You're right, these things are all relative. This is from later in the same article, which I think explains what Lew means by a "silly, totally unbelievable game".
"This idea of self-consistency or integration must be pursued further. One of the most destructive notions I've encountered in D&D is the belief that 'anything goes'. This is fine for a pick-up or silly-fun game, but contributes an air of unreality and recklessness which can be fatal to a campaign, and which in any case is offensive to many players. Inevitably, an 'anything goes' campaign tends to be one in which player skill counts for little, for two reasons. First, players have no foundation to base decisions on; never knowing what to expect, they cannot plan a rational response. Second, the 'anything goes' game tends to be dominated by dice rolls or referee manipulation. A great deal usually depends on the saving throws of characters. For example, one of the favourite ploys of the 'anything goes' referee is to devise panels of buttons or decks of cards similar to a Deck of Many Things, often involving more far-reaching changes. Players push buttons or pick cards and great things occur. Players seldom do much to earn the rewards or penalties - the cards are easy to find, and the dice determine results.
One may protest that the skillful player can avoid picking from the card deck, or fooling with the lever or button, and so on. Unfortunately, the structure of this kind of game is such that, if a player (not a character) wants to get ahead, he must take his chances. The reasoning is simple. A player can always roll new characters. In a luck-dominated game, even if half the time a player's character is seriously harmed, the rest of the time he benefits to the same degree or more. Consequently, the player who chooses not to take the ridiculous risks may die less often, but his characters will often be mediocre compared to those who dared and were lucky. The player who trusts to fate will lose many characters, but his other characters will prosper. In other words, the 'law of averages' works against the cautious player. The key is that the character run by the player does not have to act rationally because it has no separate existence. In many cases, only an insane person would accept the risks involved in cards, buttons, and levers. It's too much like Russian Roulette. But the player isn't the one who may die or be maimed; in fact, if his character is crippled, he can easily get him killed and start a new one. Thus this form of the game forces players to depend on luck and at the same time contributes an air of unreality to the entire proceeding. Even fantastic fiction, despite the name, possesses an internal self-consistency, and the characters in fantasy fiction usually act as rational, though brave, people. In Dungeons and Dragons, if the campaign is not designed correctly it becomes unbelievable, for a D&D player may, along with the fiction reader, say 'I don't believe men would do this'. Each referee must ask himself as he sets up his campaign what rules and items would seem believable if he read about them in a fantasy novel.
Even in a fantasy game, moderation and self-discipline are virtues necessary to top refereeing. While campaigns may be run on other bases, I believe that a skill-game campaign is likely to satisfy people more in the long run. Some people prefer luck and passivity, but they are seldom game players. If you feel a need to get drunk and/or stoned, however, try lottery D&D the similarities are surprising."