In regards to the physics, KSP is fairly realistic, other than it not modeling n-body physics (which isn't really relevant in scope of orbiting Earth/Kerbin).
In regards to the engineering, KSP makes its parts much stronger than the real life components. The tradeoff here is that when something finally breaks in KSP, it explodes and disappears. Engines are also more powerful, and fuel is (slighly) less heavy. This is all done to "round the numbers" in favor of the player to make sure that the limited part set can get most reasonable jobs done.
In regards to the controls, KSP's control system is not as fine-grained as a real world spacecraft. Given a keyboard and mouse and moslty eyeballing the thrust gauge, there is only so much precision you can reasonably expect from a player.
In regards to the ascent profile, due to the stronger craft and less fine-grained controls, KSP spacecraft are able to ascend using sharp(er) turns than is realistically possible in real life. In KSP, a simple ascent profile is to go straight up for about 10km, and then make a 45° turn and start building horizontal speed.
In real life, a rocket won't be as easy to turn (e.g. a 45° turn) due to the stress that puts on the vehicle; most importantly it's the incoming wind from still moving straight upwards (due to inertia) that is going to start beating down on the side hull (which you expose by orienting your craft away from its prograde vector).
To avoid that, the best approach is to orient your spacecraft as close to its prograde vector as you can, because that minimizes exposure of the less aerodynamic side hull compared to the aerodynamic nose.
This means that in reality, spacecraft tend to make their 90° (from vertical to horizontal) turn slow and steady (i.e. pretty much immediately after launch). And this is actually not that hard to do: even a minor (but consistent) deviation over a long period can do most of the work for you and doesn't require active steering (other than necessary corrections due to imperfections in the initial setup).
In KSP's defense, there is a boundary of realism that you should avoid when it negatively impacts gameplay. The level of precision required (for both thrust and attitude adjustment) to achieve a "real" ascent profile is quite high, requires a lot of calculation, and any imperfection in the simulation would render any calculation impossible to get accurate anyway. So rather than risking the player's calculated approach being useless after all that effort, KSP opted to simplify things slightly to keep things much more intuitive.
Most rockets can be flown by the seat of your pants (and a basic grasp on atmospheric friction and that an orbit is achieved through horizontal velocity), and this is a lot more fun for most people. KSP doesn't force a player to plan their entire mission down to every control input before ever launching (NASA pretty much does, other than necessary corrections of course). KSP gives the player the freedom to plan a little, and then experience the mission (and only figure out some things such as the interplanetery maneuver at that stage, not before launching).
This is also why (unmodded) KSP doesn't implement food/oxygen logistics. Not being bound by that means that there is no real drawback to orbiting a few more times before finally doing your Hohmann transfer (e.g. waiting until the Mun is in the right spot).
NASA has years to perfect their approach and calculate every little detail about the mission (e.g. the aerodynamic impact of the angling of a single uncovered solar panel, or what a % of a degree of inclination can do to affect the ascent path), but requiring a player to do so for KSP would lead to a ridiculous amount of preplanning and testing before every achieving anything. That's not fun (except for a few die hard fans) and it's simply not a good design design (gameplay wise).