Assumptions of Quantitative Investing

“Market movements are unpredictable” is not an absolute “yes” or “no” consensus in the field of quantitative investing, but rather a “basic assumption” with strong nuances.

Simply put, the consensus in quantitative investing is that individual, specific market movements (Point Prediction) are nearly impossible to predict, but through the law of large numbers, the probabilistic characteristics of market price distributions (Probabilistic Edge) can be predicted and profitably exploited.

Below is a detailed analysis of the disagreements and consensus within quantitative investing regarding “market predictability”:

1. Core consensus: Acknowledging that “most” movements are unpredictable (Efficient Market Hypothesis)

Quantitative investors generally accept a practical version of the Weak Form Efficient Market Hypothesis (Weak EMH):

• Extremely low signal-to-noise ratio: Of the fluctuations in market prices, 99% or even more is “noise” (random walk), with only a very small portion containing useful information (signals).

• No crystal ball: Almost no serious quantitative institution would attempt to make precise point predictions such as “What will the S&P 500 level be next month?”

• Rejecting “market timing”: The vast majority of fundamental quantitative strategies (such as AQR) or statistical arbitrage strategies do not rely on predicting the overall rise or fall of the market to make money. Instead, they rely on long-short hedging (Long-Short) to strip out market risk (Beta), earning only the returns from the relative strength of individual stocks (Alpha).

Supporting quotation:
Jim Simons, the “king of quant” and founder of Renaissance Technologies, once famously said: “One can predict the course of a comet more easily than one can predict the course of Citigroup's stock.” This directly captures the extreme difficulty of predicting individual assets.

2. Practical revision: Pursuing a “small statistical edge”

Although they acknowledge that “unpredictability” is the norm, quantitative investing itself is based on the premise that “the market is not completely random.” Their “prediction” is fundamentally different from the “prediction” understood by retail investors:

• A game of win rates: The well-known quantitative fund WorldQuant has disclosed that the average win rate of its models may be only 51%–53%. This means they acknowledge that they are “wrong” nearly half the time, but as long as the number of trades is large enough (the law of large numbers), this slight edge can be converted into enormous profits.

3. Differences among schools of thought

Different quantitative strategies rely on predictability to different degrees:

• High-frequency trading (HFT): They barely predict future prices; instead, they predict market microstructure and order flow (Order Flow). For example, they exploit millisecond-level latency differences between exchanges to conduct arbitrage. For them, this is not “betting on up or down,” but “jumping the queue” through physical speed.

• Statistical arbitrage (Stat Arb): They believe in mean reversion. That is: I do not predict whether a stock will rise tomorrow, but I predict that “if stock A deviates too far from its historical relationship with stock B, they will eventually converge.” This is a prediction of the spread (Spread), rather than a simple prediction of whether prices will rise or fall.

• Trend following (CTA): They completely abandon prediction. The core logic of a CTA strategy is: “I do not predict when a trend will start or end. I simply follow it when it appears, until it reverses.” They profit from irrational market exuberance or panic, rather than from predicting the future.

4. Expert views: Is the market becoming “harder to predict”?

• Cliff Asness (founder of AQR): He is a supporter of efficient-market theory, but has also pointed out that due to behaviors such as retail-investor crowding into Meme stocks, the market may become “less efficient” in the short term. However, this does not mean it becomes easier to predict; instead, it means more noise and greater risk.

• Man Group: It clearly points out that, because the future cannot be known in advance, the core of quantitative strategies must be risk control (Risk Control) and diversification (Diversification), rather than betting on a single direction.

Summary

The real meaning of “market movements are unpredictable” in quantitative circles is:

Do not try to play God and foresee the future. Instead, be like a casino house, using mathematical models to find small opportunities whose win rates are slightly higher than 50%, and strictly control risk so that probability is on your side.

For individual investors, this is indeed a consensus warning: If even quantitative funds equipped with top mathematicians and supercomputers only dare to pursue a 51% win rate, then ordinary investors trying to get rich by “predicting market movements” are pursuing an impossible goal.

One step I can take for you:
If you are interested in this way of thinking, I can explain in detail how the “Kelly Criterion” is used in quantitative investing to decide “since predictions are inaccurate, how much should be bet each time.”

This is a very profound question. The answer is: Yes, absolutely, and this is precisely the "Holy Grail" that multi-strategy hedge funds pursue.

In quantitative investing, not only can different strategies rely on conflicting assumptions, but excellent portfolio managers will deliberately seek out strategies with mutually exclusive assumptions. This is because it brings low correlation or even negative correlation, which greatly improves the portfolio's Sharpe Ratio.

Here are several classic examples of "mutually exclusive assumptions" and why they can coexist:

1. Momentum vs. Mean Reversion

This is the most classic pair of "archrivals," with completely opposite underlying logic:

• Momentum Strategy Assumption: "The strong get stronger, the weak get weaker." If an asset's price has been rising over the past period, assume its inertia will cause it to continue rising.

• Behavioral Logic: Buy high, sell higher.

• Mean Reversion Strategy Assumption: "What goes up must come down." If an asset's price deviates from its mean (too high or too low), assume it will be pulled back to the mean by gravity.

• Behavioral Logic: Buy low, sell high (buy the dip, sell the top).

Conflict Point: When a stock rises rapidly, the momentum model might issue a "buy" signal, while the mean reversion model might issue a "sell short" signal. Why they coexist:

1. Different Time Horizons: Mean reversion is often effective over very short periods (minutes/hours) or very long periods (3-5 years); momentum is usually strongest over medium-term periods (3-12 months). A quant fund can do mean reversion on a minute-by-minute basis and momentum on a monthly basis without interference.

2. Risk Hedging: When the market has a strong trend, the mean reversion strategy will lose money, and the momentum strategy will make a lot; when the market is range-bound, the momentum strategy will get "whipsawed," and the mean reversion strategy will print money. Combining them smooths out the equity curve.

2. Value vs. Growth/Momentum

In fundamental quant, this contradiction is particularly evident:

• Value Factor Assumption: The market makes mistakes and undervalues bad stocks. So, buy cheap stocks (low P/E, low P/B), which are usually recent underperformers.

• Growth/Momentum Factor Assumption: The market is efficient; expensive stocks are expensive for a reason (they are good). So, buy strong performers, which usually have high valuations.

Conflict Point: The value factor tells you to buy "gold in the trash heap," while the growth factor tells you to buy "the star surrounded by admirers." Why they coexist: This is a perfect complement. Giants like AQR have found that value and momentum factors historically exhibit a slight negative correlation. Combining them can significantly reduce volatility without reducing expected returns.

3. Short Volatility vs. Long Volatility/CTA

• Short Volatility Assumption: The market is calm most of the time; panic is temporary. Therefore, selling options (insurance) allows you to collect premium (Carry).

• Mindset: "Peace and quiet."

• Trend Following (CTA) Assumption: The market will have extreme black swan events, and volatility clusters. Therefore, hold options or follow trends, waiting for big volatility to erupt.

• Mindset: "Let the world burn."

Conflict Point: One bets the market will be calm; the other bets it will be chaotic. Why they coexist: Short volatility is like "picking up pennies in front of a steamroller" – high win rate but small gains per trade (with risk of ruin); CTA is like "buying lottery tickets" – small regular losses, but when a crisis hits (e.g., 2008, 2020 COVID), CTA makes huge gains, perfectly offsetting losses from other strategies.

Summary: The Quant's "Blind Men and the Elephant"

If the market is an elephant, different quant strategies are the blind men touching it:

• The one touching the leg says: "The market is like a pillar, stable (mean reversion)."

• The one touching the trunk says: "The market is like a snake, moving flexibly (momentum)."

A true quant master (Portfolio Manager) doesn't argue about who is right, because he knows everyone is only seeing a part. He hires both groups simultaneously, ensuring he makes money no matter which part of the elephant the market touches.

Conclusion: Conflicting assumptions are not only allowed but are necessary for building a high Sharpe ratio portfolio. If all strategies had the same assumptions, and that assumption was wrong, the entire fund would go bankrupt.

Your intuition is very sharp! If two strategies operate on the exact same stock, at the exact same time, in completely opposite directions, and for the exact same holding period, then indeed: one must lose money (ignoring fees, it's even a zero-sum game).

However, in the actual operation of a quant fund, the reason they can both be profitable (or at least profitable overall) is due to the following three key "misalignment" mechanisms:

1. Time Horizon Misalignment (Most Common Case)

This is the biggest magic that turns conflicting strategies from "enemies" into "friends." Market trends are not a straight line; they are like a winding mountain road.

• Scenario: A stock is rising from $100 to $110, with a brief pullback to $105 along the way.

• Trend Strategy (Long): Looks at the big picture. It buys at $100 and holds until $110. (Makes $10)

• Mean Reversion Strategy (Short): Looks at short-term fluctuations. When the stock surges too fast to $108, it predicts a pullback, so it "shorts" and buys back to cover at $105. (Makes $3)

Result: In the same upward move, the long trader makes money on the big trend, and the short trader makes money on the small intermediate pullback. Although their directions are opposite, because their holding periods are different, they both profit.

2. Internal Crossing Mechanism

What happens if two strategies really want to do the opposite at the exact same millisecond – one wants to buy 100 shares of Moutai, the other wants to sell 100 shares?

• Retail Investor View: Account A makes money, account B loses money, the fund manager did nothing.

• Fund View: The trading system detects this and executes an internal crossing. The fund doesn't have to trade a single share on the open market; it just transfers the shares on its books.

• Result: In this case, no "trading profit" is generated, but huge transaction fees and market impact costs are saved. In quant, saving money is making money.

Only when Strategy A wants to buy 500 shares and Strategy B only wants to sell 100 shares does the fund go to the market to buy the net 400 shares (Net Exposure). This mechanism ensures that conflicts don't cause a double loss, only reduce risk.

3. Independence of Probabilistic Advantage (When one fails, the other succeeds)

A strategy's profitability is not judged by each individual trade, but by its annual expected value.

• First Half of the Year: The market is a strong bull market.

• Momentum Strategy (Chasing trends): Big profits.

• Mean Reversion Strategy (Buying dips): Small losses (because it keeps selling too early or getting stopped out while shorting against the trend).

• Total Result: Profitable.

• Second Half of the Year: The market becomes a "monkey market" (violent oscillations, no clear direction).

• Momentum Strategy: Drawdown (buys right before a drop, sells right before a rally, gets whipsawed).

• Mean Reversion Strategy: Huge profits (buying low and selling high works perfectly).

• Total Result: Profitable.

Key Point: Since we can't predict whether the year will be a "bull market" or a "range-bound market," we just use both strategies. As long as they make a lot when they are right and lose a little when they are wrong (cutting losses), the combined annual curve will be very smooth.

Summary

You are right that in every specific conflicting moment, one must be right and the other wrong.

But quantitative investing is a game of combination. It's like a football team:

• When the forward (offensive strategy) scores, the defender (defensive strategy) might seem useless.

• When the defender (defensive strategy) makes a save, the forward (offensive strategy) hasn't scored either.

But as long as they each play their role in their respective "time segments" of the game, the team (fund) can win the match.

Therefore, what a quant fund fears most is not "strategy conflict," but "strategy synchronization" – all strategies issuing buy signals or sell signals at the same time. This means diversification has failed, and risk is amplified infinitely.

The Weak Form Efficient Market Hypothesis (Weak EMH) is the most basic and often most misunderstood level of the Efficient Market Hypothesis (EMH), the cornerstone of modern finance.

It was formally proposed by Nobel laureate Eugene Fama in 1970. To help you fully understand it, we need to break it down from four dimensions: definition, implications, impact on trading, and real-world paradoxes.

---

1. Core Definition: What is "Weak Form Efficiency"?

Definition: In a weak-form efficient market, the current price of an asset fully reflects all past market trading information.

"Past market trading information" specifically refers to:

• Historical price series

• Historical trading volume

• Past returns

Plain English Explanation: Everyone is looking at the same historical charts. If there were truly a "surefire pattern" hidden in historical price movements, savvy traders would have already discovered it. They would instantly buy or sell to capture that profit, causing the price to adjust rapidly. Therefore, any historical pattern you can see is useless for predicting future prices.

2. Core Implication: Technical Analysis is Useless

This is the most "controversial" part of the Weak EMH. If it is true, then:

1. Technical Analysis is Completely Useless:

• All candlestick patterns (like "three black crows," "head and shoulders," "golden cross/death cross").

• All technical indicators (MACD, RSI, Bollinger Bands).

• Conclusion: Chart-based stock trading is futile. You cannot achieve long-term excess returns (Alpha) by analyzing past charts.

2. Prices Follow a "Random Walk":

• Today's price change has no correlation with yesterday's price change (no serial correlation).

• It's like flipping a coin: even if you get 10 heads in a row, the probability of getting heads on the 11th flip is still 50%. Just because a stock went up yesterday doesn't mean it's more likely to go up today (no momentum) or down (no reversal).

3. Fundamental Analysis May Be Effective:

• Note that Weak EMH does not deny fundamental analysis. It only says "chart reading" is useless. But if you study financial reports, macroeconomics, and company management (which fall under public information, the realm of Semi-Strong EMH), you might still be able to make money.

3. Why Do Quant Strategies Seem to "Contradict" Weak EMH?

Your previous questions mentioned "momentum strategies" and "mean reversion strategies," both of which are fundamentally based on historical price data.

• Momentum: If it went up in the past, it will continue to go up.

• Mean Reversion: If it went up too much in the past, it will go down.

This creates a huge paradox: If the Weak EMH holds, and the past doesn't affect the future, then both momentum and mean reversion strategies should not be profitable. Yet in reality, these two strategies are among the most profitable sources for quant funds.

Academic and Industry Explanation: This proves that the market is not perfectly "weak-form efficient." The money quant investors make is essentially money from market inefficiency.

• Behavioral Finance Explanation: People are not rational. Because investors have psychological biases like "chasing trends" (leading to momentum) and "overreacting" (leading to reversals), prices do not instantly reflect historical information; there is a lag. Quant models capture this lag.

4. The True Status of Weak EMH Today

In today's investment world, the attitude towards Weak EMH is very nuanced:

A. For Retail Investors: It is the Truth

For ordinary individual investors, buying stocks based on a MACD golden cross is unlikely to be profitable in the long run. Simple chart patterns have already been exploited by thousands of quantitative algorithms. For retail investors, the market is weak-form efficient, or even stronger.

B. For Quant Institutions: It is a "Wall" to be Broken

The job of a quant institution is to find "statistically significant minor failures" of the hypothesis.

• They don't look at simple "golden crosses"; they use deep learning to uncover non-linear, extremely complex historical price patterns.

• They find that while simple candlestick patterns are useless, if combined with high-dimensional data like microstructure (order flow) and volatility surfaces, weak signals can still be found in historical data.

Summary

The Weak Form EMH is like the physics principle that "in a vacuum, a feather and a lead ball fall at the same speed":

1. Theoretically: It is the benchmark model of finance, telling us not to try to drive simply by looking in the rearview mirror (don't blindly trust technical analysis).

2. In Reality: Air resistance (human irrationality, market frictions) exists.

3. Conclusion: The market is close to weak-form efficient, but not perfectly weak-form efficient. Quantitative traders are those who use microscopes to find the "air resistance" and profit from it.

This is a very direct and soul-searching question. To answer you without beating around the bush:

For a single stock, there are indeed approaches with "extremely high win rates," but these approaches usually come with two fatal costs: either "extremely low returns" or "extremely high risk (losing everything at once)."

If your goal is a method to predict a single stock that combines "high win rate + high return + low risk," then the answer is: It doesn't exist. If it did, it would be insider trading (illegal) or a Ponzi scheme.

We can break down the "prediction of a single stock" into the following harsh realities:

1. The "High Win Rate" Trap: Picking Up Pennies in Front of a Steamroller

You can absolutely build a strategy with a win rate as high as 90% or even 99% on a single stock, but it's poison.

• Approach: Suppose a stock is currently at $100. You predict that "its decline tomorrow will not exceed 10%."

• Operation: Sell deep out-of-the-money put options, or engage in some form of ultra-short-term grid trading.

• Result: On 99% of trading days, the stock indeed does not crash 10%, and you earn a tiny bit of "premium" or "micro-profit" every day. Your win rate looks close to 100%.

• Cost: On that remaining 1% of days (e.g., earnings blowup, regulatory black swan), the stock drops 20%. You will lose all the money you made over the past 3 years in a single day, and may even owe the brokerage money.

Conclusion: Pursuing a high win rate on a single stock often means exchanging infinite tail risk (holding losing positions) for a false sense of high win rate.

2. Why is a Single Stock Harder to Predict Than the Market?

Quant institutions usually avoid heavy positions in single stocks because the signal-to-noise ratio is too low.

• Individual Randomness: Just as you can fairly accurately predict "how many people will die in car accidents nationwide next year" (law of large numbers), you absolutely cannot predict "whether your neighbor will get into a car accident tomorrow."

• Concentrated Black Swans: A single stock faces unique risks (Idiosyncratic Risk). CEO suddenly resigns, factory catches fire, major customer cancels an order... These events are mathematically discrete and unmodelable.

• Game of Major Players: On a single stock, if your capital is small, you are a victim of the random walk; if your capital is large, you are prey for the opposing players.

3. Where Does True High Win Rate Exist? (You Probably Can't Participate)

Although ordinary people cannot predict, on certain extremely short or extremely special dimensions, the movement of a single stock is predictable:

A. Microstructure Level (Milliseconds)

• High-Frequency Trading (HFT): They don't predict "whether it will go up tomorrow," but rather "whether there are more buy orders or sell orders in the next 10 milliseconds."

• Logic: If they see a large influx of buy orders (Order Book Imbalance), the probability of the price rising in the next 1 second might be as high as 80%.

• Barrier: You need hundreds of millions of dollars in equipment, servers placed in the exchange's data center (Co-location), and top-tier algorithms.

B. Event-Driven Arbitrage

• Merger Arbitrage: Company A announces it will acquire Company B for $50 per share. Company B is currently trading at $48.

• Logic: As long as the acquisition doesn't fail, the stock price will eventually rise to $50. This has a very high win rate.

• Risk: If an antitrust investigation causes the acquisition to fail, the stock price could instantly drop back to $30.

C. Insider Information (Illegal)

• This is the only way to achieve a "high win rate + high return." For example, knowing earnings data in advance. But this falls under criminal law.

4. For Ordinary Traders, What is a "Relatively High Win Rate" Approach?

If you insist on trading single stocks and want to improve your win rate, the quantitative mindset suggests abandoning point prediction and moving to conditional probability:

• Don't ask: "Will Moutai go up tomorrow?" (This is a coin flip, 50% win rate)

• Ask: "If Moutai has already broken below its annual moving average, and trading volume has increased 3-fold, and the entire baijiu sector is falling, then what is the probability it will continue to fall over the next 3 days?"

This is the essence of quant: finding "conditional biases." Even with all these conditions, the win rate might only improve from 50% to 55% or 60%. This is already the level of a top expert.

Summary

The rise and fall of a single stock is Brownian motion (random) in the short term and a reflection of fundamentals in the long term.

So-called "high win rate approaches" usually come in only two forms:

1. You are the market maker/insider (you can see the cards).

2. You are running a "high probability of making small money, low probability of losing big money" strategy (survivorship bias – as long as it hasn't blown up, the win rate looks high).

For individual investors, the most rational approach is not to look for a "surefire prediction," but to admit "I don't know whether the next move will be up or down," and therefore use money management (stop-loss, position sizing) to ensure that even if you guess wrong, you won't die.

---

One step I can take for you: Since we mentioned the relationship between win rate and payoff, many investors lose money because of "high win rate but small profits, low win rate but large losses." Would you like to learn how to calculate the "Risk/Reward Ratio" and see how you can still be profitable even with a win rate of only 40%?

The Weak Form Efficient Market Hypothesis (Weak EMH) was a watershed moment in finance, marking its transition from "alchemy and mysticism" to a "serious science."

To understand it, we need to go back to the mid-20th century, when computers were just being born and Wall Street was still drawing charts by hand.

I. Historical Background: From "Finding Cycles" to "Discovering Randomness"

Before the 1950s-1960s, financial markets lacked a proper theoretical framework. The prevailing views were divided into two camps:

1. Chartists: The forefathers of modern technical analysis. They believed stock prices, like tides or weather, had fixed "cycles" and "waves." If you could find the pattern, you could predict the future.

2. Fundamentalists: They believed stock prices were ultimately determined by company value, but lacked systematic data analysis tools at the time.

Key Turning Points (Background Events):

1. Kendall's "Failure" and Shock (1953): British statistician Maurice Kendall attempted to use the powerful computing power of the time to analyze the UK stock market, with the original goal of finding "cyclical patterns" in price movements. Result: He was shocked to find absolutely no patterns! The data looked like a drunkard's walk (random walk). He wrote in his paper: "The data behaves like a wandering drunk, with no sense of direction."

2. Introduction of Computers to Financial Research (1960s): As computers entered universities, young scholars like Eugene Fama began testing stock prices with massive datasets. They found that no matter how complex the statistical model, they could not use past prices to predict future prices.

3. Fama's Synthesis (1970): Eugene Fama published his landmark 1970 paper, "Efficient Capital Markets: A Review of Theory and Empirical Work," formally proposing the Efficient Market Hypothesis and defining "weak form efficiency" as the baseline state.

---

II. What Problem Was It Meant to Address? (Core Logic)

The Weak Form EMH was primarily proposed to explain and resolve the following three puzzling financial phenomena:

1. To Explain "Why Technical Analysis (Chart Reading) is Useless in the Long Run"

This is the Weak EMH's most direct target.

• Problem: Since everyone knows about candlestick patterns, trendlines, double bottoms, etc., why are so few people getting rich from them?

• Explanation: Because in an informationally efficient market, "historical price information" is the cheapest and easiest information to obtain.

• If "a stock always goes up on the fourth day after three consecutive down days" were true, everyone would buy at the close on the third day.

• This buying activity would directly push the price up, meaning the profit from the "fourth-day rise" would be captured on the third day.

• Conclusion: Any pattern based on past prices, once discovered, will be instantly arbitraged away and become ineffective.

2. To Explain "Why Price Randomness Does Not Mean Market Chaos"

When Kendall discovered that stock prices followed a "random walk," people panicked, thinking the stock market was a casino with no rationality. The Weak EMH provided a brilliantly counterintuitive explanation:

• Problem: Why are stock price changes random?

• Explanation: This is precisely because the market is too smart (too efficient).

• Price is a reaction to information.

• Future news (good or bad) appears randomly and is unpredictable.

• Because the market reacts very quickly, yesterday's information is already reflected in yesterday's price. Today's price change depends only on today's new information.

• Conclusion: The randomness of prices reflects the randomness of new information, not the irrationality of the market itself.

3. To Explain "There is No Free Lunch"

It aims to shatter the illusion that investors can achieve excess returns through simple diligence (e.g., staring at a trading screen all day).

• Problem: Why do so many professional fund managers fail to beat the index?

• Explanation: Because the "historical patterns" they spend so much effort studying contain no useful information for the future. In a weak-form efficient market, the next move is independent of the previous one (like a coin flip; getting 10 heads in a row doesn't change the 50% probability of the next flip).

---

Summary

Background: It was proposed in the context where, after computers were applied to financial analysis, scholars discovered that "cyclical patterns in stock prices simply could not be found."

Problem Solved: It theoretically pronounced a "death sentence" for pure technical analysis (in an academic sense). It told the world: Don't try to find the road ahead in the rearview mirror. If you want to beat the market, you must dig for information beyond public information (which led to the Semi-Strong and Strong Form EMH).

It's like a magic trick being revealed: once you know how the trick (historical pattern) is done, the trick (excess returns) is no longer magical and can no longer be used to fool people.