Porosity is easily formed in welded joints during high power laser welding due to keyhole instability. Large pores have detrimental effects on the fatigue resistance of a component and cause many failures in welded parts. This paper is aimed at predicting the maximum pore dimension in long laser welded joints, starting from the sampling of large pores in shorter joints. Two sampling strategies and, consequently, two estimating techniques, both belonging to the statistics of extremes, were explored. The first approach, extreme value type, is used to estimate the size of the maximum pore in each of a series of steel samples. In each sample, the larger single pore or two large pores which are very close are the measured maximum pore. The second approach, threshold value type, is used to estimate the size of pores larger than a critical threshold in a single sample of steel. Both approaches lead to good estimates of the largest pore distribution in short laser welded joints. However, the first one is more adequate to describe the largest pore distribution, because it allows the synergetic effect of two adjacent pores to be considered. In particular, the Gumbel distribution adequately fits the experimental data even in the case of welded joints 10 times longer than the investigated bead length.List of symbols area 1=2 max square roots of the area of the largest pore in extreme value approach area 1=2 i square roots of the area of pores larger than the initial threshold in threshold value approach c a (12a) quantile of the x 2 1 distribution D deviance function k number of pores that exceed the threshold u l sample log likelihood function L return bead length L 0 standard testing length for extreme value approach L T total investigated length n total number of pores in the total investigated length N total number of block maxima examined T return period u threshold size u 0 initial threshold size for threshold value approach y T reduced variate for plotting GEV or SEV probability plot m GEV estimated location parameter of GEV distribution m SEV estimated location parameter of SEV distribution ĵ GEV estimated shape parameter of GEV distribution ĵ GP estimated shape parameter of GP distribution ŝ EXP estimated scale parameter of EXP distribution ŝ GEV estimated scale parameter of GEV distribution ŝ GP estimated scale parameter of GP distribution ŝ SEV estimated scale parameter of SEV distribution