The system we're currently using for Shadowdawn is breaking badly in several places.  It needs to be fixed or replaced.  Problems right now are:

* If you have an effective skill of 25, you either succeed by 9 or more, or fail.  You never "just barely make it".  I'm attracted to pool-based systems like Shadowrun, Trinity, or Exalted, where the entire range of success is always available, and the bell curve stays bell shaped instead of looking like high school test scores (bell curve "centered" on 100%, so the upper tail is chopped off, and the mode is in the high 90s).  There are tasks which should be automatic successes, and I love d20's "take 10" and "take 20" mechanics for this, but automatic successes in combat are rarely appropriate.  Feng Shui's system also has appeal: you have your stat+skill, and you add in a d6 and subtract a d6.  Any 6's rolled are rerolled and added in the appropriate direction.  You're trying to beat a target number.  It does require much tighter balance than Shadowdawn currently uses.

* I don't know what the difference between success by 10, by 15, or by 20 is, but with people in the party with skills of 25, I'd better be prepared to describe each.  The best fix I can think of is to toss margin-of-success mechanics as much as possible: every roll leads to one of Critical Success, Success, Failure, or Critical Failure.  This will lead to painful combats when everybody has a defense over 15.  One option is to allow Feints to subtract your skill Rank from somebody else's defense... that doesn't fix resistance rolls, though GURPS has a klugy fix for this too: the RuleOfSixteen.  Again, pool-based systems seem a good fix.

* Durability is leading to characters wandering around at HP totals where any substantial hit causes them to start rolling to pass out, and often a death check.  I don't want characters to be bedridden for months between adventures — that doesn't seem very heroic — so I'm really tempted by systems like Feng Shui's "you heal everything between adventures.  Don't look too closely at this."  Another alternative is leaving people with their natural HP (though allowing them to be boosted as Circle Advantages), but having Durability grant 1 DR per level.

* Hit Points are breaking down; mooks noticably fall down from taking damage, but most of the party and their opposition take so much damage to reliably knock out that attacks like Daze and head blows, which offer cheap knock-outs, are The Only Way To Fight.  Also, Durability has eroded the previous wide margin between knock-out and death.

* Cyber shouldn't be boosting Talents, but it does because it boosts attributes and Talents're based on attributes.  Justin pointed out the obviously correct fix: Talents are based on natural attributes, and raise the spectre of Talent Crises for Adepts who rely on anything other than their Talents.

* Fifteen ranks of a Talent is too much.  Maybe fifteen circles are too.  If I had it to do again, I'd compress things down to about five ranks of each Talent, and maybe eight or nine circles (two or so each of apprentice, journeyman, master, warden).  Really rank 7 and rank 8 in a Talent aren't noticably different.  Half a dozen seems about right.  Maybe just four Circles, seven ranks in each Talent, and a broader variety of Talents, with a clear tree back to their base skill.


* Some folks have questioned DR — maybe there's too much running around.  For example, an average man does 1d-4 damage with a punch; assuming he has Brawling, 1d-3 (min 1): 1, 1, 1, 1, 2, 3.  That means he can't hurt a Troll without gouging at eyes.  Twelve seconds of uninterrupted beating on the Troll's head will knock out an average Troll nonetheless, but maybe this is a problem.  I don't think it is: that same man with a sword does d6 damage, wounding the Troll on half of his hits.  A handgun does three hits of 3d each in a round, enough to knock any Troll out and have him bleeding to death within seconds.  An average Troll does 2d+1, and so essentially always wounds another Troll with a bunch, and squashes most Windlings.  This doesn't seem to be a problem to me: just a reason not to start a fistfight with a Troll (or a jumping contest with a Windling, or an endurance contest with an Ork...).

* I'm less enamored of the one-second combat rounds than I once was.  It's nice to know whether each swing or bullet hit or missed, and I've never seen an autofire mechanic outside of GURPS I didn't loathe, but the system we have makes it difficult to do anything innovative in combat; even tactical maneuvering is very limited.  Round of 3-12 seconds would have less problems in that regard, but I'm not sure what to do with Autofire and very rapid attacks (Even a revolver can be emptied in that long... an M-60 puts out dozens of bullets.)  To clarify, the feature of autofire systems in White Wolf's system, Palladium's system, and Shadowrun is that it combines the whole thing into one attack: for example, in Shadowrun, a rifle does 7S damage with a single shot, but 10D when a burst of three shots is fired.  A pistol does between 7L and 9M.  That creates *weird* effects: if you're wearing 6 points of armor, and have a body of 4, you take one box of damage from the rifle bullet, six from the burst, and none to one from the pistols.  That means three bullets did six times the wounding effect of one bullet, and that one rifle bullet was negligably more armor-penetrating than a pistol bullet.  All sorts of things break.  Still, the simplicity of resolution may be worth the unreality.

GamistExperiments has some ideas I'd like to try out while looking for solutions; as Jesse suggested, running them as dreams of the Dreamers could make a good unifying thread.  It's SpiritQuest time!

BriansLatestMechanicsFixes has the accumulated fixes.