Plant breeder use yield traits to identify promising genotypes. This goal depends the magnitudes of genotype by environments and stability performance of genotypes. Therefore, twenty Egyptian extra long genotypes were grown in three locations under two years for yield, yield components earliness and fiber traits to identify promising stability genotypes. The genotype x environment interaction was significant for yield and fiber traits. It also noticed that variation due genotype x environment were further partitioned into linear and non-linear components. Genotype x environment linear was insignificant for all studied traits except for MC trait, insignificant of genotype x environment linear indicated that genotypes didn't differ genetically in their response to different environments. Pooled deviation mean squares were significant for all studied traits , indicated that the major components for differences in stability were due to deviation from the linear function. Therefore, it could be concluded that the relatively unpredictable components of the interaction maybe more important than the predictable components. The results illustrated that lines 3, 5 and 15 were stable for seed cotton yield (x^-= high, b =1 and s²d =0).While, the line (7) has high mean performance and regression coefficient equal to one but the deviation from regression was larger than zero. However, for lint percentage, some lines has high mean performances over grand mean (2, 3, 7, 8,9,12 and 14) but these lines did not parallelism with the stability parameters. Therefore, the best performing, highest value in this trait or genotypes was not necessary. The best stable genotype for fiber length is 2, 10, 11 and 12 when had  mean performances same the check variety (Giza 70) and regression coefficient (bi)was equal unity for all genotypes and deviation from regression was significant differ from zero The best genotypes according to these criteria (three indexes) are also identified in this when selection based on mean yield a lone when have yield ranks of one because all lines were similar for mean performance for yield due to low variability of these material ( Extra long staple ) while , selection based on index3 the top lines were 1, 3,5,6,9,11,15and16. Using principal components analysis to selection the better stability lines to comparison regression model, the results shown that the percentage contribution of PCA components of seed cotton yield. Each PAC1and PCA2 were more important. In addition results show that the strains 1and 3 which PCA1    equal unity and PCA2 equal zero. The two strains were stable by using regression model (x=high=1 and s^2=0).On the other hand, lines 5 and 15 were stable by using regression model but the values of PCA equal zero and PCA2 close to unity. Therefore, using the two models to identify promising genotypes stability in cotton breeding programs is very useful.