Quantitative research basics
Six introductory articles explain quantitative research, backtesting, factors, IC, NAV, drawdowns, costs, overfitting and look-ahead bias.
Six introductory concepts
01 · Quantitative research and backtestingDefinition, interpretation and a practical example02 · Factors, IC and quantile evaluationDefinition, interpretation and a practical example03 · Returns, NAV and drawdownsDefinition, interpretation and a practical example04 · Costs, turnover and executionDefinition, interpretation and a practical example05 · Overfitting and out-of-sample evaluationDefinition, interpretation and a practical example06 · Data availability and look-ahead biasDefinition, interpretation and a practical example
Put the concepts into practice
Explore individual topics
What Is Backtesting? What Historical Tests Can and Cannot Tell YouBacktesting applies trading rules to historical data. Learn how to read orders, simulated fills and skipped orders, and understand the assumptions behind the results using a public KabuForge example.What Is a Factor? Comparing Security Characteristics with NumbersLearn what a factor measures, how scores differ from rankings and trading orders, and why factor evaluation does not establish investment profitability. Public KabuForge examples keep features separate from future-return labels.What Is IC? Reading the Relationship Between Factors and Future ReturnsCompare IC and Rank IC using fictional data, and learn how labels, missing values, sample size and costs affect interpretation.What is maximum drawdown? Measuring the fall from a portfolio peakLearn how maximum drawdown is calculated and how to read an underwater chart using a short portfolio-value example.What are transaction costs? Why fees change backtest resultsUnderstand fees, spreads, slippage and turnover, and learn how to interpret results after costs.What is lookahead bias? Using information that was not yet knownLearn how lookahead bias arises from information timing and the misuse of future evaluation labels.What is point-in-time data? Understanding historical availabilityDistinguish observation dates, publication times and revisions, and understand why frozen data is not necessarily point-in-time data.What to check first in J-Quants price dataInspect existing local price data for dates, price conventions, duplicates, trading halts and freeze records.What Is the Sharpe Ratio? Why the Number Alone Is Not EnoughUnderstand excess returns, volatility, sampling frequency and annualization, and learn why a Sharpe ratio needs context before comparison.What Is Out-of-Sample Evaluation? Handling Periods You Have Already InspectedSeparate development and evaluation periods, understand how repeated inspection changes their interpretation, and keep a record of rules and data versions.What Daily Bars Cannot Tell You: Highs, Lows and Execution OrderIdentical daily OHLC values can hide different intraday paths. Learn why touching a price does not establish execution and how to state backtest assumptions.What Is Quantile Analysis? Evaluating Factors by GroupsGroup securities by factor score and examine future returns, group sizes, missing data and membership changes using an explicitly fictional example.What is momentum? Separate a price trend from a rankingUse fictional prices to distinguish a security’s own trend from a cross-sectional ranking.What is a moving average? Check its window and availabilityCalculate a simple moving average and distinguish warm-up gaps, observation time and execution.What is rebalancing? Why target and actual weights differUse a fictional two-asset example to distinguish allocation targets from executed holdings.Portfolio concentration: holdings, largest weight and HHICalculate concentration with fictional weights and check the treatment of cash.What is turnover? Distinguish one-sided and two-sided measuresCalculate turnover from fictional trades and check numerator, denominator and aggregation.What is slippage? Separate reference prices from fillsDistinguish price differences, fees, execution delays and actual fills with fictional trades.Price return and total return: check the benchmark basisUse a fictional dividend example to distinguish price return from total return.What is survivorship bias? The problem with today’s stock listUse fictional securities to examine the difference between survivors and the historical universe.Factor Coverage: Check Which Securities and Dates Reach EvaluationSeparate eligible securities, valid features and future labels before interpreting changes in the evaluation sample.Rank Normalization: Comparable Scales Still Need Clear ConventionsCompute an illustrative rank transformation and inspect ties, missing values, direction and weights.Factor Correlation: Check Whether Features Repeat the Same InformationDistinguish feature-to-feature correlation from IC and check the common sample and invalid observations.ICIR: What the Mean and Variation of IC Can Tell YouCalculate an illustrative mean-to-standard-deviation ratio and inspect sample size, overlapping labels and inferential limits.Historical Paper Replay and Forward Testing: Different EvidenceSeparate replayed history, newly observed forward records and real trading evidence.What Is J-Quants? Research Uses and LimitsJ-Quants provides Japanese market data for individual users.J-Quants Price Data: Fields and Observation DatesA daily bar represents one security on an observation date.J-Quants Financial Data: Periods and UnitsFinancial data has a reporting period and a disclosure timeline.J-Quants Disclosures: Times and Financial RevisionsA reporting period ends before its results are published.Japanese Equity Backtests with J-Quants: Input DesignDefine what information drives a decision, when that decision occurs and when execution follows.J-Quants Adjusted Prices and Stock SplitsIn a fictional two-for-one split, a price moving from 2,000 to 1,000 alongside doubled shares does not itself imply a halving of holding value.J-Quants and Point-in-Time ResearchPoint-in-time research checks whether every input was available by the corresponding decision cutoff.Factor Analysis with J-Quants Data: Inputs and EvaluationFactor analysis turns security characteristics into values and evaluates their relationship with later returns.J-Quants Listed Issues and Market SegmentsA security master changes with listings, delistings and classifications.J-Quants Data Storage and Reproducible ResearchReproduction requires matching input content, selection rules, transformations and software versions.Value investing: separate valuation measures from future returnsDefine a valuation characteristic with fictional book equity and market capitalization.Quality factors: fix a profitability definition before comparing firmsUse fictional earnings and equity to study profitability as one dimension of quality.Size factors: distinguish market capitalization from a return spreadDistinguish a company’s market capitalization from a portfolio return spread.Low volatility: measure historical variation without promising safetyUse return dispersion to distinguish a low-volatility characteristic from future safety.Multifactor research: check scale and overlap before combining scoresCombine fictional normalized scores while preserving direction, timing and validation rules.CAGR: separate compound growth from the path of lossesCalculate compound annual growth with fictional values and no external cash flows.Volatility: align return frequency and standard-deviation conventionsCalculate fictional daily volatility with explicit ddof and annualization assumptions.Sortino ratio: define the target and downside denominatorUse fictional monthly observations to define target shortfall and the full-sample denominator.
Applies to: main source candidate (unpublished) · Public source docs/en_US/GUI_RESEARCH_COURSES.md
Reviewed · KabuForge