Honestly, the numerical rating is only good for a general warning about lousy products. I've read 3-star reviews that convinced me to by the product, and 5-star reviews that are worthless. It's all in how detailed the review is and whether or not you're looking for what the reviewer is describing.

As long as I can move said selection to another game if GOG releases a game I find even better. Or remove my "best game ever" selection completely if I find a game elsewhere that deserves said rating (as it no longer applies to a GOG game).

I like Conan O'Brien's "Clueless Gamer" rating system the best, as it gives you a total freedom to interpret it any way you want. And it is not boring either as he keeps changing it, sometimes using color codes, sometimes something else.

Now I know how they felt when they named West Virginia. They figured, what happens if we expand west? Do we name it Westerner Virginia? Even More West Virginia? Definitely NOT East Virginia?

I generally use 4/5 stars as a good/great distinction. 3 stars means that the game is reasonably playable but probably also forgettable. A game that has serious issues affecting its playability or enjoyment but still has some redeeming features gets 2. So far I haven't had the misfortune of getting anything worthy of a 1, which would pretty much mean it's either completely unplayable or plain awful with no redeeming features. So far it seems that I've handed out an even mix of 3/4/5 ratings and only a couple (like the infamous MoO3) have merited a 2.

Of course, these all assume that the person likes the genre in question, and even if I don't care for the genre much, I'll still try to rate the game from an "If I DID like this genre..." perspective.

It most certainly is a % rating, or at least a non-binary rating. I think what you're saying is right: if people like a game they give it a really good rating. People don't want to have to wring their hands to determine whether one game is better than another. That's why the system is flawed and you end up with all games ending up rated from 3.5 stars and up. Just have a yay or nay rating and you end up with much more precise and informative system.

And no, I really don't care THAT much about all this, but hey, it's something to talk about when I'm bored at work.

So, what's the point in having a rating, that you will never use? This way, you can think that 6 start means a perfect game, and since it's not possible, gog doesn't give you an option to award 6 stars.

For me 5 stars means a game that is as good as it gets, the best games I have ever played. For me Planescape Torment, BG series and many others are definitely 5 star games.

Ghorpm captured it well. I would say a lot of people give 5 stars to anything they enjoyed. I would give 5 stars (or 10 out of 10) to circa 5 games. Not on GOG but overall. 3 is somewhat enjoyable experience. As long as a reviewer bothers to explain her/his opinion, it's fine. It does irritate me sometimes but well, it's just a rating of game.

Professional reviewers giving 10/10 to anything coming from AAA publisher... A Czech magazine has recently solved it well. They changed score system, so it reflects the "I had a great time with it vs. It had some glaring issues but it was fun regardless vs I think-Zzzzz..." instead of "The most epic game ever made vs. I would never buy this average game vs. I hate it" mentality.
I think the scale was:
ZOMG (5/5) - It's one of best among games of this era; it should appeal to a majority of people.
Very good (4/5) - It offers the best of genre. Fans will love it. Other people will probably find it interesting.
Okay (3/5) - Only fans will probably enjoy it. It has some flaws and some nice ideas.
So so (2/5) - Good idea, bad execution. Not very enjoyable.
Meh (1/5) - Painful to play. Bad everything.
It's a useful solution for game mags and it distinguishes it from user's reviews. It's about how big scale the game can reach instead of how much emotionally significant it was for the writer. EDIT: They still tend to give 5/5 to everything with more effective PR campaign but I appreciate the effort.

Film ratings are a little less extreme. 3 stars means I'll enjoy the film if it's a genre I like. 4 1/2 means Oscar.

Game reviewers could learn from this; it might give the scores a little more meaning.

That's more or less my system as well.

I wouldn't necessarily split the ratings evenly among percentages. In my mind it'd be more like the following to reflect what each word means in terms of overall "quality":

0-35% - 1 star
36-55% - 2 stars
56-70% - 3 stars
71-85% - 4 stars
86-100% - 5 stars

You just have to account for the inflated ratings and adjust the scale accordingly. Here's how I read the star ratings on GOG:

5 stars: pretty good
4 stars: decent
3 stars: terrible
2 stars: worst game ever made
1 star: Omerta DLC

Percentages of WHAT? What are you counting?
That's why I went with descriptive terms - they're simple to grasp. Changing one number to another solves pretty much nothing, since we don't GET numbers to begin with. A "63% game" seems to convey SOMETHING about quality, but that only becomes understandable when translated into more human terms.

I didn't make the rating system, which I find not nearly specific enough.
Assuming that 5 stars equals 100%, to me means it's a perfect game and we live in an imperfect world.
Of course explaining it twice won't do diddly for you since you either understood it but didn't agree with it, or you'll never understand it. Reviews are inherently subjective so the system is flawed no matter what, all it can do is show how certain people personally rate something.

Of course it does! But I work with the tools I've been given. 5 stars system is certainly flawed but let's be honest: most of numerical rating systems are.

I'll give you a good example. Age of Wonders 1 and Puddle are games which I rated 5 stars. AoW is my favorite game, I cannot even count how many times I played it. It's the best game for me. Full stop. Puddle on the other hand is a game that I expected to be mediocre at best. And it tuned out I had a blast playing it so the ratio "amount of fun I got" to "what I expected" is enormously high and therefore I decided to give it 5 stars and it seems fair enough for me. But it doesn't mean I think they are equally good, far from that! AoW is much, much better. If you can only get one buy AoW without hesitation! You can argue that in this case I should have given 4 stars to Puddle. But that wouldn't be fair because I cannot think of a single aspect of this game that disappointed me.

Changing the rating system wouldn't be perfect neither. If we had 1-10 system I would give 10 to AoW and 9 to Puddle. Progress? No because I would give 9 to Baldur's Gate and we have exactly the same situation: I would always try to encourage people to buy BG first and Puddle next.

Conclusion: numerical systems are always flawed, at least the way I see them. And therefore the most important is what others have already said: reviews. They are definitely not perfect but by far more informative (at least some of them)
True dat!