Ken E. Legé, Ph.D., Extension Cotton Specialist writes: Many sources of variety yield performance data are available from Texas A&M AgriLife Extension and Research trials each year. Replicated Agronomic Cotton Evaluations, or RACE trials, are large-plot, on-farm trials conducted by Extension personnel and grown and managed by grower/cooperators at multiple locations throughout the high plains. Small plot replicated variety trials are conducted by Texas A&M AgriLife Research at Halfway, Lubbock, and Lamesa, TX, each year by Dr. Carol Kelly and her cotton breeding team. Other AgriLife trials involving the comparison of varieties are conducted by other researchers under various disease, nematode scenarios or under agronomic systems, such as planting dates or irrigation regimes. For this study and factsheet, any AgriLife trial involving the comparison of varieties was compiled and analyzed for yield stability.
What is yield stability?
While it is advised that growers evaluate variety performance from trials near their farms, and/or that are grown under similar agronomic practices to their farms, most growers may only have a few locations of variety trials at their disposal. These few locations are not sufficient for a grower to evaluate the stability of a variety over a range of environments. Yield stability analyses evaluate how consistent a variety performs under a wide range of yield environments. Some varieties may perform well under a certain yield environment, but may not perform well under other yield scenarios. Knowing yield stability, or predictability of performance, of a variety increases the confidence of the variety selection decision.
For this study, every trial which each variety was tested was used for the analysis. Lint yields of a variety were regressed against the mean lint yield of each trial (Eberhart & Russell, 1966). The resulting regression equation was used to evaluate each variety’s stability as such: Y = a + b(X), where a = intercept, b = slope, X = mean lint yield of each trial, and Y = lint yield of the variety at a trial location. The coefficient of determination, or R2, describes the deviation of each datapoint from the line of the equation, such that higher R2 values indicate higher degrees of predictability. Varieties that were commercially available in 2025 were analyzed across years (ranging from 2019 to 2025). In the following table, yield environments (X) were set at 500, 750, 1000, 1250, 1500, and 1750 lbs/A. The values for each variety in the various yield environment columns indicate the deviation of variety performance within each yield environment. Positive values indicate that that variety performed better than average in that yield environment, while negative values indicate that that variety performed below average in that yield environment. For example, Armor9371B3XF averaged 53 lbs/A better than average in the 500 lb/A yield environment, or 553 lbs/A (i.e., 500 + 53). However, that variety performed 34 lbs/A below average in the 1750 lb/A yield environment, or 1716 lbs/A (i.e., 1750 – 34). The number of trials is listed in the ‘# locs’ column, and the year(s) that each variety appeared in AgriLife trials are listed in the ‘Year(s)’ column. The R2 values are presented in the ‘Predictability’ column.
How to interpret the data
Yield responses and predictability values are color coded, such that deeper shades of green indicate better performance, and deeper shades of red indicate poorer performance. Broadly adapted varieties with good predictability are indicated by green shades under every yield environment and green shading under predictability. Higher numbers of locations (# locs column) and multiple years indicate higher degrees of confidence in the data. If the number of locations and/or the number of years tested are lower, further testing in future years may potentially increase the confidence level of a variety’s response.
