Research note
15 trading strategies, explained
Pairs, momentum, VWAP, sentiment, genetic algorithms and 10 more trading strategies explained, with what the research says and where each one tends to break.
Short answer: the 15 strategies in the popular table fall into two groups. Four of them (smart order routing, TWAP, VWAP and market-making) are mostly about how an order gets filled. The other eleven are ways to hunt for an edge, and the published evidence on them ranges from strong-but-fragile to weak.
This post walks through each one: what it does, what the research says, and where it tends to fail. It describes methods. It does not tell you to use any of them.
Which trading strategies are about execution, not prediction?
These four don't try to guess where price goes next. They try to buy or sell without paying too much for the privilege.
Smart order routing
US stock trading is spread across many venues: exchanges, ECNs, dark pools and broker-dealers that fill orders internally. The SEC's 2010 Concept Release on Equity Market Structure describes how brokers offer smart order routing to reach that scattered liquidity. It also describes algorithms that split a large "parent" order into many smaller "child" orders and route them over time to different venues. Brokers have a legal duty of best execution, and FINRA Rule 5310 sets out what that means for member firms.
Where it breaks: routing logic is only as good as its picture of where liquidity sits. Hidden liquidity in dark pools and internalizers is, by design, hard to see.
Market-making
A market maker posts both a bid and an offer and earns the spread when both sides fill. The same SEC release explains that much of this work is now done by proprietary trading firms. They submit large numbers of resting orders, often cancel a very high share of them, and collect liquidity rebates from venues that pay them.
The table says market-makers "often" use statistical arbitrage. That is true of some firms, but the core business is collecting the spread and managing inventory. The main risk is being picked off by better-informed traders when prices move fast.
TWAP (time-weighted average price)
TWAP slices an order evenly across a set window, for example the same number of shares every five minutes, no matter how much volume is trading. It is simple, predictable and easy to audit. The downside is that it ignores volume. It trades as much at a quiet lunch hour as at the busy open.
VWAP and Implementation Shortfall
VWAP is the average price of the day weighted by volume: dollar value traded divided by shares traded. StockCharts' ChartSchool notes that institutions first used it to place large orders without disrupting the market. A VWAP algorithm sizes its child orders to match the expected volume profile, so it trades more when the market is busy.
Implementation Shortfall goes back to André Perold's 1988 paper, which compares a "paper" portfolio with what you actually got after costs and delays. Almgren and Chriss later framed optimal execution as a trade-off. Trading fast raises market impact cost. Trading slowly leaves you exposed to price moves for longer. Their model builds an efficient frontier between the two.
In MRT, the Market Trend page (/market-trend) shows price against a daily anchored VWAP from the most recent major swing low, plus session and weekly VWAP when intraday data is available. That is a chart reference, not an execution algorithm.
Which strategies try to profit from price relationships?
Pairs trading
You take two instruments that have moved together. When the spread between them stretches, you go long the laggard and short the leader, betting the gap closes. Gatev, Goetzmann and Rouwenhorst tested a simple version on US stocks using daily data from 1962 to 2002. Matching stocks by the distance between their normalized prices, they found "average annualized excess returns of up to 11%" for self-financing portfolios of pairs, and profits that typically exceeded conservative cost estimates.
Where it breaks: a relationship that held for years can end for good, through a merger, a new business line or a sector re-rating. The short leg keeps losing while you wait for a reversion that never comes. MRT's Pairs Divergence Monitor (/pairs) tracks spread z-scores for correlated pairs and shows its own backtest on the page. Its verdict is that the strategy was not a return source on that test, and the page calls itself a divergence monitor, not a strategy.
Seasonality and calendar spreads
Seasonality looks for patterns tied to the calendar. The best-known stock example is "Sell in May and go away", which Bouman and Jacobsen studied in the American Economic Review in 2002 under the name the Halloween indicator.
A calendar spread is a futures trade. CME Group's education material defines it as buying a futures contract in one month while selling the same contract in a different month. The trader cares about the relationship between the two months, not the outright price, and hedgers also use calendar spreads to roll positions forward.
Where it breaks: calendar effects are easy to find in old data and hard to keep once they are widely known. MRT's Cycles page (/cycles) has a Seasonals panel with the Halloween and January Barometer readings. Both are labeled as failed pre-registered backtests and shown for research only.
Which strategies follow trends, volatility and chart patterns?
Momentum and RSI
Momentum bets that recent winners keep winning for a while. Jegadeesh and Titman (1993) found that buying past winners and selling past losers produced significant positive returns over 3- to 12-month holding periods, though part of the gain faded over the following two years. Moskowitz, Ooi and Pedersen (2012) studied a related idea, time series momentum, where an asset is judged against its own past return.
The Relative Strength Index is a momentum oscillator developed by J. Welles Wilder and published in his 1978 book. It runs from 0 to 100. Readings above 70 are usually called overbought and readings below 30 oversold.
Where it breaks: momentum crashes. Daniel and Moskowitz (2016) found these crashes tend to come in panic states, after market declines and when volatility is high, at the same time as sharp rebounds. MRT's Stock Screener (/screener) has RSI(14) minimum and maximum filters, and the Market Trend page reads the broad index through its moving-average stack.
Volatility trading
Volatility traders trade the size of moves, not their direction, usually through options, futures or VIX products. Cboe describes the VIX Index as a measure of the market's expectation of near-term volatility, taken from S&P 500 option prices. Cboe also notes that implied volatility has tended to trade above the volatility that later shows up, and that volatility tends to mean-revert over time.
The volatility breakout idea in the table runs the other way: after a quiet stretch, it waits for price to break out of its range. Where it breaks: quiet markets can stay quiet, and selling volatility can lose a lot very fast when a calm market turns. MRT's Vol and Options Surface page (/vol) covers implied-vol term structure, skew, dealer gamma exposure and implied moves.
Pattern recognition
Chart patterns include triangles, flags, head-and-shoulders and cup-and-handle. The obvious problem is that a pattern is often in the eye of the beholder. Lo, Mamaysky and Wang (2000) tried to fix that with an automatic, statistical way to detect patterns. They applied it to US stocks from 1962 to 1996 and found that several patterns "do provide incremental information and may have some practical value." That is a statement about information content, not proof of a profitable trading rule.
MRT's screener has rule-based scans for cup-and-handle setups, falling wedges and harmonic patterns. A Zanger swing-playbook card in the Terminal also tags bull flags, ascending triangles and pennants.
Which strategies use data science and machine learning?
Sentiment analysis
Sentiment strategies score news, social posts and other text for mood. Tetlock (2007) measured the tone of a daily Wall Street Journal column and found that high media pessimism predicted downward pressure on prices, followed by a reversion to fundamentals. Unusually high or low pessimism also predicted high trading volume.
Where it breaks: text is noisy, sarcasm is hard to score, and a widely watched sentiment signal gets crowded. MRT has a News board (/news), a CNN-style Fear and Greed gauge (/fear-greed) and a Social IQ digest (/social-iq) that includes a sentiment field.
Machine-learning strategies
ML models such as neural networks, random forests and support vector machines look for patterns in large datasets. Gu, Kelly and Xiu (2020) compared ML methods for estimating stock risk premiums. Trees and neural networks did best, mostly because they capture nonlinear interactions between predictors. The strongest signals were variations on momentum, liquidity and volatility.
Where it breaks: overfitting. Bailey, Borwein, López de Prado and Zhu (2014) explain how easy it is to tune a strategy until it looks good in-sample and then watch it fail out-of-sample. The more models you try, the more likely the best one is luck.
Deep reinforcement learning
Reinforcement learning treats trading as a control problem. An agent takes actions, gets rewards, and learns a policy from trial and error. Moody and Saffell (2001) used a method called recurrent reinforcement learning to optimize risk-adjusted returns while accounting for transaction costs. In their tests it beat a Q-learning system. Deep RL adds neural networks to the same idea.
Where it breaks: markets are not a game with fixed rules. An agent trained on one regime can fail when the regime changes, and the overfitting problem above applies with extra force.
Adaptive strategies
Adaptive strategies change their own parameters or rules as conditions change. The intellectual background is Andrew Lo's Adaptive Markets Hypothesis (2004). It applies evolutionary ideas (competition, adaptation and natural selection) to markets and argues that many behavioral biases are consistent with people adapting to a changing environment through simple heuristics. When the environment shifts, heuristics that suited the old one can stop working. MRT's Hidden Markov Model page (/hmm) fits a model to SPY returns and realized volatility and estimates the probability of each market regime, the kind of input an adaptive rule might use.
Where it breaks: every extra knob is another chance to overfit, and an adaptive rule can end up chasing the last regime.
Genetic algorithms
Genetic algorithms "breed" trading rules. They mutate and recombine candidate rules and keep the fittest. Allen and Karjalainen (1999) used one to learn technical rules for the S&P 500 with daily data from 1928 to 1995. After transaction costs, the rules did not earn consistent excess returns over buy-and-hold in the out-of-sample periods. The rules did tend to be in the market when returns were positive and volatility low, which the authors largely traced to low-order serial correlation in index returns.
What is scalping, and why is it risky?
Scalping means many small, short trades that each try to capture a tiny price move, often using order flow and tape reading. FINRA's guide to frequent intraday trading warns that higher costs can erode returns and that trading on margin can lose more than the money you deposited. It adds that strategies involving frequent trading on margin generally are not appropriate for investors with limited financial resources, limited trading experience or low risk tolerance.
What do these strategies have in common?
The same three problems keep showing up:
- Costs. Every strategy that trades often pays spreads, fees and market impact. A backtest that ignores them is not a backtest.
- Regime change. Pairs split, momentum crashes, calendar effects fade. A rule that worked in one era may not work in the next.
- Overfitting. The more rules, parameters or models you test, the more likely the winner is noise.
If you want to test a rule yourself, MRT's Strategy Lab (/strategy-lab) compares a strategy against buy-and-hold and a seeded shuffle-null, and labels the result in-sample. For our audits of individual viral claims, see the newsletter at /newsletter.
FAQ
What is the difference between TWAP and VWAP?
TWAP spreads an order evenly over time. VWAP sizes the slices to match the expected volume profile, so it trades more when the market is busy. Both are execution methods, not ways to predict price.
Does pairs trading still work?
Gatev, Goetzmann and Rouwenhorst found excess returns in US stocks from 1962 to 2002. That result is historical, and pairs can break apart for good when a business changes. MRT's own test, shown on /pairs, did not find a return edge.
Is momentum the same as RSI?
No. Momentum is a strategy that bets recent winners keep winning. RSI is one indicator that measures the speed and size of recent price changes on a 0 to 100 scale. Traders use RSI in momentum strategies and in mean-reversion strategies.
Can machine learning predict stock prices?
Gu, Kelly and Xiu (2020) found large economic gains from ML forecasts in their study, in some cases doubling the performance of leading regression-based strategies. The forecasts are noisy, and overfitting is a serious risk, so in-sample results say little about live performance.
Which of these strategies can a retail trader actually run?
Retail traders can study most of them on charts, but several depend on infrastructure most people don't have. Market-making and smart order routing need direct venue access and fast systems. Scalping depends on very low costs, and FINRA's guidance on frequent trading sets out the risks.
Sources
- Gatev, Goetzmann and Rouwenhorst, "Pairs Trading: Performance of a Relative-Value Arbitrage Rule", Review of Financial Studies 19(3), 2006: https://doi.org/10.1093/rfs/hhj020
- Jegadeesh and Titman, "Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency", Journal of Finance 48(1), 1993: https://doi.org/10.1111/j.1540-6261.1993.tb04702.x
- Moskowitz, Ooi and Pedersen, "Time series momentum", Journal of Financial Economics 104(2), 2012: https://doi.org/10.1016/j.jfineco.2011.11.003
- Daniel and Moskowitz, "Momentum crashes", Journal of Financial Economics 122(2), 2016: https://doi.org/10.1016/j.jfineco.2015.12.002
- StockCharts ChartSchool, "Relative Strength Index (RSI)", accessed 2026-10-17: https://chartschool.stockcharts.com/table-of-contents/technical-indicators-and-overlays/technical-indicators/relative-strength-index-rsi
- StockCharts ChartSchool, "Volume-Weighted Average Price (VWAP)", accessed 2026-10-17: https://chartschool.stockcharts.com/table-of-contents/technical-indicators-and-overlays/technical-overlays/volume-weighted-average-price-vwap
- Perold, "The Implementation Shortfall: Paper versus Reality", Journal of Portfolio Management 14(3), 1988: https://doi.org/10.3905/jpm.1988.409150
- Almgren and Chriss, "Optimal execution of portfolio transactions", Journal of Risk 3(2), 2000/2001: https://doi.org/10.21314/JOR.2001.041
- US Securities and Exchange Commission, "Concept Release on Equity Market Structure" (Release No. 34-61358), January 2010: https://www.sec.gov/rules/concept/2010/34-61358.pdf
- FINRA, "Rule 5310. Best Execution and Interpositioning", accessed 2026-10-17: https://www.finra.org/rules-guidance/rulebooks/finra-rules/5310
- FINRA, "Frequent Intraday Trading: Understanding the Basics", June 4, 2026: https://www.finra.org/investors/insights/frequent-intraday-trading
- CME Group, "Futures Spread Overview", accessed 2026-10-17: https://www.cmegroup.com/education/courses/understanding-futures-spreads/futures-spread-overview
- Bouman and Jacobsen, "The Halloween Indicator, Sell in May and Go Away: Another Puzzle", American Economic Review 92(5), 2002: https://doi.org/10.1257/000282802762024683
- Cboe, "Cboe Volatility Index (VIX)", accessed 2026-10-17: https://www.cboe.com/tradable-products/vix
- Lo, Mamaysky and Wang, "Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation", Journal of Finance 55(4), 2000: https://doi.org/10.1111/0022-1082.00265
- Tetlock, "Giving Content to Investor Sentiment: The Role of Media in the Stock Market", Journal of Finance 62(3), 2007: https://doi.org/10.1111/j.1540-6261.2007.01232.x
- Gu, Kelly and Xiu, "Empirical Asset Pricing via Machine Learning", Review of Financial Studies 33(5), 2020: https://doi.org/10.1093/rfs/hhaa009
- Bailey, Borwein, López de Prado and Zhu, "Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance", Notices of the AMS, May 2014: https://www.ams.org/journals/notices/201405/rnoti-p458.pdf
- Moody and Saffell, "Learning to trade via direct reinforcement", IEEE Transactions on Neural Networks 12(4), 2001: https://doi.org/10.1109/72.935097
- Lo, "The Adaptive Markets Hypothesis", Journal of Portfolio Management 30(5), 2004: https://doi.org/10.3905/jpm.2004.442611
- Allen and Karjalainen, "Using genetic algorithms to find technical trading rules", Journal of Financial Economics 51(2), 1999: https://doi.org/10.1016/S0304-405X(98)00052-X
- MacroMomentum Terminal, Pairs Divergence Monitor backtest verdict, accessed 2026-10-17: /pairs
- MacroMomentum Terminal, Cycles page, Seasonals panel, accessed 2026-10-17: /cycles
Educational only, not financial advice. Past and backtested results do not guarantee future returns. Do your own research or speak with a licensed professional before making investment decisions.