I have Madson on 4 teams...and I thought it would be interesting to share his performance in PB this season. It may shed some light on how the game uses raw stats to generate results. With the realization it is a small sample size, the results are astounding.
First, his actual season numbers:
He actually was used for 2 IP, gave up 0 runs, 0 hits, and had no fielding chances. RH batters were 0-4 against him. LH batters were 0-2 against him. His PB ratings were Grade 8.5, Clutch 8.4. His error rating was 0.1 (he had no fielding chances).

His performance...by league

D5: 1 game, 8.3 IP, 0 hits, 0 K, 0 BB, (W,1-0), 2 errors.

C11: 1 game, 10.6 IP, 0 hits, 0 K, 0 BB, no decision, 2 errors.

C16: 2 games, 9 IP, 0 hits, 1 unearned run, 0 K, 0 BB, (Sv-1), no errors made.

F18: 1 game, 8 IP, 0 hits, 0 K, 0 BB, (W,1-0), 1 error.

So, in 36 IP, 0 hits allowed, 0 BB, only baserunners allowed were on errors. Only allowed 1 unearned run.

So, his 0 hits allowed (actual stats), really worked in reducing hits allowed in PB. The PB Grade of 8.5/Clutch 8.4, seems too low to allow these types of results.

Interestingly, his actual useage of 2 IP, meant that I used him in 1 game (3/4 leagues), and when he came in for long relief, he went as long as possible...ending up with more IP than actually expected. In 1 league, he had two appearances, the first one for less than 2IP, then the second one as long as possible. I realize that I used him much more than he was allotted, but it was a fun interesting experiment.