Statistical Methods for Determining Optimal Rifle Cartridge Dimensions

We have designed and carried out a statistical study to determine the optimal cartridge dimensions for a Savage 10FLP law enforcement grade rifle. Optimal performance is defined as minimal group diameter. A full factorial block design with two main factors and one blocking factor was used. The two main factors were bullet seating depth and powder charge. The experimental units were individual shots taken from a bench-rest position and fired into separate targets. Additionally, thirteen covariates describing various cartridge dimensions were recorded. The data analysis includes ANOVA and ANCOVA. We will describe the experiment, the analysis, and some results.

💡 Research Summary

**
The paper presents a statistical investigation aimed at identifying the optimal cartridge dimensions for a Savage 10FLP law‑enforcement rifle, with “optimal performance” defined as the smallest group‑mean radius (MR) of shot groups. The authors employed a full‑factorial experimental design with two quantitative factors—bullet seating depth (six levels ranging from 0.005 in to 0.030 in in 0.005 in increments) and powder charge (ten levels ranging from 25.3 gr to 26.2 gr in 0.1 gr increments)—yielding 60 treatment combinations. Because the experiment was conducted in four lots of 100 cases each, the lot was treated as a blocking factor to account for systematic differences among the case batches. Random assignment of cases to treatment combinations resulted in an uneven number of observations per cell (5–8 shots), and after discarding 20 shots due to wind gusts, human error, copper fouling, and target‑hole ambiguities, the final data set comprised 380 valid observations.

The response variable, MR, was calculated as the average Euclidean distance from each bullet hole to the group centre, allowing each shot to be treated as an individual experimental unit. Initial analysis used SAS GLM to perform a two‑factor ANOVA with lot as a block. The overall model was significant (p = 0.0008), but the main effect of seating depth was not (p = 0.57), whereas powder charge was marginally significant (p = 0.0378). The interaction between seating depth and powder charge was highly significant (p = 0.0006), indicating that the effect of one factor depends strongly on the level of the other. Contour and surface plots suggested two “valleys” of low MR: one around 25.5–25.9 gr powder with seating depths of 0.015–0.025 in, and another near 25.9 gr with similar depth ranges. However, the authors did not provide confidence intervals or formal post‑hoc comparisons, limiting practical interpretation.

To incorporate additional cartridge geometry information, the authors recorded thirteen covariates (case length, neck inner/outer diameters, neck thickness, headspace, primer pocket depth/diameter/weight, case weight, case volume, bullet overall length, bullet weight, and a binary “case mouth square” variable). An ANCOVA model including all covariates was fitted. The overall ANCOVA model remained significant (p = 0.0009), but only primer weight emerged as a statistically significant covariate (p < 0.001). Primer weight had two levels (3.1 gr and 3.2 gr); the lighter primer yielded a smaller MR, leading the authors to recommend the 3.1 gr primer for better accuracy. All other covariates showed p‑values well above 0.05, suggesting either negligible influence or insufficient power to detect effects.

Critical appraisal reveals several methodological shortcomings. First, the per‑cell replication (5–8 shots) is low for a 6 × 10 factorial design, resulting in limited statistical power and an unbalanced design that can bias estimates. Second, while lot was used as a block, the authors did not test for a significant block effect nor report block‑adjusted means. Third, environmental factors (wind speed, temperature) and the documented copper fouling were not modeled as covariates, despite evidence that they may have introduced systematic error. Fourth, the authors relied solely on ANOVA/ANCOVA without conducting model diagnostics (normality of residuals, homoscedasticity) or exploring alternative modeling strategies such as response‑surface methodology (RSM) or mixed‑effects models that could better accommodate the hierarchical structure (shots nested within lots). Fifth, multicollinearity among the thirteen covariates was not assessed, and no variable‑selection procedure (stepwise, LASSO) was employed, raising the risk of overfitting. Finally, the practical recommendation (optimal powder charge and seating depth) is vague because the interaction surface is not quantified with confidence bounds, and the effect size of the significant covariate (primer weight) is marginal (0.1 gr difference).

In summary, the study demonstrates a commendable attempt to apply experimental design and statistical analysis to rifle cartridge optimization, but the limited replication, data loss, inadequate control of nuisance variables, and simplistic analytical approach constrain the robustness of its conclusions. Future work should aim for a balanced, adequately replicated design, incorporate environmental measurements as covariates, employ response‑surface designs (e.g., central composite) to locate the true optimum, conduct thorough model validation, and present results with confidence intervals and practical tolerance specifications. Such enhancements would yield more reliable guidance for shooters and manufacturers seeking to fine‑tune cartridge dimensions for maximal accuracy.