It's not 20% of a statistic. 20% of the games but not statistic. If a defense gives up 250 yards passing on 3 games and 1 game gives up 180 yards passing it's not going to move the needle very far. It would only change the average by 17.5 yards.
There had been 5 games from which they compiled DVOA when I wrote that. So 1 game out of 5 would be 20% of the games and responsible for 20% of the DVOA calculation.
It may not seem like much, but 17 yards would be the difference (in NFL passing defense, for example, which uses YPG) between a team being ranked 9th and in the top ten defenses, and being ranked 16th and dead on league average. One game out of five absolutely has an outsize impact...again, it's 20% of the accumulated statistics at the time it was calculated, as opposed to 6.25% which would be one game out of a 16 game season. So in essence you could say one game out of five is over three times as meaningful to DVOA when calculated through week 6 as opposed to week 16, so you need to look at outliers in that light IMO.