Understanding the Ebb and Flow of Guests + Guest AI

[metsfan1025]

Hello everyone,

I know there have been some complaints about guest AI but I was hoping to have an informational discussion about it. I have mostly been playing through the career mode and I got to say, I feel like something is a little off about the guest AI but I cannot put my finger on what exactly are the root causes.

I am going to begin by talking about guests in RCT2 as a point of a comparison and because I am much more familiar with the game.

Guests in RCT2 seemed to follow two main behavior sets: normal and "headed to" behavior.

-A normal guest wanders around somewhat randomly along paths and occasionally partakes in appropriate activities if they happen to go in that direction, i.e. riding a ride within their intensity preferences or doing something neutral like purchase a balloon.
-A guest can begin to "head to" either a ride (if they just rode it and liked it, have a voucher, etc.) or a nearest appropriate facility to fulfill a need (hunger, thirst, sickness, etc.).
-A "headed to" guest will generally move in the direction of where they want to go. However, even if a ride is reachable by paths, guests actually could get trapped/lost trying to get where they wanted to go (I forget if this only occurs if the guest does not have a map). For example, if a large park has a general "U" layout, a guest in the upper right corner may get trapped in path systems there trying to head to a ride in the top left corner without realizing they can go down and around.

I believe this last point is attributable to the paths being more of just blocks where guests can randomly choose to turn or not, or choose to head in the general direction of the ride.

Besides this one point, there was a general consistency in behavior that also made management statistics more consistent. For example, you could usually expect monthly ride profits to be pretty similar month to month, and if there was a discrepancy, you would usually have a good idea why (i.e. rain, major ride breaking down etc.).

Now on to Planet Coaster.

First of all, I believe that with everything properly connected by at least one path, a guest in Planet Coaster cannot actually get lost. The way paths connect seems to be more of a proper node and edge system, so I assume guests can always just get from point A to B with some inherent park-layout knowledge (which I assume is just some Djisktra's algorithm type stuff). I also assume this is how the staff roster's operate, although I am not sure what exact order between components the staff member chooses.

My bigger problem is understanding how the guests actually behave specifically. It seems like they more frequently "head to" attractions. For example, a distant, empty area in RCT2 will still have some guests who happened to randomly walk there. This will not happen in PC. However, if you place a new ride in that area, you will get at least some foot traffic there from guests heading directly to that ride.

However, they also seem to head toward things in large groups, especially when queue's are full or new rides open up.

I think a good way to illuminate this thought is in the Silversmith Manor career mission, which is essentially laid out as a long spiral path from the entrance to the River Rapids ride in the center. I began this scenario by building a few gentle rides by the entrance. The jump in park rating caused many guests to enter, almost all of whom decided they wanted to head toward the rapids ride most leaving my gentle rides nearly empty. However, when many got to the rapids ride to discover a full queue, there was another huge exodus back to the gentle rides at the park entrance, and then the rapids ride was not nearly as full.

I have not fully followed guests enough to fully understand behavior but I feel as though these types of swings happen frequently and cause huge variations in monthly profits (i.e. if guests happen to flood the queue of an expensive ride one month).

Unlike RCT2, in PC I notice very large differences in monthly ride incomes for no discernible reason; the above is one of my best guesses why. This type of variation makes goals such as "maintain profit of XXX for 3 months" a nightmare. I have heard some suggest always having full queue scenery--I tried this and I cannot say for sure if this has any impact honestly. In my mind I imagined this just made guests less angry to wait in a long queue and would not impact whether they joined the queue.

And beyond all this there are still points I do not understand such as why guests will simply refuse to purchase certain shop brands.

Sorry if that was all a bit jumbled, just thought I would spill out some of my observations about this game as i try to figure out the optimal park designs and management strategies. Has anyone else noticed any of this or any other specific behaviors they wish to discuss?

 

[Jack95970002]

First, of all, great post... your analysis of RCT2 was spot on-- I don't think I ever thought about guest behavior when playing that game... it seemed very natural.

In PC, I think you are right-- the guests have a much more "head to" mindset.

What first sparked my interest in guest activity was their disinterest in gift shops-- in my most recent park I created a sort of little "main street," with the necessities you'd expect... once the simulation goes on for a while, this part of the park essentially becomes uninhabited (with no rides there). While I am quick to place necessities around the more popular paths, (between major attractions) I do wish that guests walked a bit more randomly, like in RCT2.

In my first park, after I built a lot, I decided to switch from free entry/pay for rides to pay entry/free rides... and my reserve plummeted. There isn't much of a transfer between guests entering and leaving... it's more so, enter en masse, and then stay. I wish guests left and arrived more often, both to make pay-to-enter more viable economically, and to bring traffic to that "main street" I mentioned, the idea being guests would use the shops there when they were entering or leaving. (They do when they enter, at least for priority passes, but I find that their hunger/thirst/toilet needs are full upon entering.)

So granted that we know that guests have a "headed to" priority, I'd like to know what specifically encourages them to buy gifts when they do... I'd presume they buy them simply when they're in the vicinity but I'd like something more concrete, because I often have the problem of negative profit and vendors quitting in gift shops.

Cheers.

 

[breezerHOG1]

Its strange how in many little ways the classics are still better. I totally agree with the "wander" vs "head to" movements in RCT vs PC. Thats the problem with making "guests smart" not everything needs to be realistic, but the game needs to have a good guest flow for the economic system to function properly. Seeing an area of a park empty despite scenery is strange, seeing guests flood a ride one week and not again for a month is strange, seeing certain shops not sell anything and vendors quitting is strange. Mechanics waiting for a queue to empty before repairing the ride is strange. The lack of breakdown variations is strange. Guests not buying maps is strange. I would have rather seen "over-flow queues" instead of fast pass and separate entrance exit stations for coasters would be nice too.

 

[Lurker1]

Randomly wandering is a realistic way to tackle a park. Often if you go to a park (especially for the first time) you might be aware of a few of the rides, but you almost certainly won't know your way around. I might look at the map once or twice and say I'll head that way, but I don't constantly think about the shortest route. I also stop if there is something of interest. Just because I'm heading to one ride, it doesn't mean I'm not tempted by rides along the path.

Awareness of queues before you get to the gate is key in most parks. Parks use screens with times to influence guest flow around a park. Too many guests in one area of the park, they'll inflate the times on the queue boards to encourage some people to head to other rides. Adding queue boards which guests pass and pick up information about queues could aid the AI.

As well as wandering or heading directly for a ride I also sometimes decide to go and see/do things unrelated to attractions. Good spot to watch the coaster? I'll stop to watch it woosh past. Quite area with a choice for lunch? I'll visit or remember for later. Somewhere with excellent scenery, I'll go to see just that. Sometimes I just want to rest my legs on a bench out of the crowds!

Guests in planer coaster have a single track mind. They choose a target (seemingly either a ride or the closest shop) and they head along the shortest route to it. Unless they get blocked, nothing interupts them.

Two ways I'd improve AI

Interruptions : guests plans are queued, things change in priority with various triggers. If they pass something they like, there is a chance they'll visit it while passing.

Randomness : guests need more choices. Don't just decide to take the fastest route. Take the long route, decide to explore a path just because, go somewhere quiet, busy.

 

[1096bimu]

From what I can tell

• Guests never wonder, they always know if a road leads to nowhere and they will never go down that path
• Guests will never discover a ride and get on it, they just know that a ride exists by telepathy and head straight for it if they want it
• Guests will not head straight to stores or facilities like rides, they discover those instead
• Guests will always take the shortest path regardless of how congested or uninteresting it is

Also no, all income and money related challenges or objectives are incredibly easy, if you know the system you can manipulate it to your advantage.

 

[Bitter Jeweler]

if you know the system you can manipulate it to your advantage.

That's always been the case in the RCT series though.

 

[tyczj]

That's always been the case in the RCT series though.

or any game for that matter

 

[breezerHOG1]

That's always been the case in the RCT series though.

yes but after all these years there could have been some advances made

 

[Bitter Jeweler]

or any game for that matter

Except Myst. [haha]

 

[tyczj]

yes but after all these years there could have been some advances made

one could say there were advancements made since people know where they go on their own, they know the shortest route to get there, they now have collision detection between other guests, they are somewhat aware of surroundings, travel in groups.

However some of these advancements did dumb down the game a bit and take away some of the fun the other RCT games had but that does not take away the advancement made

 

[breezerHOG1]

Also no, all income and money related challenges or objectives are incredibly easy, if you know the system you can manipulate it to your advantage.

That's always been the case in the RCT series though.

yes but after all these years there could have been some advances made

one could say there were advancements made. However some of these advancements did dumb down the game a bit and take away some of the fun the other RCT games had

I was talking about the economic system which could be better
guest animations and AI is not a "major" step up to the game IMO its just nice graphics

 

[metsfan1025]

From what I can tell

• Guests never wonder, they always know if a road leads to nowhere and they will never go down that path
• Guests will never discover a ride and get on it, they just know that a ride exists by telepathy and head straight for it if they want it
• Guests will not head straight to stores or facilities like rides, they discover those instead
• Guests will always take the shortest path regardless of how congested or uninteresting it is

Also no, all income and money related challenges or objectives are incredibly easy, if you know the system you can manipulate it to your advantage.

I agree with how you describe guest behavior but I am curious, how exactly do you manipulate it? To complete profit quests (i.e. the Gulch one which requires maintaining some profit for 3 months) the only applicable strategy I could think of is building a very compact park with as many rides as I can afford. However, I find because of the way people move around I get diminishing returns on profit when I add rides and can beat the profit mark on random months but not sustain it for three.

As for the "head to" behavior as opposed to random wander, I actually think this is a more realistic behavior as most people would have a map and probably some plan of what they want to go on. I'd say the major difference is that they are a bit smarter about it; for example, they'd probably have planned routes that allow them to ride rides in order that are close together, and they may deviate from the path if they see an interesting ride with little to no line.

Even the swarming type behavior has some realism to it; for example, at common lunch times you can expect ride lines to be shorter as many people stop to get food. The problem is, when this type of behavior happens in PC guests become unhappy and you get notifications like "many guests are hungry" and that all your food stalls are over crowded.

I guess it's difficult to say what the best system is because in this case, realism makes management a bit clunkier. Though on the other hand, the lack of realism makes other parts more difficult--i .e. without guests leaving at the end of the day and entering in the morning like a true park, profiting solely on entrance tickets is nearly impossible (even though this is how every amusement park I've ever been to runs).

 

[Bullethead]

From what I can tell

• Guests never wonder, they always know if a road leads to nowhere and they will never go down that path
• Guests will never discover a ride and get on it, they just know that a ride exists by telepathy and head straight for it if they want it
• Guests will not head straight to stores or facilities like rides, they discover those instead
• Guests will always take the shortest path regardless of how congested or uninteresting it is

.

I agree and think this is a good synopsis. I do have a few extra points, though.

Guests do wander, but it's usually for only a couple of seconds before they decide where they're going next. Then they go to "heading to". I think this has an insignificant impact on overall behavior.

I like how they don't have to discover rides. After all, the new thing has been in the news or at least social media. Guests at the park during construction would have been tweeting about it.

When travelling long distances, guests will take advantage of multiple train/monorail stations if that's faster than walking the path, even if the track is a longer distance.

Despite all guests constantly tweeting (having thoughts), none of them read each other's tweets. Thus, guests have no idea how long a ride's queue is until they get to the ticket booth. This is what causes the mass migrations mentioned in the OP. HOWEVER, this soon settles out. Every time you add something to a park, you immediately get a mob of new customers, many of whom have the same idea so go to the same place. But after a while, this all evens out until you add the next new thing.

It seems like maybe 75% of all guests come in groups of 4, with groups of 2-3 making up most of the rest. Solo adults and teens are rather rare. Groups typically walk in a 2x2 square formation so all 2-way paths must be wide enough for 4 at least. In high-traffic areas, you need room for at least 8 abreast. It's also a good idea not to create chokepoints, like a single path (even a wide one) leading into a group of many shops.

 

[SoapyIllusion]

my biggest issue with the topic is just how hard this is to measure. Having some visualization would go so long. Maybe the devs and marketing were banking on the ambiguity. But I think as an overall point it looks bad on the game. I mean its not hard to say you think your seeing intelligence but it could be luck, could be your imagination. Parkitect while not that much further ahead, having more heat maps and the guest tracking features adds to the simulation.

 

[seeker7]

Randomly wandering is a realistic way to tackle a park. Often if you go to a park (especially for the first time) you might be aware of a few of the rides, but you almost certainly won't know your way around. I might look at the map once or twice and say I'll head that way, but I don't constantly think about the shortest route. I also stop if there is something of interest. Just because I'm heading to one ride, it doesn't mean I'm not tempted by rides along the path.

Two ways I'd improve AI

Interruptions : guests plans are queued, things change in priority with various triggers. If they pass something they like, there is a chance they'll visit it while passing.

Randomness : guests need more choices. Don't just decide to take the fastest route. Take the long route, decide to explore a path just because, go somewhere quiet, busy.

I aggree 100%. Guests should be exploring, too, not just take the most efficient route from point A to B.

 

[Inge Jones]

I aggree 100%. Guests should be exploring, too, not just take the most efficient route from point A to B.

It was nice in RCTW when I'd see a few peeps wandering around in my quiet wooded area, apparently just to get a few moments peace from the noise of the park

 

[1096bimu]

I agree with how you describe guest behavior but I am curious, how exactly do you manipulate it? To complete profit quests (i.e. the Gulch one which requires maintaining some profit for 3 months) the only applicable strategy I could think of is building a very compact park with as many rides as I can afford. However, I find because of the way people move around I get diminishing returns on profit when I add rides and can beat the profit mark on random months but not sustain it for three.

As for the "head to" behavior as opposed to random wander, I actually think this is a more realistic behavior as most people would have a map and probably some plan of what they want to go on. I'd say the major difference is that they are a bit smarter about it; for example, they'd probably have planned routes that allow them to ride rides in order that are close together, and they may deviate from the path if they see an interesting ride with little to no line.

Even the swarming type behavior has some realism to it; for example, at common lunch times you can expect ride lines to be shorter as many people stop to get food. The problem is, when this type of behavior happens in PC guests become unhappy and you get notifications like "many guests are hungry" and that all your food stalls are over crowded.

I guess it's difficult to say what the best system is because in this case, realism makes management a bit clunkier. Though on the other hand, the lack of realism makes other parts more difficult--i .e. without guests leaving at the end of the day and entering in the morning like a true park, profiting solely on entrance tickets is nearly impossible (even though this is how every amusement park I've ever been to runs).

Don't build popular rides close to the entrance, build the most popular rides furthest away from the entrance so you have space to build stores and less popular rides, guests don't mind the walk. Also the more they walk the more food/drink they have to buy.

I think the best park layout is one super long 10m route that leads to everything. any kind of forks and loops makes the shops/rides less efficient and leaves areas your janitors don't visit.

If you want to beat profit objectives you just have to make your rides more profitable, know the rate of return for each ride, as in monthly profit over building cost. Most static rides suck because their return is like 30%. A short coaster with max scenery and optimized routine can make $10k so about 100% return. A cheaty short track ride can get 300%.

 

[metsfan1025]

Don't build popular rides close to the entrance, build the most popular rides furthest away from the entrance so you have space to build stores and less popular rides, guests don't mind the walk. Also the more they walk the more food/drink they have to buy.

I think the best park layout is one super long 10m route that leads to everything. any kind of forks and loops makes the shops/rides less efficient and leaves areas your janitors don't visit.

If you want to beat profit objectives you just have to make your rides more profitable, know the rate of return for each ride, as in monthly profit over building cost. Most static rides suck because their return is like 30%. A short coaster with max scenery and optimized routine can make $10k so about 100% return. A cheaty short track ride can get 300%.

That's a good point about building coasters instead of a ton of small flat rides.

For the one long path approach though, what happens when a guest decides to head to a ride on the complete opposite end of the path? Isn't that a long time the guest takes not spending money? I fail to see why that is better than just having everything as absolutely clustered as possible.

 

[Narad]

I built a park that had a main street coming in and about half a dozen flat rides and a couple of coasters off that path. I put all the ride/coaster exits to the back so the guests had to walk a longer path around toward the far side of the park and pass through a mall area to get back to the rides. I hoped this would create more traffic for the shops and increase sales.

The guests weren't getting lost and seemed like things went pretty well. Then I put all the exits back out the front to the main path and my revenue nearly doubled. I thought I would get much more in shop sales forcing people to go through the mall, but results were contrary. My shop revenue went up when they didn't pass through all the time. I believe what I demonstrated is they don't get distracted much along the way. The head to mentality is strong.

Certainly forcing them to walk around a big loop to get back to rides dropped my revenue because they spend a bunch more time walking and not buying (like was pointed out above). A lesson learned.

Sad that even my utterly cowtastic cosmic cow shop couldn't distract them. [shocked]