline.


13. Why Is Momentum Attractive to Retail Investors?

Section titled “13. Why Is Momentum Attractive to Retail Investors?”

The main idea in this section is to turn the constraints faced by retail investors into a rules-based system.

13.1 Less Time Needed for Large-Scale Financial-Statement and Valuation Analysis

Section titled “13.1 Less Time Needed for Large-Scale Financial-Statement and Valuation Analysis”

Momentum can be calculated mainly from:

  • adjusted price
  • return
  • volatility
  • volume
  • sector
  • benchmark

Therefore, a basic version can be built using market data.

However:

“Not requiring valuation” does not mean valuation is useless. It only means a momentum model does not necessarily require valuation inputs.


13.2 No Need to Watch the Market Every Day

Section titled “13.2 No Need to Watch the Market Every Day”

Momentum is generally a medium- to low-frequency strategy.

For example:

  • Weekly scan
  • Monthly rebalance
  • Quarterly rebalance

MSCI Momentum Indexes moved to quarterly reviews after the 2025 methodology update and also introduced turnover-mitigation measures.

The key point is:

Strategy frequency should be determined by backtesting, not by assuming that “more frequent is always better.”


13.3 No Need to Predict Which Stock Will “Explode” Next

Section titled “13.3 No Need to Predict Which Stock Will “Explode” Next”

Momentum is more of a reaction model than a prediction model:

Do not guess which stock will suddenly become strong
→
Observe which stocks are already demonstrating relative strength
→
Only include them when predefined rules are satisfied

This can reduce subjective, narrative-driven stock selection.


Avoid:

All-in strongest stock

Instead use:

Top-ranked basket
+
liquidity filter
+
trend filter
+
market regime
+
position sizing
+
exit rules

Rules can be defined in advance:

IF entry conditions = true:
eligible = true
ELSE:
eligible = false

Potential benefits include reducing:

  • FOMO
  • reluctance to execute a stop
  • changing the strategy impulsively after a decline
  • chasing prices upward by changing the rules

However, the rules themselves must first be reasonably backtested. Mechanically following a bad rule can still lose money.


Assume:

  • Portfolio = $100,000
  • Maximum risk per trade = 0.5%
  • Entry = $100
  • Stop = $95

Maximum loss budget per trade:

RiskBudget=100000×0.005=$500RiskBudget = 100000\times0.005 = \$500

Risk per share:

RiskPerShare=100−95=$5RiskPerShare = 100-95 = \$5

Number of shares:

Shares=5005=100Shares = \frac{500}{5} = 100

In general:

PositionSize=PortfolioValue×Risk%∣Entry−Stop∣PositionSize = \frac{PortfolioValue\times Risk\%} {|Entry-Stop|}

This controls individual-trade risk more directly than simply investing a fixed amount such as $5,000 in every stock.


If a fixed stop is not used, position weights can be based on volatility:

Weighti∝1σiWeight_i \propto \frac{1}{\sigma_i}

Meaning:

  • Higher volatility → lower weight
  • Lower volatility → higher weight

Then normalize:

wi=1/σi∑j1/σjw_i = \frac{1/\sigma_i} {\sum_j1/\sigma_j}

True Range:

TRt=max⁡{Ht−Lt∣Ht−Ct−1∣∣Lt−Ct−1∣TR_t = \max \begin{cases} H_t-L_t\\ |H_t-C_{t-1}|\\ |L_t-C_{t-1}| \end{cases}

ATR:

ATRn=MAn(TR)ATR_n = MA_n(TR)

For example:

Stop=Entry−2×ATR14Stop = Entry-2\times ATR_{14}

An ATR-based stop adjusts the stop distance according to the stock’s own volatility.


A stop-loss order does not guarantee execution at the stop price.

If a stock gaps down:

  • Stop = $95
  • Next available execution price = $89

The actual trade may be executed close to $89.

Therefore, position sizing still needs to account for gap risk.


I recommend evolving MarketLens from “one momentum formula” into a multi-layer architecture.

┌──────────────────────────┐
│ 1. Universe / Liquidity │
└────────────┬─────────────┘
↓
┌──────────────────────────┐
│ 2. Market Regime │
│ SPY trend / volatility │
└────────────┬─────────────┘
↓
┌──────────────────────────┐
│ 3. Sector Momentum │
└────────────┬─────────────┘
↓
┌──────────────────────────┐
│ 4. Stock Momentum │
│ 12-1 / 6-1 / RS / ROC63 │
└────────────┬─────────────┘
↓
┌──────────────────────────┐
│ 5. Trend Quality │
│ SMA200 / 52W High / VCP │
└────────────┬─────────────┘
↓
┌──────────────────────────┐
│ 6. Risk Control │
│ vol / sizing / stops │
└────────────┬─────────────┘
↓
Candidate List

16. MarketLens Momentum Score: A Reproducible Version

Section titled “16. MarketLens Momentum Score: A Reproducible Version”

The following is not an official formula from any institution. It is a transparent MarketLens baseline that can be backtested and modified.

For each stock:

S1=MOM12−1S_1=MOM_{12-1} S2=RAM6S_2=RAM_{6} S3=RAM12S_3=RAM_{12} S4=ROC63S_4=ROC63 S5=PHigh52WS_5=\frac{P}{High_{52W}}

Convert all signals to a 0–100 scale:

Qk=PercentileRank(Sk)Q_k = PercentileRank(S_k)

16.3 Step 3: Build the Momentum Core Score

Section titled “16.3 Step 3: Build the Momentum Core Score”

A possible starting model is:

CoreMomentum=0.30Q12−1+0.15QRAM6+0.20QRAM12+0.15QROC63+0.20Q52WCoreMomentum = 0.30Q_{12-1} + 0.15Q_{RAM6} + 0.20Q_{RAM12} + 0.15Q_{ROC63} + 0.20Q_{52W}

The weights sum to:

0.30+0.15+0.20+0.15+0.20=10.30+0.15+0.20+0.15+0.20=1

These weights are not “optimal parameters.”

They must be validated using out-of-sample backtesting.


For example:

TrendGate={1,P>SMA200 and SMA200t>SMA200t−200,otherwiseTrendGate= \begin{cases} 1, & P>SMA200 \text{ and } SMA200_t>SMA200_{t-20}\\ 0, & \text{otherwise} \end{cases}

The simplest version is:

FinalScore=CoreMomentum×TrendGateFinalScore = CoreMomentum\times TrendGate

However, a binary gate can be too rigid.

Another option is a soft trend score:

ConditionTrend Points
Price > SMA200+40
SMA200 rising+30
Price > SMA50+20
SMA50 > SMA200+10

Total score: 0–100.


Using SPY as an example:

MarketTrend={1,SPY>SMA2000,SPY≤SMA200MarketTrend= \begin{cases} 1,& SPY>SMA200\\ 0,& SPY\le SMA200 \end{cases}

A stricter version is:

BullRegime=1ifSPY>SMA200∧SMA200t>SMA200t−20BullRegime= 1 \quad\text{if}\quad SPY>SMA200 \land SMA200_t>SMA200_{t-20}

During a risk-off regime, MarketLens could:

  • reduce the number of candidates;
  • increase the selection threshold;
  • reduce total exposure;
  • rather than simply switching the whole system completely on or off.

This should be determined by backtesting.


17. Metrics You Must Calculate in Backtesting

Section titled “17. Metrics You Must Calculate in Backtesting”

Do not look only at “win rate.”

CAGR=(VendVstart)1/n−1CAGR = \left( \frac{V_{end}}{V_{start}} \right)^{1/n} -1

First define the running peak:

Peakt=max⁡s≤tVsPeak_t = \max_{s\le t}V_s

Drawdown:

DDt=Vt−PeaktPeaktDD_t = \frac{V_t-Peak_t}{Peak_t}

Maximum Drawdown:

MDD=min⁡tDDtMDD = \min_t DD_t
Sharpe=E[Rp−Rf]σ(Rp−Rf)Sharpe = \frac{E[R_p-R_f]} {\sigma(R_p-R_f)}

If annualizing daily returns:

Sharpeannual≈252×rˉsrSharpe_{annual} \approx \sqrt{252} \times \frac{\bar r} {s_r}

Using only downside deviation:

Sortino=E[Rp−Rf]σdownsideSortino = \frac{E[R_p-R_f]} {\sigma_{downside}}

A simplified measure is:

Turnovert=12∑i∣wi,t−wi,t−1∣Turnover_t = \frac{1}{2} \sum_i |w_{i,t}-w_{i,t-1}|

If a momentum strategy has excessive turnover, transaction costs can consume part of the strategy’s edge.


At minimum, compare against:

  • SPY
  • Equal-weight universe
  • Simple 12-1 momentum
  • MarketLens composite model

Otherwise, you cannot tell whether the more complex model is actually better than a simple baseline.


You must not use information that only became known today as though it had already been available yesterday.

For example:

Announcement date = 15 May

You cannot use EPS announced on 15 May in a signal dated 14 May.


If you download only the 7,000 stocks that exist today and then backtest ten years into the past:

Companies that were delisted or went bankrupt will disappear from the historical universe.

This can make the strategy look better than it really was.


You must handle:

  • splits
  • reverse splits
  • dividends
  • mergers
  • symbol changes

Price momentum should preferably use a consistent adjusted-price convention.


Suppose you test:

  • ROC 61
  • ROC 62
  • ROC 63
  • ROC 64
  • ROC 65

If ROC63 performs best, that does not mean 63 has some special magical property.

A better question is:

Does the entire 60–66 day parameter neighborhood produce reasonably similar results?


The following signals are highly correlated:

  • 12-1 return
  • 12M RS
  • 52W High
  • ROC
  • SMA trend

Therefore, having five signals does not mean you have five independent pieces of evidence.

A composite score should account for signal correlation.


A backtest should include:

NetReturn=GrossReturn−Commission−Spread−SlippageNetReturn = GrossReturn - Commission - Spread - Slippage

Momentum may have worked on average historically:

≠\neq

Momentum works in every market environment.

This is one of the most important risk concepts in the entire system.


19. Data Requirements and Implementation Priorities

Section titled “19. Data Requirements and Implementation Priorities”

Your existing Massive / market-price data can already support:

  • ROC63
  • SMA200
  • SMA200 rising
  • 12-1 Momentum
  • 6-1 Momentum
  • 52-week High Ratio
  • RS Percentile
  • Risk-adjusted Momentum
  • SPY Market Regime

This is the most valuable first phase to implement.


You need a complete US-stock universe with:

  • ticker
  • adjusted price history
  • liquidity
  • preferably market cap
  • sector / industry classification

Then add:

  • Sector Momentum
  • Universe ranking
  • Top-decile / top-quintile portfolio tests

Then add:

  • quarterly EPS
  • analyst estimates
  • earnings announcement dates

This enables:

  • SUE
  • Earnings Momentum
  • PEAD-related filters

Jegadeesh, N. & Titman, S. (1993)
Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency
The Journal of Finance, 48, 65–91.
https://doi.org/10.1111/j.1540-6261.1993.tb04702.x


Kenneth R. French Data Library — Momentum Factor
Monthly MOM construction uses prior (2–12) returns and winner-minus-loser portfolios.
https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/det_mom_factor.html

Developed Markets Momentum methodology:
https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/f-f_developed_mom.html


MSCI Momentum Indexes Methodology
Core methodology includes 6-month and 12-month price momentum, a one-month skip, risk-free-rate adjustment, three-year weekly volatility, Z-score standardization, and momentum weighting.
https://www.msci.com/indexes/documents/methodology/2_MSCI_Momentum_Indexes_Methodology_20250417.pdf

MSCI announced in 2025 that Momentum Index reviews would move to quarterly frequency with partial-turnover implementation from the August 2025 review:
https://app2.msci.com/webapp/index_ann/DocGet?pub_key=kiUSObVqj%2Fs%3D

MSCI announced a further methodology update on 19 August 2026 clarifying the treatment of securities with missing momentum scores:
https://app2.msci.com/webapp/index_ann/DocGet?format=html&lang=en&pub_key=FWS5q6euxPk%3D


Moskowitz, T. J. & Grinblatt, M. (1999)
Do Industries Explain Momentum?
The Journal of Finance, 54, 1249–1290.
https://doi.org/10.1111/0022-1082.00146


George, T. J. & Hwang, C.-Y. (2004)
The 52-Week High and Momentum Investing
The Journal of Finance, 59, 2145–2176.
https://doi.org/10.1111/j.1540-6261.2004.00695.x


Moskowitz, T. J., Ooi, Y. H. & Pedersen, L. H. (2012)
Time Series Momentum
Journal of Financial Economics, 104, 228–250.
https://doi.org/10.1016/j.jfineco.2011.11.003


Asness, C. S., Moskowitz, T. J. & Pedersen, L. H. (2013)
Value and Momentum Everywhere
The Journal of Finance, 68, 929–985.
https://doi.org/10.1111/jofi.12021


Bernard, V. L. & Thomas, J. K. (1989)
Post-Earnings-Announcement Drift: Delayed Price Response or Risk Premium?
Journal of Accounting Research, 27, 1–36.
https://doi.org/10.2307/2491062


Daniel, K. & Moskowitz, T. J. (2016)
Momentum Crashes
Journal of Financial Economics, 122, 221–247.
https://doi.org/10.1016/j.jfineco.2015.12.002


If the entire approach is condensed into one sentence:

Momentum is not simply “chasing stocks that are going up.” It is a systematic approach that uses fixed lookback periods, relative ranking, trend, volatility, diversification, and risk control to judge whether existing strength is worth continuing to follow.

For MarketLens, I would divide the model into three layers:

Market Regime→Sector Momentum→Stock Momentum\boxed{ Market\ Regime \rightarrow Sector\ Momentum \rightarrow Stock\ Momentum }

Instead of relying only on ROC63, Stock Momentum can consider:

12 ⁣− ⁣1+6 ⁣− ⁣1+RiskAdjusted+RS+52WHigh+Trend\boxed{ 12\!-\!1 + 6\!-\!1 + RiskAdjusted + RS + 52WHigh + Trend }

Then use:

Position Sizing+Diversification+Exit Rules\boxed{ Position\ Sizing + Diversification + Exit\ Rules }

for risk control.

The most important research principle is:

Build a simple baseline first, then add signals one by one. Every added signal should demonstrate that it improves risk-adjusted performance out of sample, rather than merely making the model look more sophisticated.