Some companies will have spectacular failures as well as spectacular successes. Successful companies are able to learn from both and iterate.
Unreal and Unreal Tournament were absolute block busters that paved the way for Fortnite, just as Paragon did. You wouldn't have arrived at Fortnite without first going through the ups and downs that preceded it.
By your logic, a company can only be considered successful if they never make mistakes, which is a severely flawed position.
I don't really consider Unreal and UT because, while they were definitely successful and I personally love UT, they're the product of a different time. The last truly successful UT was in 2004 (UT3 in 2007 didn't have quite the same long-lasting appeal and even that was 12 years ago!). There's also Gears of War that was quite popular on consoles but again, a completely different formula and nothing remotely close to what Fortnite is achieving today.
Gears too seemed to lead the pack on pulling together various gaming trends ahead of a lot of its competition. Cover-based Shooter wasn't its own "genre" until Gears (even admitting that fad may have passed, Gears was pretty central to the fad). Gears was also ahead of the pack in adding things like tower-defense elements to FPS/3PS multiplayer (Horde mode).
That's before you take into account their pre-Unreal successes, too. Epic had some big hits in the Shareware era that are perhaps more interesting pivots simply because of how much tougher distribution and sales were in that era.
Anyway, Epic seems to have a sample size greater that "striking gold once", for whatever it is worth.
I mean, that literally describes every innovation. Luck is always part of the equation, you're just assuming luck was a bigger part of the equation than most people do, and I'd argue you have little evidence to support that position.