The missing element for all three of those (which are action games) is latency.
With mobile touch-screen controls the game has to somehow register a button-press as soon as a button-press is made.
But it must also never (to a first approximation) register a button-press when a button-press was not requested (e.g. an accidental brush of a finger against the screen).
It's hard enough to do this anyways, but the "input device" does no favors for the game developer either. There's a reason that input devices on consoles and computers have evolved over the years into the shapes they have, they are meant to be easy to hold in a certain position (so that you always know where the buttons are) and perhaps more important, they have tactile features that allow your finger to quickly find and stay near the buttons. It's the reason that buttons and the D-pad are not fully-flush with the shell, even though that might have been more stylish.
Imagine trying to play Super Mario World with a NES-shaped rectangular controller where there are no tactile buttons. This might have worked if you only needed Left, Right, Jump, but even that game required all of the following for the best gameplay: Left/Right, Up/Down (for pipes and to literally scroll the screen up), Jump, Run/Hold/Fireball, Spin Jump, Pause, Drop Stored Item and Shoulder Left/Right (to scroll the screen). And if the input wasn't right you'd either jump into a spike you had meant to run under or accidentally run off a cliff you'd meant to jump past.
The controls on Mario seem complex only because they were so completely intuitive when you were actually playing them but I assure you they were far more difficult to implement than you think. It's not until you start playing platformer games with horrible controls that you can realize the great job done by the development teams at Sega (for Sonic) and Nintendo.
With mobile touch-screen controls the game has to somehow register a button-press as soon as a button-press is made.
But it must also never (to a first approximation) register a button-press when a button-press was not requested (e.g. an accidental brush of a finger against the screen).
It's hard enough to do this anyways, but the "input device" does no favors for the game developer either. There's a reason that input devices on consoles and computers have evolved over the years into the shapes they have, they are meant to be easy to hold in a certain position (so that you always know where the buttons are) and perhaps more important, they have tactile features that allow your finger to quickly find and stay near the buttons. It's the reason that buttons and the D-pad are not fully-flush with the shell, even though that might have been more stylish.
Imagine trying to play Super Mario World with a NES-shaped rectangular controller where there are no tactile buttons. This might have worked if you only needed Left, Right, Jump, but even that game required all of the following for the best gameplay: Left/Right, Up/Down (for pipes and to literally scroll the screen up), Jump, Run/Hold/Fireball, Spin Jump, Pause, Drop Stored Item and Shoulder Left/Right (to scroll the screen). And if the input wasn't right you'd either jump into a spike you had meant to run under or accidentally run off a cliff you'd meant to jump past.
The controls on Mario seem complex only because they were so completely intuitive when you were actually playing them but I assure you they were far more difficult to implement than you think. It's not until you start playing platformer games with horrible controls that you can realize the great job done by the development teams at Sega (for Sonic) and Nintendo.