Been a long time since I roamed this realm. But since I stumbled upon this topic today I'll take a stab at it.
I'm the one who wrote the pdf you're referencing in your post. I will try to respond to your post.
Firstly, glad to see the reason why 1.x ratings aren't compatible with the 2.x engine makes sense now. No wonder people still (I can't even begin to tell you how it boggles my mind) accuses the 2.x engine of being random and/or inferior when they've never actually used it properly.
I believe the ratings you are listing do multiple things at once. Yes, their main impact is on the tendency of the player as in the Jimmy Mann vs. Wayne Gretzky example (if I was drafting a team for my 1979-80 project I would pick Mann every time, but I do love a good goon and that has no bearing on this topic at all so we'll leave that lol), but that's not all they do.
If, for example, we wanted to create a player who will shoot from everywhere, score lots of goals and get next to no assists, we might envision a player like this:
SC 99, PA 45, SK 10
That type of player will shoot the puck very often (around 64% of the time to be specific).
Now we have a player who will shoot the puck a lot, but we also want him to have some skill and accuracy with his shooting so we give him lots of PH (99 in fact).
Put this player on the ice and you might expect an Ovechkin type player. But you will get a scrub. He won't shoot much and he most certainly wont score very much. Why? I can't answer that question with 100% accuracy, but I do know from A LOT of testing that this player needs more of what I call "accumulated skill points". That is, players need to have a fairly high sum of total ratings in those particular categories (PA, SC, SK & PH).
Ovechkin is a very good example of this. Every year when I create my ratings for the 2.1 I gather more stats than I dare confess about each player and run my formulas to generate a first draft of the ratings. I then set up a mock NHL league and manually put together the accurate lines for each team (down to calculating accurate deployment percentages for all the lines *sigh* I'm such a geek) and then I start running season after season, tweaking things as I go.
Ovechkin is an example of a player I typically call a "bastard" during this phase because of his unique skills. Given how the Sim generates assists it's near impossible creating a 50+ goal scorer who will shoot from everywhere but never get any assists. The Sim loves assists off of rebounds, so a player who shoots a lot will inevtiably pick up his fair share of those.
But another issue with Ovechkin goes back to the ratings listed above. If his PA and SK are set too low (which they typically are by default from my formulas) he just won't be very productive, and as a result the Caps as a team aren't very good either. So typically I end up boosting both his PA and SK gradually until his goalscoring reaches a level that is realistic for him. And by that point the Caps will be a very good team.
So in order to make a player "good", he will need to have a certain amount of "skill points".
Another factor, which is very much related to what other people are posting about how players on All-Star calibre teams are struggling to score "as much as they should", is that the simulation engines are very much hierarchical in how scoring is distributed across a team. If a team has one or two players who are vastly more skilled than their teammates, they will pile up a ton of points.
But if you have three lines of first line talent on your roster, scoring will be fairly evenly distributed among them and you may have six or seven guys with 70 points instead of one or two of them hitting 100+.
And there's a key word in all of this rating business:
balance.
Each individual player has to be given ratings that will balance his play in such a way that the performance you get in the Sim is as close to what you'd expect from that player in real life.
You also have to balance every player on the roster so that they will perform both individually
AND as a team as realistically as possible. And this is where a lot of GMs and (perhaps especially) people who create ratings lose themselves in all kinds of erroneous assupmtions about how the Sim actually works.
Balance is everything, both when it comes to creating excellent ratings and building a winning team.
As for your question about making a player "slow" my answer would be this. Speed is not really a factor in the simulation. Once a player decides to skate with the puck, he will do that. Would higher SK make him a bit more elusive when it comes to hitting? Maybe. But once he has decided to skate with the puck he will continue to skate with the puck until he either enters a new zone, gets hit or there's a stoppage in play. Players make one decision with the puck in the offensive zone. They don't receive a pass, skate around with the puck looking for an open teammate or a shot on net.
So will a low SK make a player slow? Not really, there may be an effect on his ability to avoid being hit, but I haven't seen much evidence of that. In my experience there are other things that have a bigger impact on the hittablity of a player.
I will conclude this post by giving all GMs out there a little bit of advice. As we've kept hearing for the past 10 years or so, there is no longer any room in the modern game of hockey for enforcers, goons or pugilists. Well, thankfully, in the world of STHS ratings not only is there still room for them, they are quite possibly the most important part of building a winning team. This post pretty much explains why.