What Is Quantile Analysis? Evaluating Factors by Groups

Quantile analysis ranks securities by factor value at the same time, divides them into groups and compares subsequent outcomes such as returns. It examines differences between high and low values and the progression through intermediate groups.

Define the factor, its direction and the future-return interval first. Check whether larger values are intended to be preferred, alongside the group labels.

For illustration, assign five invented securities to two upper, one middle and two lower observations. This small sample cannot produce equally sized groups. Check the actual implementation and configuration for its partition rule.

Equal-weight means in this teaching example are +3% for the upper group, +1% for the middle and −1% for the lower. The outcomes increase with the group’s score. This deliberately constructed single example does not establish stable predictive ability.

The upper-minus-lower difference is four percentage points. It is not an executable trading profit. Shorting feasibility, portfolio weights, costs and execution conditions require separate evaluation.

Ties at group boundaries can change the partition depending on whether equal values stay together or target group sizes take priority. Ordering ties by code creates no new predictive information. Record the partition rule and daily group counts.

Missing features or future returns can shrink the evaluation universe. Do not replace missing returns with zero. Preserve exclusion reasons and coverage information, and check whether the remaining universe has become selectively favorable.

Large daily changes in upper-group membership may imply more trading if those rankings drive holdings. Group membership turnover and actual portfolio turnover based on allocations are different calculations.

Separate portfolio outcomes including transaction costs from evaluation labels. IC provides another view of the overall relationship alongside differences between groups.

KabuForge’s public example contains three securities and therefore at most three quantiles. A single security can dominate, while ties or missing values further reduce information. Favorable group returns do not establish a relationship across the Japanese equity market.

Check overlapping windows when computing multi-session future returns each day. The factor diagnostics tutorial separates features from future labels and teaches inspection of counts, missingness and group membership before conclusions.

KabuForge GUI research courses.

This article teaches research procedures. It does not recommend trades or guarantee future returns.Disclaimer

Explore the factors learning path

• Factor Coverage: Check Which Securities and Dates Reach Evaluation
• Rank Normalization: Comparable Scales Still Need Clear Conventions
• Factor Correlation: Check Whether Features Repeat the Same Information