Carrot Farming : Complementary System For Pure RNG Loot Generation

Recremen

Farming is not only the bedrock of society, it is the bedrock of typical RPG content, perhaps most especially for MMORPGs. ESO has, at least for me, generally struck just the right balance for the average time needed to farm something, whether it be materials for crafting gear or the gear itself via zone and dungeon drops. Obviously most of us would love to skip farming altogether and just get rewards for completion of content, but from an MMO design perspective it's a bit of a necessary evil so that people keep coming back. Without repeatable content, the game would quickly die. They've even improved some of the grinds to more frequently give some kind of reward, even if it's not the best trait or armor weight for you. Still, are there opportunities to improve it further? The answer is: very yes!!

At its core, RNG loot systems are designed to encourage players to repeat the content. While some players may get what they want right away, others must try many many times, and this keeps people coming back to zones, queuing up for dungeons, and generally keeping the game looking busy. It also stalls for time while the devs churn out new content, new DLC, new costumes, etc. But not all RNG systems are the same! My two goals are to convince you that a pure RNG system is not the ideal solution, and to offer a complementary, customizable system that still fits the needs of the devs and does NOT rely on a token system.

“Wait... WHAT?” you shout indignantly. “No TOKENS???”



Yep, no tokens. Here's why!

While token systems have a lot of potential, they come equipped with some severe limitations that are not appropriate for every loot opportunity. The most important is that they do not encourage repeating content. An average person only needs a few pieces of gear from any given dungeon or zone. They will only have to grind up exactly enough tokens to afford the pieces they want, and then they'll quit. The devs could just up the cost until it fits with how much they want people to farm on average, but then that unduly punishes people who want a lot of gear from the same dungeon or zone. No good! Token systems are good for games where you are shoehorned into one specific item, and not so good for games like ESO that allow any weapon and any armor.

“Okay, so wait, isn't pure RNG just as bad?”

No!! It is actually much worse. The problem with pure RNG systems is that you are never actually guaranteed to get a particular item you are looking for, no matter how much time you spend farming for it. While the exact details depend on drop rates and the number of attempts to get the gear, it's an unavoidable mathematical fact that your odds never converge to 100%. Since content takes time to complete and people have lives to live, this creates the very real scenario that even some of our best farmers don't wind up with the gear they're after. I'm sure we've all seen enough threads along the lines of “This is my 147th VMA run, let's see if I get a sharpened inferno staff... nope” to last a lifetime. This is frankly not the kind of game I want to be playing, and I am sure the vast majority of people agree. People should, at some point, be rewarding with what the gear they're going for when they are consistently and successfully completing the associated content. We can call it an ethical problem or we can just say “I don't like the taste of this bologna”, either way it's obviously that something should be changing.

So what, then, can we do to make a better system? Nothing. We aren't the devs, silly goose-quack!! But we can still suggest some complementary designs.

I like the recent change to VMA weapons and monster helm drops in that you are guaranteed to get one every time you complete the content. This really was a stroke of genius that gets people at least in a position where they can try some kind of build out based on the gear from the content they're doing, while still encouraging repeat visits to get That Perfect Trait. Having to roll once just to get any gear and then roll again to try and get something good was one of the worst doubling ups I've seen. My suggestion is to continue in this fashion : by guaranteeing that you will eventually get every piece of gear from content after a certain number of playthroughs.

That probably sounds ludicrous at first, but that's just because it's hard to express the exact nuances of the system I'm proposing in a single sentence. I am not saying that people should play through some dungeon x number of times and then suddenly get a chest full of All The Things. The one-per-run rule is still very good. The randomness of it is still very good. But we should (efficiently) keep track of what people are getting from the loot table so that they don't get a bunch of repeats on the same character.

Essentially, when someone completes content and gets a roll on the loot table, you cross off the item they get from that loot table for that character. Then when they go to complete the content again, if they would have gotten the same item they instead get the next one on the list (or however you want to handle it, there's plenty of room for dev discretion on that implementation detail). If there are no more items left on the list, the tracker gets cleared and they have a chance at everything all over again! There would of course be separate trackers for each zone, trial, and dungeon, and each character on an account should have their own set of trackers.

To efficiently keep track of loot history takes an exceedingly small amount of memory, which is important considering this system must scale across all accounts, and be tracked separately for each character on that account. The key to remember is that it only takes a single bit to track a particular item. A value of 0 shows that the item has not yet been looted, and a value of 1 means it has. Each item is associated with a unique index on the data structure, so you end up keeping track of the bare minimum amount of information per character.

Let us take the infamous Maelstrom Arena weapons, for example. There are 12 weapons which may come in 9 traits (obviously not all of the traits drop in Maelstrom, this is just to illustrate the point). That's 108 combinations, or 108 bits to keep track of, or 14 bytes. So for the price of two 64-bit system integers you can track all of Maelstrom's signature weapons, and then some. Scale this up to 12 potential alts for an account and you can keep track of that player's whole potential Maelstrom profile using less information than it takes to keep track of this sentence I just typed. (168 bytes versus 205)

If we multiply that by our 7 million players, we get just a little bit over a single gigabyte. If we multiply that by the number of zones, sets per zone, etc. we're going to wind up with less than a single terabyte of storage needed for keeping track of all current and future loot tables. Talk about efficiency!! If y'all are really hurting for a spare terabyte SSD I can hook you up. It's for the good of the game and all. ;-) Obviously, this memory can also be dynamically allocated if you're especially in the mood for counting pennies, but in general I'd wager it's better to ensure you have the space required in a worst-case scenario.



So what does this tracking of loot accomplish? Everything, darling, everything.

Players are guaranteed to eventually get the gear they want if they're willing to play the content. It's still a roll on the table every time, so it's still totally random what you're getting, you're just picking from a smaller and smaller table every time. Thus you'll still have some people getting it instantly and some people lagging behind, but now you're not penalized for a butterfly flapping its wings in Taiwan and setting off a chain reaction of events that leads you to getting your seventh Defending Maelstrom Bow. You can also still get repeats if you want, you just need to keep going until your tracker resets or work on it with a different character.

This also encourages several good behaviors for the players. It should make people more likely to repeat content because they won't be discouraged by the chance for getting nothing they need. It will reward playing on multiple characters so you can get repeats of good items faster. It will also foster a sense of healthy peer pressure for guildies who have heretofore been reticent to do gear grinding because the odds are just so darn abysmal. YES I AM TALKING TO YOU. I HOPE YOU ARE READING THIS AND I HOPE YOU KNOW IT IS ABOUT YOU, SPECIFICALLY.

you.




It is also a system that lets the devs play around a lot with the specifics. For example, let's say they want to, on average, delay the attainment of the most desirable traits so that people play for longer. No problem! You just stuff your loot table with more instances of the non-endgame traits until you reach the desired odds. They'll still all get dropped from the table as they get looted, and it will still take at most (by our example) 108 playthroughs to get a single copy of a specific gear piece (or more or less depending on the number of potential drops for that dungeon or zone), but honestly if you don't think people should have what they came for by 108 successful runs through the content then you are just a horrible people!!

There are also some concerns related to this system that ought to be addressed. In the case of one-handed weapons, for example, it is imperative that they drop in pairs. This sounds cheaty-face until you remember that all two-handed weapon users get their goods in one go, and it is only equitable to do the same for one-handed/dual-wield weapons. Otherwise they'd have to keep going all the way to the tracker reset just to get another chance at finishing their set! Not cool!! Another concern relates to the new item trading system for dungeon drops. The check should be done only when the loot drops and should not take into account anything received in a trade. Otherwise it could be abused if someone happened to have high odds for getting a popular item and they keep trading it away for gold or ERP favors. Finally, there's the economic issue related to rare overland zone drops, like Sharpened Spriggans Swords. Everyone being guaranteed to eventually get these may cause the prices to drop dramatically. To that, I shrug. It's really the only possible downside for players, but it's more good for everyone than it is bad for the people trying to make bank.

So yeah, I'm hoping we can see a system like this roll out eventually! It's a great, customizable way to tailor the loot system while still being fair to players and exorcising all the lingering math demons that conspire to make us miserable. If you have any questions or concerns, please post them, I'd love to see what people think of this!

Xerton

@Recremen great post! Want to know the big problem of it all?
It actually makes sense and would make playing the game more fun and enjoyable.
So it won't happen.

stevepdodson_ESO888

all RNG systems really do is make developers lazy and we don't see any new content for ages...just re-hash old stuff with a few whistles and bells added and call it a "seasonal event"

think of Pavlov or Skinner and you have it in a nutbag (or is that nutshell)...it always feels like a kick in the nutbag anyway

Gandrhulf_Harbard

There are three main problems with ALL RNG systems IMO:

1) They do not reward skill.
2) They do not reward effort.
3) They are lazy game design elements.

What other field of endeavour would you engage with if the payoff for your time and effort was entirely random?

The honest answer is none.

If you play tennis (or indeed any other sport)you would soon quit if no matter your skill and effort the result of every point played was determined by the roll of a dice (especially the roll of a rigged dice - which is what RNG loot tables are).
You would not go to work for a random pay-packet.
You would not pay weekly for a random grocery delivery.

Random has two huge advantages to game developers though: a) it require no intellect or imagination to come up with an RNG system, b) because of that RNG systems are invariably the cheapest option.

RNG Loot Tables is an element of game design that should have been consigned to history a decade ago.

All The Best

Dr_Resilient

The RNG right now is awful and frustating. The real TES players don't want very easy content, but at same time, don't want to run 231 times one dungeon to get a weapon with the right trait. Let's make this whole thing fair..

idk

@Recremen

Adding my cat pics doesn't make reading the extremely long explanation easier.

Not being rude or mean but I don't by Zos will read it due to length with lots of fluff. Try to narrow it down. Get to the point with what your saying and suggesting.

AdamBourke

This makes sense, for the most part. I would argue two points though:

The first is that dual wielding should not drop at the same time. There are advantages to DW and disadvantages, and this is one of them. I agree they shouldn't have to play through until the loot table reset - but this can be counteracted by having all weapons in the "10101010" twice (yes - even 2H weapons!).

The second is that I think that you should not be guaranteed anything. there should be a 90% chance of hitting the new loot table, and 10% chance of getting a completely random item. I can't really explain why I think this though.

To expand on the idea - there could then be achievements for collecting everything "Maelstrom Armour Collector", "Hel Ra Weapon Collecter" etc. This might encourage people to run them until the 108th time...

Reorx_Holybeard

Great writeup and interesting ideas. I think it would work great for limited/special sets like Maelstrom/Master but I believe you are vastly underestimating the total number of unique set items and the resulting effect it would have.

From here I get a total of 263 sets comprised of 55,508 different items which is a lot bigger than your 1000 item comparison if I read you right. Probably not a completely unreasonable number to be handled in a system like ESO but not a trivially small number either.

Personally, I think a lot of the loot issues can be partially solved by another trait re-balancing and adjusting exactly what traits drop where. The bonus of doing this is that it involves minimal work and doesn't require a new token or loot system to be implemented, mainly a bunch of design time to figure out the details.

Recremen

Giles.floydub17_ESO wrote: »

@Recremen

Adding my cat pics doesn't make reading the extremely long explanation easier.

Not being rude or mean but I don't by Zos will read it due to length with lots of fluff. Try to narrow it down. Get to the point with what your saying and suggesting.

I appreciate the criticism but I get way more traction on a topic when I add funny and nearly-topical pics and when I take the time to explore as much of the subject as possible.

idk

Recremen wrote: »

Giles.floydub17_ESO wrote: »

@Recremen

Adding my cat pics doesn't make reading the extremely long explanation easier.

Not being rude or mean but I don't by Zos will read it due to length with lots of fluff. Try to narrow it down. Get to the point with what your saying and suggesting.

I appreciate the criticism but I get way more traction on a topic when I add funny and nearly-topical pics and when I take the time to explore as much of the subject as possible.

It's not the pics. It's the post is to worry. Traction you get from players is great but how important is it if Zos refuses to read such a long and overly wordy post? Maybe the suggested changes are not something your interested in having Zos consider it.

Recremen

Reorx_Holybeard wrote: »

Great writeup and interesting ideas. I think it would work great for limited/special sets like Maelstrom/Master but I believe you are vastly underestimating the total number of unique set items and the resulting effect it would have.

From here I get a total of 263 sets comprised of 55,508 different items which is a lot bigger than your 1000 item comparison if I read you right. Probably not a completely unreasonable number to be handled in a system like ESO but not a trivially small number either.

Personally, I think a lot of the loot issues can be partially solved by another trait re-balancing and adjusting exactly what traits drop where. The bonus of doing this is that it involves minimal work and doesn't require a new token or loot system to be implemented, mainly a bunch of design time to figure out the details.

@Reorx_Holybeard

One terabyte is equal to 1,099,511,627,776 bytes; 1 byte is equal to 8 bits. So 8,796,093,022,208 bits to play with in a terabyte. Incidentally, your list includes all the crafting sets, which are not zone drops, so the true number of potential items is 40808. So that's 40808 * 12 characters per account * 7 million players = 3,427,872,000,000 bits needed to keep track of all possible drops for all possible characters for all 7 million players. So yes, less than a terabyte! Thanks loads for the link to the item sets, saved me a lot of info gathering!!

Recremen

Giles.floydub17_ESO wrote: »

Recremen wrote: »

Giles.floydub17_ESO wrote: »

@Recremen

Adding my cat pics doesn't make reading the extremely long explanation easier.

Not being rude or mean but I don't by Zos will read it due to length with lots of fluff. Try to narrow it down. Get to the point with what your saying and suggesting.

I appreciate the criticism but I get way more traction on a topic when I add funny and nearly-topical pics and when I take the time to explore as much of the subject as possible.

It's not the pics. It's the post is to worry. Traction you get from players is great but how important is it if Zos refuses to read such a long and overly wordy post? Maybe the suggested changes are not something your interested in having Zos consider it.

The devs have been absolute darlings when it comes to lengthy, well-researched posts in the past, and in fact we tend to get more dev response from such posts than for small blurbs. I'm actually more worried about the cat pics when it comes to the devs since the forum doesn't have the best format and it comes off as kind of clunky and unprofessional, but they seem to be low-key memelords and cat junkies so it's probably fine. The cat pics are actually to keep the players interested in the post while the long content was specifically tailored for dev consideration. Don't worry, I know what I'm doing!

Panth141

Incredibly eloquent and well thought out post.

To anyone who has just scrolled through the comments: this one is worth scrolling back up and reading!

@ZOS_RichLambert - take a look at this, please!

Lexxypwns

Am I a spiteful jerk if I want loot to stay the way it is because I farmed perfect traits on BSW, Treasure hunter, mother's sorrow, spinners, lich, warlock, and a few other sets and it took me ages?

Reorx_Holybeard

Recremen wrote: »

Reorx_Holybeard wrote: »

Great writeup and interesting ideas. I think it would work great for limited/special sets like Maelstrom/Master but I believe you are vastly underestimating the total number of unique set items and the resulting effect it would have.

From here I get a total of 263 sets comprised of 55,508 different items which is a lot bigger than your 1000 item comparison if I read you right. Probably not a completely unreasonable number to be handled in a system like ESO but not a trivially small number either.

Personally, I think a lot of the loot issues can be partially solved by another trait re-balancing and adjusting exactly what traits drop where. The bonus of doing this is that it involves minimal work and doesn't require a new token or loot system to be implemented, mainly a bunch of design time to figure out the details.

@Reorx_Holybeard

One terabyte is equal to 1,099,511,627,776 bytes; 1 byte is equal to 8 bits. So 8,796,093,022,208 bits to play with in a terabyte. Incidentally, your list includes all the crafting sets, which are not zone drops, so the true number of potential items is 40808. So that's 40808 * 12 characters per account * 7 million players = 3,427,872,000,000 bits needed to keep track of all possible drops for all possible characters for all 7 million players. So yes, less than a terabyte! Thanks loads for the link to the item sets, saved me a lot of info gathering!!

Another interesting number is the number of bytes per character since that is the amount of data that needs to be shuffled around between servers when you zone and between the servers and the client. At 40808 bits that is only 5101 bytes which is roughly the amount of memory needed to store your inventory items. So technically feasible with these rough numbers.

I think we can potentially reduce the number of data needed per set by choosing a random seed and using that to "randomly" sort a fixed item list. For example, with a list "1, 2, 3, 4" and the random seed "78" it would always sort the list to "4, 1, 3, 2" which you would use for the loot RNG, one after the other. Using something like this you would only need a random seed per character and then a list index per set. This works out to only 530 bytes per character if you use a 64 bit seed and 16 bit indexes.

The only problem with such a system is that it makes it potentially harder for people to loot multiple of an item. For example, if I wanted 2 Sharpened Spinner Swords I would have to loot 201 different items to get one Sword and then potentially another 201 items to get another one. On the other hand this system would guarantee you would get a Sword in 201 loots as opposed to the current system which only makes it likely. I guess it depends on whether you get lucky with the current RNG or not....

Recremen

Lexxypwns wrote: »

Am I a spiteful jerk if I want loot to stay the way it is because I farmed perfect traits on BSW, Treasure hunter, mother's sorrow, spinners, lich, warlock, and a few other sets and it took me ages?

This is still a good system for you because there will always be That New Set you'll want to farm up! Think of it as investing in your future.

Recremen

Reorx_Holybeard wrote: »

Recremen wrote: »

Reorx_Holybeard wrote: »

Great writeup and interesting ideas. I think it would work great for limited/special sets like Maelstrom/Master but I believe you are vastly underestimating the total number of unique set items and the resulting effect it would have.

From here I get a total of 263 sets comprised of 55,508 different items which is a lot bigger than your 1000 item comparison if I read you right. Probably not a completely unreasonable number to be handled in a system like ESO but not a trivially small number either.

Personally, I think a lot of the loot issues can be partially solved by another trait re-balancing and adjusting exactly what traits drop where. The bonus of doing this is that it involves minimal work and doesn't require a new token or loot system to be implemented, mainly a bunch of design time to figure out the details.

@Reorx_Holybeard

One terabyte is equal to 1,099,511,627,776 bytes; 1 byte is equal to 8 bits. So 8,796,093,022,208 bits to play with in a terabyte. Incidentally, your list includes all the crafting sets, which are not zone drops, so the true number of potential items is 40808. So that's 40808 * 12 characters per account * 7 million players = 3,427,872,000,000 bits needed to keep track of all possible drops for all possible characters for all 7 million players. So yes, less than a terabyte! Thanks loads for the link to the item sets, saved me a lot of info gathering!!

Another interesting number is the number of bytes per character since that is the amount of data that needs to be shuffled around between servers when you zone and between the servers and the client. At 40808 bits that is only 5101 bytes which is roughly the amount of memory needed to store your inventory items. So technically feasible with these rough numbers.

I think we can potentially reduce the number of data needed per set by choosing a random seed and using that to "randomly" sort a fixed item list. For example, with a list "1, 2, 3, 4" and the random seed "78" it would always sort the list to "4, 1, 3, 2" which you would use for the loot RNG, one after the other. Using something like this you would only need a random seed per character and then a list index per set. This works out to only 530 bytes per character if you use a 64 bit seed and 16 bit indexes.

The only problem with such a system is that it makes it potentially harder for people to loot multiple of an item. For example, if I wanted 2 Sharpened Spinner Swords I would have to loot 201 different items to get one Sword and then potentially another 201 items to get another one. On the other hand this system would guarantee you would get a Sword in 201 loots as opposed to the current system which only makes it likely. I guess it depends on whether you get lucky with the current RNG or not....

@Reorx_Holybeard

This is getting a little too far into the implementation details to be a useful discussion since we don't know how much info per character is "shuffled around between servers". I, for instance, would not keep all of this information tracked all the time, I would just access the specific loot tracker for a specific character when they load into a specific zone or dungeon. There's no need to drag all that data everywhere you go, it's a simple database lookup at best (or whatever method they're using to store that data.

Granted, I'm making the assumption that they are using loot tables, pure RNG, and a few other things, but those seem either observably true, or easily translated into whatever details they are using.

Dawnblade

Simple solutions are almost always the best.

IMO the issues with the RNG system today stem from poorly balanced traits and loot tables bloated with trash items.

Balance traits so the difference between each is much smaller and remove a lot of the garbage from the loot tables, and the current system would be fine.

Oakmontowls_ESO

One concern I have with this purposed system is that it would kill the farming to sell market. This would make it so once you got the bis item to sell, your character is now useless to farm that particular item in any reasonable amount of time. It would also hurt instances where people run content that they already have the bis items just so their friend who still needs the items has an extra chance of getting it.

Recremen

Dawnblade wrote: »

Simple solutions are almost always the best.

IMO the issues with the RNG system today stem from poorly balanced traits and loot tables bloated with trash items.

Balance traits so the difference between each is much smaller and remove a lot of the garbage from the loot tables, and the current system would be fine.

Making all traits desirable or making the difference in performance too small to notice will stop people from repeating content or make traits an uninteresting aspect of gear attainment, respectively. The proposed system is already quite simple and lets the devs fine-tune how much, on average, they want people to repeat the content.

Recremen

Is there a dev I should tag on this? @ZOS_BrianWheeler ?