Um...to duplicate completely the circumstances under which the pseudorandom function returns a value both a seed and another number indicating how much has been pulled already from the function, IIRC. So then you''ll need both, and if another number is pulled, that number''s value during the bug is lost. I suppose it would be possible to retain some indication of this number. Yeah, that''s probably a good idea. But it still wouldn''t cover fatal errors.

And I think I covered that a huge portion would not be subject to bugs in the use of the random numbers, so the number of bugs relating to this would be small. Oh well.

I think you''re confusing the word "uncertain" with "inscrutable". Some probabilities can be deduced, but one cannot be certain of the outcome of anything that involved a random variable.

Oh, and as to the analogy of reducing complex systems to probabilities, it''s probably a good idea to reduce any extremely complex system to probabilities because if it''s beyond the computer''s mind to predict, it''s also beyond the player''s mind to predict. So then, let''s all agree that we should reduce coin flips, rolls of a die, where the ball falls in roulette, and deck shuffling to randomization. And now back to the parts of a game that actually matter to a player

I think it''s probably a good idea to think very hard about trying to introduce a variable that is very hard to determine by a player instead of one the is completely random. Good examples would be adding a moving target before random air velocities during the flight of a bullet, or an opposing player before adding random laser beams that need to be dodged.

Randomness is most appropriate when all or nearly all information is given to the player, although you''ve probably got more serious design problems in that case. And even if there are no other problems, it may not even be appropriate then. We look to the Game Design apex, Scorched Earth, for guidance. Scorched earth is completely deterministic. Not only is all information guided completely by variables that are not random(except at the beginning), but all the information is known the player as well. Wind is random, but easily deducible to the player. Oddly enough, everyone I know plays with wind disabled--it''s no fun when you don''t notice the wind changes and then shoot yourself because it blows your death''s head back at you. That still makes for one of the most legendary games of all time.(it being free and uncrippled didn''t hurt either)

Two games come to my mind when thinking about randomness in games, Risk and Axis and Allies. Risk is an example of too much randomness, battles can swing heavily one way or the other based on the dice and not the units involved or any particular strategy used by one or more players. The game leaves too much up to chance, the strategy element is severely limited.

Axis and Allies on the other hand is a game in which randomness is properly curbed. True, the game still revolves around dice rolling, and inexperienced players will say the game relys heavily on luck, but to experienced players, strategy plays a much larger role than dice. You play the game by the averages, counting on the dice to level themselves out, but there is still those battles that go a little bit awry or a little better than you planned, all this must be accounted for in your strategy. Proper purchasing (unit building) and deployment are the key factors of the game, no the luck.


Desert Fox

Gee, thanks for the pointless and snide comments, as well as defining the law of averages which I already alluded to...rather than "clear up misconceptions" all you''ve done is act like an ass...


Let me ask you a simple question: which is more consitent, the same thing happening every time, or things happening is roughly equal ratios from over the long run?

For example, let''s say I play a game where one out of every 100 times I click on a unit it blows up for no apparent reason. Is that consistent behavior?

Of course, you could reply yes, the vehicle consitently blows up 1/100 times...

Similarly, say you launch a program that worked fine yesterday, but crashes today. Is that consitent behavior? Hey, sure it is! Obviously you did something in between, and if you believe in determinism you must believe that the crashing behavior is wholly consistent with what you did. One time bugs? Those are consistent behavior!

Once every thousand year bugs happen on average once every thousand years...looks like consitency to me!

The point here is if the player does what appears to be the exact same thing, they generally expect to get the same result. If I tell a unit to go someplace it should probably go there the same way it has always gone there in the past, unless there is some obvious difference that is apparent to the player. (Like th old path is blocked, or more dangerous, or whatever)

Basically, there had better be a rational explanation for what is happening. If units randomly get lost, the only real rational explanation is that the units are idiots. Which could work in some settings, but in an RTS pathfinding is bad enough already. If a unit gets lost, was it dumb? Could it not read a map? Or was it a bug, or bad pathfinding? Note that this could be partially dealt with just by doing something like putting a thought balloon question mark over the unit''s head. The idea is that the player will at least understand what is happening and why. Similarly, if unit''s flee to to low morale a thought ballon exclamation mark could indicate they are scared and not just that you accidentally hit the scatter hot-key.

The player must understand *what* is happening and *why* it is happening. So in the case of low morale, they should be able to tell *that* the units are fleeing and *why* they are fleeing. If your morale rules are that there is a chance they will flee as soon as they hear gunfire that will appear pretty random to the player, unless they understand what is happening and why. (Like, the units pause for a second with a sad face and an ear in a thought balloon, indicating they heard a bad sound. And the player knows these units are very cautious)

You probably want to avoid really low percentages. Like saying there is a 1% chance units will run from a noise is dumb. It won''t happen often enough to appear consitent, rather it will seem like an unexplainable bug or really really terrible luck. Basically at 1% it will be something you don''t plan for, but a real annoyance when it does occur. If you bump that up to say 10% now it is something the player will learn to adjust to, and will actually be less annoying in some ways, because they have learned to plan for it.

I would also point out that in something like a table top wargame turns take a while and all the players generally know all the rules. In an RTS game you might not have time to carefully consider everything that is happening, and you probably won''t know the exact rules or be able to properly do the calculations to apply them.

JM

quote: Original post by Floppy
blah blah

quote: Original post by Flarelocke
And I think I covered that a huge portion would not be subject to bugs in the use of the random numbers, so the number of bugs relating to this would be small. Oh well.

You did say most bugs would be limited to particular types, but you said so only in passing. Everything else you said represents an exaggerated opinion of the problems randomness might pose. Hard to find bugs related to randomness are usually not something you have to worry about if you understand when and how random numbers are used in your game. You keep talking about fatal and hard-to-reproduce bugs, as if a person writing a game that uses random variables wouldn't bother to test all the possible outcomes (usually not that many).

quote: I think you're confusing the word "uncertain" with "inscrutable". Some probabilities can be deduced, but one cannot be certain of the outcome of anything that involved a random variable.

I presume you're talking to Floppy here. His use of the word uncertain was quite appropriate. The outcome of a random experiment is uncertain but will fall within expected parameters (particularly evident after many experiments or when something is significantly more likely than something else).

quote: So then, let's all agree that we should reduce coin flips, rolls of a die, where the ball falls in roulette, and deck shuffling to randomization. And now back to the parts of a game that actually matter to a player

Such a statement illustrates you're either unaware or intentionally ignoring the usefulness of random variables in the construction of models which predict the behavior of real-life systems.

quote: I think it's probably a good idea to think very hard about trying to introduce a variable that is very hard to determine by a player instead of one the is completely random.

Why is anything either "completely random" or completely determined? Do you not allow for random outcomes that fall within likely but uncertain parameters?

quote: Randomness is most appropriate when all or nearly all information is given to the player, although you've probably got more serious design problems in that case. And even if there are no other problems, it may not even be appropriate then.

I'm not sure what you mean by "serious design problems". In any case, the player doesn't need all the information, he only needs to understand when and how a random variable manifests itself.



Edited by - chronos on February 27, 2002 8:07:21 PM

I only really meant to comment about bugginess in passing in the first place.

quote: Such a statement illustrates you''re either unaware or intentionally ignoring the usefulness of random variables in the construction of models which predict the behavior of real-life systems.

I probably am, but we''re not predicting the behavior of real-life systems, we''re creating systems from the ground up, in which case prediction is not necessary at all.

quote: Why is anything either "completely random" or completely determined? Do you not allow for random outcomes that fall within likely but uncertain parameters?

I can''t find what I was replying to at the moment, so I feel like I''m answering a question nobody asked. Maybe there was an edit in there somewhere that I was addressing.

quote: Why is anything either "completely random" or completely determined? Do you not allow for random outcomes that fall within likely but uncertain parameters?

Yeah, I allow for them, but in most cases, why bother? Why not just adjust the numbers to account for them. That has the advantage of consistency. Randomness of chaos. I''d rather be immersed in the former.

quote: Gee, thanks for the pointless and snide comments, as well as defining the law of averages which I already alluded to...rather than "clear up misconceptions" all you''ve done is act like an ass...

I''m not sure who you''re talking to, but if you don''t want pointless and snide comments, post as something other than Anonymous, because it''s hard for posters to differentiate between one anonymous and another.

quote: Original post by Flarelocke
I only really meant to comment about bugginess in passing in the first place.

In any case, the issue was discussed more than just in passing. You represented it time and time again as something that would make debugging very difficult.

quote: I probably am, but we''re not predicting the behavior of real-life systems, we''re creating systems from the ground up, in which case prediction is not necessary at all.

That''s hardly the point. The statement I responded to implied that the usefulness of randomness was limited to things like flipping coins and playing roulette. I countered by saying that random variables are useful in describing the behavior of things as complicated as real-life systems. The logical conclusion is that randomness is a good way to represent the behavior of many kinds of systems, including some of those you build "from the ground up".

quote: Yeah, I allow for them, but in most cases, why bother? Why not just adjust the numbers to account for them.

How do you adjust the numbers to account for mutually exclusive outcomes? You can''t have them all at once, so you have to arrive at a particular outcome according to some rule. One way to accomplish this is to determine the outcome at random according to some statistical model.

quote: That has the advantage of consistency.

You say that as if random outcomes did not exhibit consistency. If the outcomes are consistent with player expectations, that''s good enough for me.

quote: Original post by Flarelocke

Gee, thanks for the pointless and snide comments, as well as defining the law of averages which I already alluded to…rather than "clear up misconceptions" all you''ve done is act like an ass…

I''m not sure who you''re talking to, but if you don''t want pointless and snide comments, post as something other than Anonymous, because it''s hard for posters to differentiate between one anonymous and another.

I was talking to Floppy, hence the quote I included from Floppy at the end of my post.

I don''t see what being an AP has to do with being snide. Unless some AP was snide and the person is returning the favor but misdirecting, which is not the case here.

JM

Personally I think true randomness has no place in a game.
As an example, in the game "Majesty," on some levels there are lightning bolts that randomly strike the ground; if you''re unlucky a bolt will strike your level 50 wizard and kill them, effectively ending the level for you. Is this fun? Not hardly. However "random" elements that are predicatable on at least some level could definitely enhance gameplay. Weather is one example. Consider a war game set in the seventeen hundreds (before weather prediction). The weather for any given day could not be predicted precisely ahead of time (say the day before as battle plans are being drawn up), but the player should be able to make educated guesses based on the season, the weather of the previous days, and the current atmospheric conditions. This way the weather could become a gameplay element itself; just as in real life, a good general must be able to use weather to their advantage. Of course this would require a decent, deterministic weather model (meaning it should follow rules with a sprinkling of true randomness thrown in so that just as in real life absolute prediction is impossible).

Take Solitaire.

Imagine trying to play solitaire if the cards always came out A,1,2,3,4,5,6,7,8,9,10,J,Q,K.

On RTS games, if there were no random element, then the foes would become way too predictable. They would be "Pattern" games instead of "Strategy" games.

Random elements are critical to alot of games.

If you have a head to head game, it may be desireable to remove random elements that can give an unfair advantage.

Of course, you can add a "Catchup" feature that artifically keeps the completition close until near the end.

Example: A Car Race Game. If I am behind the leader, my car goes 5% faster than his. When I get close to him, then we are equal (but now he can see me in his mirror)

I believe that randomness does have its place in many games.

Is our existance a life-long virtual reality simulation from another realm?