The more a game is player led, the more the players are responsible for 'failure'. This includes 'failure by dice' - the players are making the decisions re: action and risk.
The more a game is referee led, the more the referee is responsible for 'failure'. This includes 'failure by dice' - the referee has set the obstacles in the path of the players.
I've been thinking about this for a while, running the Crown of Kings - a referee directed 'quest' - reading around the edges of a prospective Traveller sandbox campaign (hence, 'referee') - with Book 0 An Introduction to Traveller [get Classic Traveller free here] being a fine introduction to roleplaying games in general - which has included re-reading the advice of running and playing a sandbox game in Stars Without Number. I've also been playing LA Noire, an absolutely rubbish game, in that it is almost entirely linear, with success almost guaranteed, and the only penalty for failure is to play the exact same component of the case again.
And all this got me thinking about the reaction to the my self-mocking rant on lauding the Pathetic Aesthetic. In some places, it was discussed as if I was advocating dick-refereeing - placing characters in unwinnable situations by fiat, or delighting in killing them off. If you are responsible for the failure of your character, in the process of making meaningful decisions on the part of your character, then failure has been part of play. It is the consequence of your characters interaction with the 'world'. However, if the majority of your play has been largely referee led, then it is likely at least some of the times your characters' failures were the responsibility of the referee, which is likely to leave a far more sour taste in the mouth. It will feel like play undone, or play unrealized.
*Note, when I write 'player led', I'm not talking about collaborative storytelling, or anything like that - I'm talking about players having a significant degree of freedom in determining the actions of their characters.