So in the previous post, I lay out my argument as to why pure random tables can be bad.  It really depends on the situation.  If this is something the player either must get, or will really want to get, then you need to be careful.  You should always have in mind “It should take at least X time to do this, and no more than Y time.”  Let’s pick on crafting some more.  Let’s say to get a really good weapon crafted, you need high quality ore.  Normal ore happens most of the time when you harvest, and the HQ ore happens only rarely.  I feel as a designer, I should know roughly how long it will take a player to get something.  Here’s what I know about my game:

• The time cost for harvesting is 30 seconds (this includes harvest time, inventory management, travel between nodes)
• Average player can harvest for 50 times before inventory is full
• Town travel time cost is 2 minutes from this zone

OK, now to play with the numbers.  Let’s choose pure random first.  If I say there’s 1 in 100 chance of getting a rare ore, how long is it before I’m reasonably sure the player will get a rare ore?  Well if the player does the activity 250 times, then there is (statistically) a 92% chance the player will get the rare ore.  That’s pretty darn good, so we’ll leave it at that.  That means 125 minutes of harvesting and 10 minutes of traveling, so 135 minutes to get that rare ore.  That’…a long time.  Is that sword worth over two hours of playing?  In two hours of questing/killing, could a combat oriented character get something better?  Of course 8% of your player base still doesn’t have it, and if your game sold like Skyrim did (3.5 million in two days) that means 280k players have no rare ore to show for the effort.

So let’s look at the problem a little differently.  I want this sword to take no less than 20 minutes, and no more than 30 minutes to complete.  Punching in some numbers, if I have the rare ore drop 1/200 times, in 40 harvests or 20 minutes, 81% of my players will not have gotten the ore.  That sounds reasonable.  Now if I boost the chance of the ore dropping by 10 every harvest after the 40th, by the 60th harvest I have a 100% chance of the ore dropping.  Of course every time the ore is dropped, the drop rate and the harvest counter needs to be reset.

Now why do I feel this is a better way of handling things?  After all it is more complex, and complexity can lead to errors. For me it comes down to control.  As a designer, I should have a strong grip on how long certain actions should take the player.  If I don’t, it becomes harder to design good play experiences.  Not impossible of course, just harder.  If I have a solid understanding of how long it will take a crafter to get a good weapon, I can use that to balance how long it will take a combat oriented character to earn a similar weapon, ensuring that both paths feel equally valid.

Hmm, a little abrupt on the ending, but I really want to end this here.  So I will.

This is something that I’ve been chewing on for a bit.  Is a pure random loot table really all that great?  I’m beginning to think not.  Let’s establish some things first.  Let us assume you have a good random number generator (RNG) to start.  Without a good RNG all of this is moot anyways as your players will figure it out and abuse it.  Second is that you either have a large list of items you want to distribute to your players, you have items that you rarely want to distribute, or both.  If you just have a small list you’re probably fine with pure random.  Third is that you have a large player base, 100k+ at least.  More is better because ‘randomness’ isn’t really visible in small sample sizes.  Now let’s take a look at some numbers.

Assume 10 items in the list, with the ‘ultra rare’ only occurring with a 1/10 chance.  Keep in mind this does not mean you will get the item in 10 drops.  Each time you get the drop (kill, chest, vase, reward, doesn’t matter) there is a 9/10, or a 0.9 chance of not getting it.  So if I want to find the odds of this happening 10 times in a row, I’d do 9/10 * 9/10 * 9/10 and so on ten times, or 0.9^10 which equals 0.349 or 34.9%.  That’s pretty good odds that you won’t get your item in 10 drops.  In 20 drops though this drops to 12.1%, and at 30 drops you’ve plummeted down to 4.2% odds of not getting the special item.  This is why I say a small list can get away with pure random.  Depending on the pacing of course, getting 30 chances at a drop shouldn’t take all that long.

But what happens when that is increased?  Say, we have a really powerful item that we only want to have a 1/100 or 1% chance of dropping?  It is very easy for the mind to think “hmm, that means it will take the player 100 tries to get this item, which will be a bit of effort, but then this is the uber item”.  This is bad thinking.  First let’s extend the example above.  If there’s a 1% chance of getting it, that means there’s a 99% chance of not getting it.  So that 100 tries that takes a while would result in  (99/100)^100 = 36.6% chance of not getting it.  That means over one-third of your player base has not received the item you intended them to receive with that level of effort.  That can cause problems because there’s a strong chance that the unlucky 1/3 of your players are going to feel some strong resentment that they are so unlucky.  They will blame you, the developer of the game, not the RNG.  Oh, and there’s the fact that about 40% of your player base got the item with 1/2 the expected effort.  (99/100)^50 = 60.5% not getting it, so 39.5% getting it.

Honestly, this does cause problems.  They may have changed things, I haven’t played in a long time, but in Age of Conan the crafting skills suffered from this.  At launch there was a flat percentage chance of a rare resource dropping when gathering.  The result of this was some players had an abundance of rare mats, and some players took a very, very long time to get their rare mats.  Which you needed to advance your skills.  Which you had to gather yourself, you couldn’t buy it and turn it in for the quest.  Much frustration over this, as I tended to be in the unlucky group.

OK, so pure random can bad, so what’s the answer?  That shall be in the next post.