His track record seems clear at first: he struggled as a starter, but has done well as a reliever. In 42 innings as a starter (all in 2002), he compiled a 5.10 ERA, striking out 14% of batters faced and walking 9%. For comparison's sake, the 2005 Joel Pineiro struck out 13% of batters he faced and walked 6% (The league average is about 16% for Ks and 9% for BBs). One big problem for Soriano was facing lefties (.870 OPS allowed).
Conversely, in 69 2/3 career innings as a reliever, he has a 2.07 ERA. His K% is 16% and his walk rate is 6%. He still was more successful against righties, but he was still good against lefties.
Most qualitative observations I've heard note that Soriano lacks a good change. Conventional wisdom (and I've heard little dispute from even the most die-hard statheads) is that pitchers with two good pitches excel in relief roles, but struggle as starters, because hitters adjust to a smaller repertoire after fewer innings. Once Soriano makes it through the lineup twice, so the story goes, he has run out of things with which to fool batters.
This is a broad generalization of Soriano's performance so far, however. It's normal for a younger, inexperienced pitcher to struggle at first, especially if his development has been quick (as Soriano's was). Moreover, his 2002 season ended with injury, suggesting that he was not only young and inexperienced at the time of his starts, but also may have been hurt. Comparing his starts to his relief appearances may be comparing apples to, you know, that other thing.
Furthermore, it's not entirely clear that Soriano was less effective the longer he stayed in the game:
Situation Batters K% B% Balls in
Pitches 1-15 34 15% 6% 79%
Pitches 16-30 40 18% 13% 65%
Pitches 31-45 34 12% 9% 76%
Pitches 46-60 29 21% 14% 66%
Pitches 61-75 30 10% 0% 90%
Pitches 76-90 20 25% 5% 50%
Pitches 91-105 11 9% 9% 73%
Inning 1-3 102 11% 12% 75%
Inning 4-6 74 18% 5% 72%
Inning 7-9 9 22% 0% 67%
If Soriano were easier to figure out after a couple times through the rotation, we'd expect his numbers to get consistenly worse the longer he pitched...but that's clear from these numbers. You might say that he stopped missing bats after about 60 pitches, but then he looks unhittable after 75 pitches. There's probably not a large enough sample size to draw many conclusions.
In the context of the Mariner's current rotation, Soriano could be one of the better starters, but only if he performs as well as his career reliever numbers, not his career starting numbers. As a starter, he had similar numbers as Gil Meche's last three years--both have K% and BB% of 16 and 9 respectively, and Soriano actually gave up HRs more frequently as a starter (4.7% to Meche's 3.5). As a reliever, Soriano's HR rate plummets to .4% (yes, that's right: a tenth of his % as a starter), and his K rate a walk rate improve significantly, too. But can Soriano bring his more recent effectiveness into a starting role? I don't think his numbers can tell us.
That leaves us with qualitative rather than quantitative analysis. Soriano does have electric stuff, but does he have enough pitches to make it as a starter? If Soriano makes an okay starter but a dynamite reliever, which do we want?