There have been massive strides forward in all fields relating to game design, but not all of them are equal in terms of what makes a game good. In the end it's the base idea + execution of said idea that carries the game. We'll get to that later, but first I want to explain what "modern standards" can do for a game to make it better, and how many game design choices are now the 'correct one' compared to back then, by comparing a game released 30 years ago with one released just last year.
"Head Over Heels" by Jon Ritman and Bernie Drummond is an Isometric Platformer (a sub-genre that became a thing as it allowed 8-bit platforms to create three-dimensional environments) released in 1987 where you control two dog-like creatures, the eponymous Head and Heels. These cuddly characters are actually spies that were captured while trying to overthrow the evil Emperor Blacktooth and free his subjects, and the game starts with you helping them bust out of jail. The 'gimmick' of HoH over other games of its ilk is that Head and Heels are a team that works best together... but the game starts with them seperated, so your first goal is to unite them. Each character has their own set of abilities, but when united you gain full access to all of them. Once they're united you must choose whether you simply return to base, or try to complete the original mission by reclaiming the crowns of the five worlds under Blacktooth's rule, and thereby successfully overthrowing him. (For an 1980s game, that choice there is a phenominal amount of option left in the hands of the player.)
The above paragraph is the first example of how modern games have moved forward: The exposition is internalized in a modern game, meaning you don't have to read it in a document/manual. In fact, the rest of the manual (which explains the controls and how some of the gameplay aspects of the game work) can also be internalized within a modern-day game, and easily at that. It could be a series of graphical images, a rolling demo, an in-game tutorial, or even just something that's revealed via in-game dialogue or a 'show, don't tell'-esque exposition. In short, tons of methods that are left up to the game's author to choose from. HoH did not have that luxury; it was originally coded on the Sinclair Spectrum, a platform with only 48 kilobytes of memory. There was no room to spare for... well, anything. Whatever part of the game could be shunted onto a piece of paper that came in the game box, was put there without second thought. (Note that some aspects of HoH's gameplay were left to player experimentation. You are not told what things are safe to touch, for example, nor which platforms will crumble once touched. This is the 'trial and error' aspect of player experimentation that has to be used carefully to get it 'right', and HoH doesn't always do so. It's fair to say that the game loves to troll the player.)
'Technological limitations' is the backbone of game design as far back as the late 1990s, but it's how game designers approached it that reveals so much about the evolution of game design. <-- Keep this in mind, as it's more important to this discussion than you'd think.
Anyway, on to the modern game. "Lumo" by Triple Eh? Studios is an Isometric Platformer released in 2016 where you control a human that somehow gets sucked into a Spectrum computer and finds itself in a bizarre, magical realm from which it must escape. The aforementioned exposition (plus any further parts of the story you may come across) are all found within the game itself. Likewise the controls are kept simple enough that Lumo barely explains them, at best you're given a short cutscene that shows what certain things do, like the Light Wand. The rest is left to player experimentation. This keeps the entire gaming experience within the game itself, you don't need to read a booklet, or a 200-page spring-bound 'Survivor's Guide' to play the game (Hello, Fallout 1). That doesn't mean that a modern-day game shouldn't have a manual but it changes the reason for its existence, and is another discussion entirely.
Next up, the controls. Head Over Heels, being an isometric game intended to be played with a single-button joystick, is forced to make some tough choices. First off, it can't... it would actually need a three-button joystick to cover all of the controls, which inevitably means that some keyboard interaction is required. Fortunately the keys are customizable but within limits. Another part is the directions on the joystick. Push 'Up' and you move up to the right and not straight up... this is an isometric game, after all. You get no choice in this matter. Lumo, on the other hand, gives you three options in that regard: 'Up' being up to the right, up to the left, or just straight-up 'Up'. Lumo also offers full control customization, at least for the keyboard. Lumo also only needs two buttons beyond the directional controls: One for jumping and one to toggle the Light Wand. Lumo, like HoH, allows the player to pick up certain things, but the fact that Lumo doesn't need a seperate 'Pick Up/Drop'-key shows clear progress in UI and game design. One final note on UI and controls is comparing the menus in the two games - Lumo has standard-fare menu controls, but HoH has the player one-way-cycling through the available options. Miss the choice you were aiming for, and you have to cycle through them all again. Vital memory-saving technique in 1987, utterly pointless today... which explains why you don't see it anymore.
Head Over Heels has one difficulty setting: Nail-bitingly hard. This is a 1980s game, "Dark Souls" doesn't hold a candle to most of them in terms of difficulty. There are no saveslots or passwords here, but HoH does have a strange form of checkpoint system, but only within the current play session. More than that, HoH operates on a 'Lives' system. Each character starts with eight lives, but more can be found in-game via hard-to-reach power-ups. If either character loses all of their lives, it's Game Over. Lumo offers two difficulty settings: 'Adventure' or 'Old-School'. The former has infinite lives, autosaving and a map (which isn't as useful as you'd think). The latter has no saving, no map and a limited number of lives. As someone who has played 1980s games when they were new, I only recommend 'Old-School' difficulty to people that are either masochistic, autistic, or both. Modern-day players, however, should try it out as a teaching lesson.
Now, I could go on in comparing these two games here, but I think I've made my point by now: The primary difference between games of Then and games of Now, is the amount of options available to the developer, and how they use them. You wanted to program a game on a Spectrum, for example, you needed to know the system's limits and program the tightest code imaginable, with next to no margin for error. This cramped environment, strangely enough, gave ample room for experimentation - developers trying various things, just to see what they could squeeze into their games, and as expected not all of it came out right. But slowly and surely a few basic 'rules' did emerge, primarily among them that players, like the developers, wanted all kinds of choices in their games. In short, progress was made.
Today's developer is spoilt for choice; he can pick any perspective, any form of exposition delivery, any control scheme and literally every aspect of the gameplay, it's all up to him and he faces no barriers but himself, he is not held back by any technological limitations. With that in mind, look at the games we're getting today. The majority of them are re-hashes of older games with a few added choices, but not necessarily the right ones. Lumo for example, is not just an Isometric Platformer, it also has mini-games based on games such as "Zaxxon", "Nebulus", "Spindizzy" and even "Horace Goes Skiing", yet all made to fit within the framework of the core game. It detracts from the primary experience of the game, but an argument can be made that a little change of pace can be good. Again, that's another discussion.
We're also getting developers that are experimenting with their tools, but most of them are actually just re-inventing the wheel because they haven't studied games older than the ones they played as a kid, and therefore don't realize that It's All Been Done Already. Worse yet, the AAA games have abandoned all forms of experimentation or imagination - to the point that 'AAA-game' is becoming a game genre in and of itself!
Finally we have the players themselves. They've become accustomed to having choices in their game, that game decisions that favor the player are the norm, and finally that the game includes everything needed to play it - that last part is probably the single-biggest factor that divides 'old' games from 'new' games. When a player is used to having an auto-map in his game (for example) the reaction to its absence in another game will be negative. This is the problem with introducing modern gamers to old games (even the good ones) - it's the 'system shock' of having to play a game that is so constrained, but also so 'all over the place' ("What do you mean, I must read this paper first before playing?"). Another point is keeping the player engaged with the game - an overly repetetive game will not keep a player's interest. This was the case back then, and it's even more important today. Back then we put up with because we were starved for choice. Now we're drowning in it, so we choose not to play those games.