Alright, looked down my list.
Games that do it well, @raged_norm hit it on the head. Hidden roles and objectives really thrive on this. Keyflower, Argent: The Consortium, Archipelago.
Despite Tigris & Euphrates working well without it, still better hidden. Not knowing how important that kingdom or battle is to your opponent is a part of the game.
Also games that don’t really need to be hidden but would be a pain to calculate realtime and I can’t see huge benefits to available scores: Concordia, Great Western Trail, Cascadia.
Games that do split scoring well (known + unknown kicker):
DuneImp tops this list. 80-90% known information and the possibility of holding back intrigue cards is on-theme. I love almost knowing but never being quite sure.
New Fronteirs also 90% known. Goals are well deployed as there are only 8 of them, and a player chooses from nearly half of them when choosing, and there are mechanisms to reveal hidden goals. A good mix of control and uncertainty that makes for interest rather than laziness.
Lords of Waterdeep: You generally kind of know. Most goals are of equal value and equally easy to pursue, so you can look at the score track and count the completed quests and you kind of know. But you don’t know so you get some excitement and endgame theater without any real cost to the game. Ticket to Ride and Carcassonne same reasoning.
Isle of Skye I particularly like. Much of the score is public. Private objectives are visible but uncalculated and not worth calculating. So it is easy to check and see what another player will value or not, and therefore engage with them and strategize, but you’re not going to take the time in-game to specifically figure out if you are four points ahead or four points behind. So you get control of the game (check) and also the feel goods and endgame theater (check). Smart design.
Ark Nova I haven’t thought about much but the open score, endgame trigger, and endgame bonuses seem to meld well.
Split Scoring done poorly:
London 2e is a bookkeeping necessity. Tracking the points in your stacks would be cumbersome, And some scores (money, poverty, loans) fluctuate up and down until the end of the game. So I wouldn’t change it. But the end result is that your score is about 30% track, 30% stack, and 30% endgame. You don’t know where you are during the game and you don’t know what won or lost you the game when you are finished. Would definitely activate realtime scoring in a digital version.
Troyes feels like a gimmick. Your goals make up a large part of your score and I’d rather see what you’re trying to do and interact with it. Not egregious, give this one 50/50.
Settlers of Catan could get the DuneImp argument - there is something to holding a dev card and I don’t know if you have a soldier or a point and how close you are to the end of the game. But you might still get that with the soldier / monopoly / year of plenty uncertainty and the endgame never feels good when more than one card is flopped.
Formosa Tea has been a while. But I remember first timers not understanding how much of their score came in end game and thinking they were doing well or even winning only to realize they had half of the winner’s score. Hidden scoring hid half the game from their strategy. Good game, though, remedied with repeat plays. Not sure it can be fixed due to the bookkeeping and fluctuations side.
Gugong it’s been a while but I remember a huge chunk of points showing up at the end and no real reason for it.
Castles of Burgundy, Hansa Tuetonica, and Grand Austria Hotel have split scoring and I don’t really have an opinion good or bad.
Games where hidden scores are better made public:
Deus just lost my first game and, like Formosa Tea, the lack of score info blinded me to a ton of what was actually happening in the game. No benefit to the hidden score, just a trope.
Puerto Rico and Samurai already discussed. High interaction, scores orient you and help you interact.
Oceans? I’m wondering about this one. It has all the hallmarks of other games that have benefited from the screens coming down. Give it a try?
Thurn & Taxis and Dominion I’m a bit ambivalent. Low interaction games to begin with. The main touchpoint is ending the game while you are ahead, so knowing who is ahead and maybe trying to end the game is useful information. I could go either way here. I enjoyed T&T on BGA better when scores were available.
Then there are any number of games where scores aren’t “hidden” so much as “uncalculated” but I’ve found in digital implementations the game thrives on running calculations:
Barenpark, Agricola, Beyond the Sun, Race/Roll ftG, etc