The two greatest statistical outliers of all time are Bradman and Gretzky. A good way of measuring this, for sports like these, is to look at averages, and how many standard deviations they are above the rest (or a Z score). Statistician Charles Davis did the heavy work for multiple sports, although excluding the ice hockey players.
The results:
Bradman has a z-score of 5.0 when considering batsmen only when taking batting average in Tests
Pele has a z score of 3.7 when considering strikers, and goals per game at international football.
Ty Cobb has the best z score in baseball, with a z score of 3.6 when looking at his batting average.
Jordan has a z score of about 3.4 in basketball - about the same as Wilt Chamberlain, so to argue he is more dominant that him would be wrong.
I've found another site where someone used the same methodology to look at Gretzky's record - although using total career points instead of points per game (as most players had a similar number of games played today than 60 years ago, not the case in other sports). For that, Gretzky had a z score of about 4.6 - and in terms of points per game 4.9 (with Lemieux, who only played about 600 fewer games) on 4.7.
For reference, Jordan averaged almost 31 points per game - Charles Davis calculated that for Jordan to have a z score of 4.5, he'd need to have averaged 43 points a game - more than a third more than his actual average. Equally, applying the metric to other sports with less statistical averages gives a gold career needing more than 25 Majors; or a tennis player would need to average winnning almost 2 Grand Slams a YEAR across their entire career.
All that said, I voted Bradman instead of Gretzky as cricket is a sport closer and nearer to my heart than Ice Hockey.