That's not a 51% out of 100% given as a review - that's 51% of people given two choices, recommend or not, choosing to recommend the game.
...and that is even with it being generally true that people are more probable to speak up to trash on something than to praise it, and with reviews coming from people that chose not to recommended the game including such contradictory information as having played the game for 127 hours and saying "I didn't like it."
Now, if you knew that the 51% I was talking about was 51% of nearly 2000 people saying "yes, get this game" and you still wouldn't buy anything under 70%... that's a different story entirely, especially since there is no way at all to guarantee that you are getting info from people with preferences similar to yours (i.e. there might be 2000 people that have played the game and chose to review it, but not all 2000 like what you like so some of the "don't get this game" reviews might actually be listing things you like and the reviewer doesn't as reasons to avoid the game). I mean, I know that 70% of my friends, let alone a group of complete strangers with nothing in common other than the broad quality of "likes videogames", don't agree with me about what is or isn't a good video game even half of the time.
What you seem to miss is that there's a built in sampling "bias" for these things, especially when it comes to multiplayer games. I use bias in quotation marks because it has special meaning in statistics that differ from how it's used elsewhere. You have to take into account the greater body of results when understanding any metric. At 51% approval Sword Coast Legends is at the bottom 12% of games on Steam. It's kind of like how Hoard of the Dragon Queen has a 52.5% at ENWorld presently. That puts HotDQ in the bottom 13% for D&D products here. The point is that it's not especially common for a steam product to acquire a worse rating than what Sword Coast Legends managed. Or even more to the point, it's very easy to find a product that people like more.