About Sawyer's take on encouraging the player to continue playing rather than reloading, I'd support it. That's a mild progress on C&C. Assuming the system is well done, unskilled players will (chose to) be punished, be it even by reduced penalties, whereas till now they virtually didn't exist by the virtue of reloading after a failure.
About generating random numbers and storing them before the game actually needs them, well... AFAIK Japanese game designers have attempted to go that route for quite some time already, and it hasn't prevented people from abusing their systems (Fire Emblem comes to my mind). It's maybe an improvement, but not a solution.
The thing is that modern game designers still haven't understood (or rather refuse to acknowledge) is that some of the randomness (let's call it the "bad" randomness) was historically used to substitute for a failing/absent AI :
- what kind of enemies would the player face once inside the dungeon ? Dice throw.
- how would the enemies deploy ? Dice throw
- how would the enemies fight once confronted to the player ? Dice throw.
- ...
If a game has a good AI, and any modern game should, these dice throws aren't needed.
This "bad" randomness would then be inherent to the quality and challenge level of its AI. On MEDIUM (normal), it could order Orcs to block a passage from the North, have Goblins shoot from a passage to the W, and Kobolds sneak in from a passage to the South for a backstab. On EASY, it would "fail" to organize its troops in anything but a chaotic melee, or fail to send out scouting parties to avoid being ambushed by the player. Hobgoblin shamans would use their spells in a less than optimal fashion, never buffing. On HARD, enemies would know how to fight in formations, and call in for reinforcements if the AI "feels" that the player is about to retreat. Spells would be used to yield maximum hassle and pain for the player.
The result of the battle would also not suffer from "random" outcomes. If the player is skilled, he/she should always win, in the opposite case, they should always lose, till he/she adapts, and learns how to become skilled at the game.
The other kind of randomness would still be needed, to simulate actions (skill checks, attack rolls, etc.) executed by in-game actors
in lieu of the player, for example (since not every video-game is action-oriented). Still, an untrained goblin fledgeling should not be able to kill a highly skilled knight thanks to a lucky crit, a newbie rogue should not be able to pick highly complex locks even if they spent hours clicking on the "retry" button.
One can see that even that kind of "tolerable" randomness should be limited whenever possible, which is what actually happens with the human presence of a "DM" in P&P parties. This then leads to quite predictable results, statistically always the same provided the input ( = player skill ) is the same.
Chess games is a good example, because there is no randomness in the game, no dice throws whatsoever, yet there isn't a single game that looks like another, and the result is almost guaranteed to be favorable to the more skilled on the long run. A vastly rewarding, challenging and fun game to me.
Finally, about games played to be "won", indeed, but games shouldn't be designed to let the player win easily. A good AI helps with that, not any variable amount of randomness here or there.
TL;DR : My guts tell me that the less random a game, the better it is.
