RollerCoaster Tycoon

References

1. ^GameSpot's new-releases post of 22 March 1999 listed MicroProse and Hasbro Interactive's RollerCoaster Tycoon, developed by Chris Sawyer, as new on store shelves that Monday. IGN's review page gives 31 March 1999 instead, and Computer Games Magazine's interview of January 2000 says only that it was released in March.WebGameSpot, New Releases, 22 March 1999 · IGN, RCT review (1999) · Computer Games Magazine, Sawyer interview, 2000revised
   How we know

   New on store shelves for Monday, March 22, 1999 is MicroProse/Hasbro Interactive's Roller Coaster Tycoon, developed by Chris Sawyer (the guy behind the classic Transport Tycoon).

   Released Mar. 31st, 1999

   Released in March, it was the best-selling game of 1999 according to PC Data

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Revised after a check by a second AI model, 2026-10-26.

   In the claim ledger

2. ^ abThe manual's rules for building a coaster piece by piece: the last piece must join the station platform to close the loop; track may pass over or under itself with enough clearance but may not cross itself at the same level; a rise can't turn straight into a drop without a straight piece between; supports reach only so high; and track goes underground only where the land is prepared.ManualRollerCoaster Tycoon manual
   How we know

   The last piece of track must connect to the first piece (the Station Platform) to create a closed loop.

   However, it is a violation of the safety code to build intersections, so you cannot have the track cross itself at the same level. … In order to switch from a rising track to a dropping track, you must have at least one piece of straight track in the middle. … There is a limit to how high above the land you can safely raise track using supports. … You cannot build track underground unless you prepare the land for it.

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Checked by a second AI model, 2026-10-26.

   In the claim ledger

3. ^ abNearly every shop or stall has a front with a serving window, and that side must face a path for guests to use it; the Information Kiosk serves from all sides. Shops need no entrance, exit or queue.ManualRollerCoaster Tycoon manual
   How we know

   Unlike a ride, a little shop doesn’t need an Entrance or Exit, and a queuing line is also unnecessary.

   almost all shops have a predetermined front side with a built in access window, and that front side must be positioned next to a path in order for the guests to be able to use it. The only exception is the Information Kiosk, which has access windows on all sides.

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Checked by a second AI model, 2026-10-26.

   In the claim ledger

4. ^ abSawyer has said he wanted the guests "uncontrollable": they can't be forced to do anything and make up their own minds from what they see, feel and experience, run by "quite a primitive algorithm". Testers complained that guests got lost too easily in odd-shaped parks.InterviewArcade Attack, Chris Sawyer interview (2018)
   How we know

   Having the guests “uncontrollable” in the game appealed to me a lot – i.e. you can’t force them to do something they don’t want to do as they make their own minds up based on what they see and feel and experience.

   Programming the guests was a really fun challenge, and quite a satisfying one when you see the results of quite a primitive algorithm looking quite “human”. … I remember testers criticising the guests’ ability to get lost far too easily if you built an odd-shaped park, perhaps one with a lake in the middle

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Checked by a second AI model, 2026-10-26.

   In the claim ledger

5. ^The manual's tutorial scenario gives the player until October of Year 1 to bring 250 guests into the park and raise the park rating to at least 600.ManualRollerCoaster Tycoon manualrevised
   How we know

   In the tutorial, you have until October of Year 1 to attract 250 guests into the park and build the park’s rating (a measure of reputation) to at least 600.

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Revised after a check by a second AI model, 2026-10-26.

   In the claim ledger

6. ^Sawyer's own site says RollerCoaster Tycoon was released through Hasbro Interactive early in 1999, followed by two add-on packs, Added Attractions (Corkscrew Follies) in late 1999 and Loopy Landscapes in September 2000; an Xbox version published by Infogrames Interactive in 2003; and RollerCoaster Tycoon 2, published by Infogrames in October 2002.PublisherChris Sawyer, site biography
   How we know

   Chris's second big project was RollerCoaster Tycoon, released through Hasbro Interactive early in 1999, followed by two add-on packs Added Attractions / Corkscrew Follies in late 1999 and Loopy Landscapes in September 2000.

   Published by Hasbro Interactive in 1999, Xbox version published by Infogrames Interactive in 2003

   Published by Infogrames Interactive in October 2002

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Checked by a second AI model, 2026-10-26.

   In the claim ledger

7. ^Computer Games Magazine wrote in January 2000 that RollerCoaster Tycoon was created in a small village near Dunblane, Scotland, mainly by one person, Chris Sawyer.InterviewComputer Games Magazine, Sawyer interview, 2000
   How we know

   MicroProse's RollerCoaster Tycoon was created in a small village near Dunblane, Scotland, mainly by one person, Chris Sawyer.

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Checked by a second AI model, 2026-10-26.

   In the claim ledger

8. ^Sawyer has said RollerCoaster Tycoon began as a sequel to Transport Tycoon in 1996. He had rewritten the isometric world so that track could run on bridges over bridges, and while watching vehicles on those bridges he wondered what they would be like as roller coasters. He adapted the code he had written, had basic roller coasters running within a month or two, and abandoned the sequel. OpenRCT2's readme says the engine of RollerCoaster Tycoon 2 was based on Transport Tycoon.PublisherChris Sawyer, "10 things" feature · Arcade Attack, Chris Sawyer interview (2018) · Retro Gamer 138, Chris Sawyer interview · OpenRCT2 v0.5.6revised
   How we know

   RollerCoaster Tycoon originally started as a sequel to Transport Tycoon back in 1996, but the sequel was abandoned and the code modified to handle roller coasters instead of transport vehicles.

   I got it to the point where I had multi-level bridges over bridges and little block-shaped trains merrily running under and over all these bridges while depth-sorting correctly.

   While watching the vehicles moving around on multi-level bridges in Transport Tycoon 2 I started to wonder what it would be like if those vehicles were roller coasters

   I started adapting the code I had already written and within a month or two I had some basic roller coasters

   Eventually I decided to take a breather from the Transport Tycoon sequel and just have a bit of fun seeing if I could adapt the isometric system to handle a simple roller coaster.

   The engine was based on Transport Tycoon, an older game which also has an equivalent open-source project

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Revised after a check by a second AI model, 2026-10-26.

   In the claim ledger

9. ^Sawyer told Arcade Attack that Bullfrog's Theme Park was a fun game but its roller coasters were "rubbish", which got him thinking, and that a book called "White Knuckle Ride", with photos of big roller coasters, made him think how fun it would be to build one in a Transport Tycoon-like isometric world.InterviewArcade Attack, Chris Sawyer interview (2018)
   How we know

   Bullfrog’s Theme Park game had been published and was a really fun game to play, but quite honestly the rollercoasters in it were rubbish, and it got me thinking.

   It started me thinking about how fun it would be to be able to build something like that in a Transport Tycoon-like 3D isometric game world.

   Checked against the evidence by its author, an AI model, 2026-10-26.

   Checked by a second AI model, 2026-10-26.

   In the claim ledger

10. ^Sawyer has said he bought himself out of the Transport Tycoon 2 contract and worked on "Project White Knuckle" full time, funding it himself and keeping it under wraps. The game kept the name White Knuckle through development, even after Hasbro Interactive became involved; when Hasbro bought MicroProse, the "Tycoon" connection made it RollerCoaster Tycoon.InterviewRetro Gamer 138, Chris Sawyer interview · Chris Sawyer, "10 things" feature
    How we know

    I bought myself out of the Transport Tycoon 2 contract and effectively cancelled the project, and started working on 'Project White Knuckle' full time, funding development myself and keeping the project under wraps

    Even when Hasbro Interactive became involved, the game was still going to be called White Knuckle, but then Hasbro bought Microprose, and the "Tycoon" connection was just too good an opportunity to miss, so the game became "RollerCoaster Tycoon".

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

11. ^Sawyer has said that people in the games industry told him a roller-coaster game would only ever be a niche game, and that the negative reactions gave him serious doubts during its two years of development.InterviewRetro Gamer 138, Chris Sawyer interview · Computer Games Magazine, Sawyer interview, 2000
    How we know

    all the advice was that it would only ever be a niche game, would never make any money

    there were times during the two years of development when I had serious doubts about the game, often because of the negative reaction from other people in the game industry.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

12. ^Sawyer's site says RollerCoaster Tycoon is 99% x86 assembly, with some C to interface with Windows and DirectX, built with Microsoft Macro Assembler 6.11c, Visual C 5 and the DirectX 5 SDK. In a 2015 interview he said Transport Tycoon and RollerCoaster Tycoon on the PC were "entirely" programmed in x86 assembler code.PublisherChris Sawyer FAQ · Retro Gamer 138, Chris Sawyer interview
    How we know

    It's 99% written in x86 assembler/machine code (yes, really!), with a small amount of C code used to interface to MS Windows and DirectX.

    The game was written and compiled using MS Macro Assembler V6.11c, MS Visual C V5, MS DirectX 5 SDK, plus assorted custom-written tools.

    Both Transport Tycoon and RollerCoaster Tycoon on the PC were entirely programmed in x86 assembler code.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

13. ^Sawyer said assembly let him make the simulation more detailed: it needed so much processing that it had to be highly optimised, which he thought impossible in a high-level language. The cost is that assembly can't be recompiled for another platform.InterviewRetro Gamer 138, Chris Sawyer interview
    How we know

    I've no doubt it enabled a much more detailed level of complexity in the game

    achieving that same level of optimisation in a high level language would have proved impossible. The downside of writing in assembler code is that it can't be recompiled to run

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

14. ^Dunkin described it as a combination of Disney's Coaster and Bullfrog's Theme Park, with Theme Park its only real competition. He faulted it for running at only two speeds, stop and go, and for its 21 scenarios, with no way to make more and no open-ended scenario.WebGameSpot, RCT review (1999)
    How we know

    MicroProse's RollerCoaster Tycoon is a combination of two somewhat popular PC games: Coaster, an older game from Disney in which you design roller coasters, and Bullfrog's Theme Park … Since Theme Park is pretty much the only real competition

    the game only runs at two speeds - stop and go … Finally, there are only 21 scenarios and no way to make more, and there's no generic "start from scratch and build till you drop" scenario either.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

15. ^Alan Dunkin's GameSpot review of 13 April 1999 scored it 8.6 out of 10.WebGameSpot, RCT review (1999)
    How we know

    GameSpot Score 8.6 great

    Alan Dunkin, GameSpot POSTED: 04/13/99

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

16. ^Bates wrote that it was a lot like Theme Park, "but so what? It's also a lot of fun", and that MicroProse had beaten Bullfrog's Theme Park 2 to the punch.WebIGN, RCT review (1999)
    How we know

    It's a lot like Theme Park but so what? It's also a lot of fun.

    MicroProse beat Bullfrog's Theme Park 2 to the punch with Roller Coaster Tycoon, a Chris Sawyer game.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

17. ^Jason Bates's IGN review of 2 April 1999 scored it 8.5 out of 10.WebIGN, RCT review (1999)
    How we know

    Great IGN Staff 2 Apr 1999 8.5

    Roller Coaster Tycoon is, quite simply, a lot of fun. --Jason Bates

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

18. ^Jeff Green's review in Computer Gaming World (July 1999) called it "an utterly charming gem of a strategy game from MicroProse", a game in which you get to clean up vomit, and a change of pace for players tired of conquering galaxies and fighting wars.MagazineCGW 180, RCT reviewrevised
    How we know

    to clean up vomit. Finally!

    TYCOON, an utterly charming gem … of a strategy game from … change of pace for those tired of … conquering galaxies, fighting … wars, and building global

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

19. ^Computer Games Strategy Plus's awards, posted on 6 March 2000, named it Strategy Game of the Year, with Heroes of Might and Magic III and Jagged Alliance 2 as finalists. The citation praised it for keeping the player on the fun parts, unlike Theme Park.WebComputer Games Strategy Plus awards 1999
    How we know

    Posted on 03/06/2000 Strategy Game of the Year RollerCoaster Tycoon

    Unlike Theme Park, it never forces the player to get bogged down in the minutia of running a park, doing an amazing job at focusing the player in the fun parts.

    Finalists Heroes of Might and Magic III … Jagged Alliance 2

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

20. ^Computer Games Magazine wrote in January 2000 that RollerCoaster Tycoon was the best-selling game of 1999 according to PC Data, with over 700,000 copies sold in the US. The magazine reasoned from that figure, as about 80% of the market, that it had likely sold more than a million in under a year.InterviewComputer Games Magazine, Sawyer interview, 2000
    How we know

    Released in March, it was the best-selling game of 1999 according to PC Data, selling over 700,000 copies in the US alone. (This represents about 80% of the market, which means it's likely the game has sold more than a million units in less than a year).

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

21. ^Computer Gaming World reported in July 2002, announcing the sequel, that RollerCoaster Tycoon had sold 4 million copies.MagazineCGW 216, RCT sales
    How we know

    Infogrames is on track to release a sequel this fall to Rollercoaster Tycoon, which has now sold 4 million copies.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

22. ^Sawyer recalled that sales started slow but steady and that it was a shock, a few months later, to learn the game was among the top-selling PC games of the year, with players from strategy fans to grandparents.InterviewArcade Attack, Chris Sawyer interview (2018)
    How we know

    sales initially were quite slow but steady. But it was a shock a few months later though to find out that sales had ramped up considerably and the game was now amongst the top selling games for the PC for that entire year.

    not just boys who liked strategy games but also girls, younger children, and adults and even grandparents too.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

23. ^Asked why the game isn't difficult in the usual strategy-game sense, Sawyer said it should be "fun to play with rather than just fun to play against", and that a harder game might turn hours of park building into bankruptcy and the end of the game.InterviewComputer Games Magazine, Sawyer interview, 2000
    How we know

    I think my philosophy was that it should be fun to play with rather than just fun to play against.

    it might have turned a very positive game into a rather frustrating one, where all your hard work designing and building a park might result in bankruptcy and the end of the game.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

24. ^A test run gives each custom ride three ratings, excitement, intensity and nausea, which the manual scales from 0 to 10, though intensity can go over 10. Higher excitement lets the ride charge more and makes it more popular; it comes from the ride type and design, speed, pacing, nearby scenery, the view, close passes by paths and other rides, and novelty. Intensity comes from the forces, inversions and speed; a ride rated over 10 may find no riders even for free. Spinning, sharp turns and inversions raise nausea.ManualRollerCoaster Tycoon manualrevised
    How we know

    purpose for testing is to determine the important characteristics of the ride — excitement, intensity, and nausea factor.

    The higher this number, the more you can successfully charge for admission to the ride, and the more popular the ride will be with your guests.

    The overall speed of the ride, its pacing, and the look of it from the outside can add to the excitement, as can nearby scenery … The age of the ride is a factor, too; novelty adds excitement. … The overall rating depends on all the forces, accelerations, changes of direction, inversions, and speed experienced by a rider over the course of the track. You can achieve intensity ratings over 10 through misguided design decisions, but don’t be surprised when no one is willing to ride it, even for free! … Spinning rides tend to have high nausea ratings, and the more turns (especially sharp ones) and inversions you include, the worse it gets.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

25. ^The Guest Info window lists each guest's current thought and action, and its second tab totals every guest's thoughts in order of frequency, so a complaint near the top flags a problem. Each guest's own window shows their recent thoughts, their spending in the park, and their preferred ride intensity and tolerance for nausea.ManualRollerCoaster Tycoon manualrevised
    How we know

    There are two display tabs in this window. the first lists the current thought and action of each guest in the park

    The second tab summarizes for you the actions and thoughts of every guest, and presents them in descending order of frequency.

    Condition is a detailed analysis of the visitor’s physical status and needs, plus information on his or her ride intensity preference and nausea tolerance. … Thoughts lists this visitor’s most recent few thoughts about your park. … Funds details the guest’s spending while in the park.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

26. ^The manual teaches needs as chains of cause and effect: guests who eat will need bathrooms and litter bins, salty food makes them thirsty, and a hungry guest looks out for a food stall.ManualRollerCoaster Tycoon manual
    How we know

    If you serve food, your visitors will need bathrooms and trash bins; provide them before there’s a problem. Salty foods make guests thirsty, so build drink stalls.

    if a guest is hungry, he or she will keep an eye out for a food stall

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

27. ^The manual says how often guests get lost depends entirely on the footpaths: avoid dead ends, complex routes and long trails to nowhere, and sell park maps at Information Kiosks, which make a guest much less likely to get lost. A lost guest can be picked up and dropped on a central path.ManualRollerCoaster Tycoon manual
    How we know

    How often guests … become lost depends … entirely on the system … of footpaths in your … park. Try to avoid … dead ends, complex

    Having a map … handy significantly reduces a … guest’s chances of getting lost.

    Move the mouse pointer across the landscape until the visitor you’ve retrieved is over a central footpath (preferably near rides).

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

28. ^The park rating, from 0 to 1000, reflects the guests' overall impression of ride design, layout, tidiness, value and efficiency. The manual warns that it shows that there is a problem but never what the problem is.ManualRollerCoaster Tycoon manual
    How we know

    This number (between 0 and 1000) reflects your guests’ overall impression — of ride design, park layout, tidiness, value, efficiency, etc.

    although the rating will never tell you what a problem is, only that there is a problem.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

29. ^The manual says the player has borrowed from the bank to improve the park, and that this loan is the working capital; besides it, the only money is what guests spend in the park. The park value, another report, weighs land and equipment, the quality and profit of the rides, and the park rating.ManualRollerCoaster Tycoon manualrevised
    How we know

    You have borrowed a certain amount from the bank to make improvements to the park under your management. This loan constitutes your working capital, also called cash-on-hand (COH). That and whatever income you derive from guest spending in the park are your only sources of funding.

    calculated by expert assessors to reflect what the park is worth in terms of the value of the land and equipment owned, the quality and profitability of the rides and the park as a whole, and the park’s overall public reputation, as determined by the Park Rating.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

30. ^ abOpenRCT2 is an open-source re-implementation of RollerCoaster Tycoon 2 that needs the original game's files. Its readme says it adds improved guest and staff AI, and brings back mechanics of the first game that the sequel dropped, such as the in-game mountain tool and the "have fun" objective; with the first game's files it can play the original scenarios.DeveloperOpenRCT2 v0.5.6
    How we know

    **OpenRCT2** is an open-source re-implementation of RollerCoaster Tycoon 2 (RCT2).

    improved guest and staff AI … It also re-introduces mechanics from RollerCoaster Tycoon that were not present in RollerCoaster Tycoon 2. Some of those include; mountain tool in-game, the *"have fun"* objective

    If you have the original RollerCoaster Tycoon and its expansion packs, you can [point OpenRCT2 to these] … OpenRCT2 requires original files of RollerCoaster Tycoon 2 to play.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

31. ^In OpenRCT2 a guest weighing a rated ride it isn't already heading for, in dry weather, turns it down when its intensity is below the guest's preferred minimum or above the preferred maximum (capped at 10), each widened by up to 2.55 as the guest gets happier; or when its nausea rating is above the guest's tolerance of 3, 6, 8 or 10, raised the same way. A guest already heading for the ride checks only that its intensity is no more than 10. At the ride, a refusal adds a thought ("X looks too intense for me", "I want to go on something more thrilling than X", "Just looking at X makes me feel sick") and lowers the guest's happiness target by 8.CodeOpenRCT2 v0.5.6
    How we know

    188    static constexpr RideRating_t NauseaMaximumThresholds[] = {
    189        300,
    190        600,
    191        800,
    192        1000,
    

    2005                    if (ride.id == guestHeadingToRideId)
    ⋮
    2007                        if (ride.ratings.intensity > RideRating::make(10, 00) && !gameState.cheats.ignoreRideIntensity)
    ⋮
    2009                            GuestRideIsTooIntense(*this, ride, peepAtRide);
    2010                            return false;
    

    2032                        if (!isPrecipitating || !GuestShouldRideWhileRaining(*this, ride))
    ⋮
    2034                            if (!gameState.cheats.ignoreRideIntensity)
    ⋮
    2036                                // Intensity calculations. Even though the max intensity can go up to 15, it's capped
    2037                                // at 10.0 (before happiness calculations). A full happiness bar will increase the max
    2038                                // intensity and decrease the min intensity by about 2.5.
    2039                                RideRating_t maxIntensity = std::min(intensity.getMaximum() * 100, 1000) + happiness;
    2040                                RideRating_t minIntensity = (intensity.getMinimum() * 100) - happiness;
    2041                                if (ride.ratings.intensity < minIntensity)
    ⋮
    2043                                    if (peepAtRide)
    ⋮
    2045                                        insertNewThought(PeepThoughtType::moreThrilling, ride.id);
    2046                                        if (happinessTarget >= 64)
    ⋮
    2048                                            happinessTarget -= 8;
    ⋮
    

    2383    static void GuestRideIsTooIntense(Guest& guest, Ride& ride, bool peepAtRide)
    ⋮
    2385        if (peepAtRide)
    ⋮
    2387            guest.insertNewThought(PeepThoughtType::intense, ride.id);
    2388            if (guest.happinessTarget >= 64)
    ⋮
    2390                guest.happinessTarget -= 8;
    ⋮
    2392            ride.updatePopularity(0);
    ⋮
    2394        guest.choseNotToGoOnRide(ride, peepAtRide, true);
    

    46        moreThrilling = 4,    // "I want to go on something more thrilling than X"
    47        intense = 5,          // "X looks too intense for me"
    48        haventFinished = 6,   // "I haven't finished my X yet"
    49        sickening = 7,        // "Just looking at X makes me feel sick"
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

32. ^In OpenRCT2 a guest refuses a ride that has crashed unless its happiness is at least 225 of 255, thinking "I'm not going on X - it isn't safe". Free transport rides are exempt.CodeOpenRCT2 v0.5.6
    How we know

    1951            // Assuming the queue conditions are met, peeps will always go on free transport rides.
    1952            // Ride ratings, recent crashes and weather will all be ignored.
    1953            auto ridePrice = RideGetPrice(ride);
    1954            if (!ride.getRideTypeDescriptor().flags.has(RtdFlag::isTransportRide) || ride.value == kRideValueUndefined
    1955                || ridePrice != 0)
    

    1985                // If happy enough, peeps will ignore the fact that a ride has recently crashed.
    1986                if (ride.lastCrashType != RIDE_CRASH_TYPE_NONE && happiness < 225)
    ⋮
    1988                    if (peepAtRide)
    ⋮
    1990                        insertNewThought(PeepThoughtType::notSafe, ride.id);
    1991                        if (happinessTarget >= 64)
    ⋮
    1993                            happinessTarget -= 8;
    ⋮
    1995                        ride.updatePopularity(0);
    ⋮
    1997                    choseNotToGoOnRide(ride, peepAtRide, true);
    1998                    return false;
    

    72        notSafe = 30,         // "I'm not going on X - it isn't safe"
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

33. ^In OpenRCT2 a guest's happiness moves 4 points at a time toward a target. With each update the target drifts a point toward the middle of the 0-255 scale, and it sinks while the guest is low on energy, hungry or thirsty, or badly needs the toilet. A guest whose nausea target is 128 or more may throw up while walking, more likely the sicker it is.CodeOpenRCT2 v0.5.6
    How we know

    678    void Guest::updateMotivesIdle()
    ⋮
    680        // Idle peep happiness tends towards 127 (50%).
    681        if (happinessTarget >= 128)
    682            happinessTarget--;
    683        else
    684            happinessTarget++;
    ⋮
    686        nauseaTarget = std::max(nauseaTarget - 2, 0);
    ⋮
    688        if (energy <= 50)
    ⋮
    690            happinessTarget = std::max(happinessTarget - 2, 0);
    ⋮
    693        if (hunger < 10)
    ⋮
    695            happinessTarget = std::max(happinessTarget - 1, 0);
    ⋮
    698        if (thirst < 10)
    ⋮
    

    790        uint8_t newHappiness = happiness;
    791        uint8_t newHappinessGrowth = happinessTarget;
    792        if (newHappiness >= newHappinessGrowth)
    ⋮
    794            newHappiness = std::max(newHappiness - 4, 0);
    795            if (newHappiness < newHappinessGrowth)
    796                newHappiness = newHappinessGrowth;
    ⋮
    798        else
    ⋮
    800            newHappiness = std::min(255, newHappiness + 4);
    801            if (newHappiness > newHappinessGrowth)
    802                newHappiness = newHappinessGrowth;
    ⋮
    805        if (newHappiness != happiness)
    ⋮
    807            happiness = newHappiness;
    ⋮
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

34. ^In OpenRCT2, while a guest off a ride consumes what it carries, a drink raises its thirst meter (fuller is higher) by 7 a step, and food raises its hunger meter by 7, lowers its thirst meter by 3 and fills its bladder by 2: any food makes a guest thirstier.CodeOpenRCT2 v0.5.6
    How we know

    723    /* rct2: 0x0068FA89*/
    724    void Guest::updateConsumptionMotives()
    ⋮
    726        if (timeToConsume == 0 && hasFoodOrDrink())
    ⋮
    728            timeToConsume += 3;
    ⋮
    731        if (timeToConsume != 0 && state != PeepState::onRide)
    ⋮
    733            timeToConsume = std::max(timeToConsume - 3, 0);
    ⋮
    735            if (hasDrink())
    ⋮
    737                thirst = std::min(thirst + 7, 255);
    ⋮
    739            else
    ⋮
    741                hunger = std::min(hunger + 7, 255);
    742                thirst = std::max(thirst - 3, 0);
    743                toilet = std::min(toilet + 2, 255);
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

35. ^In OpenRCT2 a guest's full periodic update runs once every 512 ticks, and on every second one (once every 1,024 ticks) a walking or sitting guest in the park picks one thought at random from those that apply, tired, hungry, thirsty, needing the toilet, or running out of cash, and heads for the nearest food stall, drinks stall, toilet or cash machine to match. A guest leaving the park thinks only "I want to go home". On the alternate updates a guest with nausea of 140 or more thinks "I feel sick", and at 200 or more thinks "I feel very sick" and heads for first aid.CodeOpenRCT2 v0.5.6revised
    How we know

    833    void Guest::tick128UpdateGuest(uint32_t index)
    ⋮
    835        const auto currentTicks = getGameState().currentTicks;
    836        const bool ticksMatchIndex = (index & 0x1FF) != (currentTicks & 0x1FF);
    837        if (ticksMatchIndex && !peepFlags.has(PeepFlag::positionFrozen))
    ⋮
    839            updateConsumptionMotives();
    840            return;
    

    983        if ((index & 0x3FF) == (currentTicks & 0x3FF))
    ⋮
    985            /* Effect of masking with 0x3FF here vs mask 0x1FF,
    986             * which is used in the encompassing conditional, is
    987             * to reduce how often the content in this conditional
    988             * is executed to once every second time the encompassing
    989             * conditional executes. */
    

    991            if (!outsideOfPark && (state == PeepState::walking || state == PeepState::sitting))
    ⋮
    993                uint8_t num_thoughts = 0;
    994                PeepThoughtType possible_thoughts[5];
    ⋮
    996                if (peepFlags.has(PeepFlag::leavingPark))
    ⋮
    998                    possible_thoughts[num_thoughts++] = PeepThoughtType::goHome;
    ⋮
    1000                else
    ⋮
    1002                    if (energy <= 70 && happiness < 128)
    ⋮
    1004                        possible_thoughts[num_thoughts++] = PeepThoughtType::tired;
    ⋮
    1007                    if (hunger <= 10 && !hasFoodOrDrink())
    ⋮
    1009                        possible_thoughts[num_thoughts++] = PeepThoughtType::hungry;
    ⋮
    1012                    if (thirst <= 25 && !hasFoodOrDrink())
    ⋮
    1014                        possible_thoughts[num_thoughts++] = PeepThoughtType::thirsty;
    ⋮
    

    1036                if (num_thoughts != 0)
    ⋮
    1038                    PeepThoughtType chosen_thought = possible_thoughts[ScenarioRand() % num_thoughts];
    ⋮
    1040                    insertNewThought(chosen_thought);
    ⋮
    1042                    switch (chosen_thought)
    ⋮
    1044                        case PeepThoughtType::hungry:
    1045                            GuestHeadForNearestRideWithFlag(*this, false, RtdFlag::sellsFood);
    1046                            break;
    1047                        case PeepThoughtType::thirsty:
    1048                            GuestHeadForNearestRideWithFlag(*this, false, RtdFlag::sellsDrinks);
    1049                            break;
    1050                        case PeepThoughtType::toilet:
    1051                            GuestHeadForNearestRideWithSpecialType(*this, false, RtdSpecialType::toilet);
    ⋮
    

    1068            if (nausea >= 140)
    ⋮
    1070                PeepThoughtType thought_type = PeepThoughtType::sick;
    1071                if (nausea >= 200)
    ⋮
    1073                    thought_type = PeepThoughtType::verySick;
    1074                    GuestHeadForNearestRideWithSpecialType(*this, true, RtdSpecialType::firstAid);
    ⋮
    1076                insertNewThought(thought_type);
    

    44        sick = 2,             // "I feel sick"
    45        verySick = 3,         // "I feel very sick"
    

    51        goHome = 9,           // "I want to go home"
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

36. ^In OpenRCT2 a guest with energy of at least 55, happiness of at least 45 and £5 or more in pocket always stays in the park (energy 70 and happiness 60 in a park without money). Any other guest has about a 1 in 20 chance at each check of deciding to go home.CodeOpenRCT2 v0.5.6
    How we know

    3027        /* Peeps that are happy enough, have enough energy and
    3028         * (if appropriate) have enough money will always stay
    3029         * in the park. */
    3030        if (!guest.peepFlags.has(PeepFlag::leavingPark))
    ⋮
    3032            if (getGameState().park.flags.has(ParkFlag::noMoney))
    ⋮
    3034                if (guest.energy >= 70 && guest.happiness >= 60)
    ⋮
    3036                    return;
    ⋮
    3039            else
    ⋮
    3041                if (guest.energy >= 55 && guest.happiness >= 45 && guest.cashInPocket >= 5.00_GBP)
    ⋮
    3043                    return;
    ⋮
    3048        // Approx 95% chance of staying in the park
    3049        if ((ScenarioRand() & 0xFFFF) > 3276)
    ⋮
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

37. ^In OpenRCT2 a walking guest checks whether it is lost each time it reaches the next point on its way (unless it has stepped into water), and the pathfinder also checks at a dead end, a path whose only way on is back. In a park with at least two rides every second check adds to its lost counter; at 254 the guest thinks "I'm lost!" and its happiness target falls by 30, and the counter drops back to 230, so the thought repeats on later checks.CodeOpenRCT2 v0.5.6revised
    How we know

    1923        // Check if in a dead end (i.e. only edge is where the peep came from)
    1924        if (!(edges & ~(1 << direction)))
    ⋮
    1926            // In a dead end.  Check if peep is lost, etc.
    1927            peep.checkIfLost();
    1928            peep.checkCantFindRide();
    1929            peep.checkCantFindExit();
    

    5475        const auto [pathingResult, _] = performNextAction();
    5476        if (!(pathingResult & PATHING_DESTINATION_REACHED))
    5477            return;
    ⋮
    5479        if (getNextIsSurface())
    ⋮
    5481            auto surfaceElement = MapGetSurfaceElementAt(nextLoc);
    ⋮
    5483            if (surfaceElement != nullptr)
    ⋮
    5485                int32_t water_height = surfaceElement->getWaterHeight();
    5486                if (water_height > 0)
    ⋮
    5488                    moveTo({ x, y, water_height });
    5489                    setState(PeepState::falling);
    5490                    return;
    ⋮
    5495        checkIfLost();
    ⋮
    

    1331    void Guest::checkIfLost()
    ⋮
    1333        if (!peepFlags.has(PeepFlag::lost))
    ⋮
    1335            if (RideGetCount() < 2)
    1336                return;
    1337            peepFlags.flip(PeepFlag::unknown21);
    ⋮
    1339            if (!peepFlags.has(PeepFlag::unknown21))
    1340                return;
    ⋮
    1342            timeLost++;
    1343            if (timeLost != 254)
    1344                return;
    1345            timeLost = 230;
    ⋮
    1347        insertNewThought(PeepThoughtType::lost);
    ⋮
    

    58        lost = 16,            // "I'm lost!"
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

38. ^In OpenRCT2 the park rating starts from 1150 (1050 in a park set to a difficult rating). The guest count adds from -150 at none to +3 at 2,000 or more. Happy guests, those above 128 on a 0-255 happiness scale, add from -500 at none to nothing once five in six guests are happy. Guests leaving the park who can't find the exit cost 7 points each beyond the first 25.CodeOpenRCT2 v0.5.6
    How we know

    383        int32_t result = 1150;
    384        if (park.flags.has(ParkFlag::difficultParkRating))
    ⋮
    386            result = 1050;
    ⋮
    389        // Guests
    ⋮
    391            // -150 to +3 based on a range of guests from 0 to 2000
    392            result -= 150 - (std::min<int32_t>(2000, park.numGuestsInPark) / 13);
    ⋮
    394            // Find the number of happy peeps and the number of peeps who can't find the park exit
    395            uint32_t happyGuestCount = 0;
    396            uint32_t lostGuestCount = 0;
    397            for (auto peep : EntityList<Guest>())
    ⋮
    399                if (!peep->outsideOfPark)
    ⋮
    401                    if (peep->happiness > 128)
    ⋮
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

39. ^OpenRCT2's park rating then weighs the rides: the average uptime gives -200 to 0; rated rides whose average excitement is near 3.7 and intensity near 5.2 give up to 0 from -100; and the summed excitement and intensity of all rated rides, each capped, give -200 to 0.CodeOpenRCT2 v0.5.6
    How we know

    426        // Rides
    ⋮
    428            int32_t rideCount = 0;
    429            int32_t excitingRideCount = 0;
    430            int32_t totalRideUptime = 0;
    431            int32_t totalRideIntensity = 0;
    432            int32_t totalRideExcitement = 0;
    433            for (auto& ride : RideManager(gameState))
    ⋮
    435                totalRideUptime += 100 - ride.downtime;
    436                if (RideHasRatings(ride))
    ⋮
    438                    totalRideExcitement += ride.ratings.excitement / 8;
    439                    totalRideIntensity += ride.ratings.intensity / 8;
    440                    excitingRideCount++;
    ⋮
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

40. ^OpenRCT2's park rating loses 4 points for each piece of litter at least 7,680 ticks old, up to 150 pieces, and a penalty for casualties; the result is clamped to 0-999 and recalculated every 512 ticks. A cheat can force the rating to a set value.CodeOpenRCT2 v0.5.6
    How we know

    477        // Litter
    ⋮
    479            // Counts the amount of litter whose age is min. 7680 ticks (5~ min) old.
    480            const auto litterList = EntityList<Litter>();
    481            const auto litterCount = std::count_if(
    482                litterList.begin(), litterList.end(), [](auto* litter) { return litter->getAge() >= 7680; });
    ⋮
    484            result -= 600 - (4 * (150 - std::min<int32_t>(150, litterCount)));
    ⋮
    487        result -= park.ratingCasualtyPenalty;
    488        result = std::clamp(result, 0, 999);
    489        return result;
    

    333        // Every ~13 seconds
    334        if (currentTicks % 512 == 0)
    ⋮
    336            park.rating = CalculateParkRating(park, gameState);
    

    378        if (gameState.cheats.forcedParkRating != kForcedParkRatingDisabled)
    ⋮
    380            return gameState.cheats.forcedParkRating;
    

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

41. ^In a 2010 interview Persson described the builder game he was working on when he came across Infiniminer as "a lightweight Dwarf Fortress", "kinda like if DF and Rollercoaster Tycoon had a baby".InterviewGamasutra, Notch interview, 23 March 2010revised
    How we know

    It was supposed to be a lightweight Dwarf Fortress. Kinda like if DF and Rollercoaster Tycoon had a baby.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

42. ^Persson wrote in October 2011 that he started on Minecraft in the summer of 2009 with a top-down isometric mining game pitched as "Dwarf Fortress meets Rollercoaster Tycoon meets Dungeon Keeper", then played Infiniminer, saw that a first-person view with blocky graphics worked well, and began prototyping that.Dev blogNotch, "Happy birthday, us!", 16 October 2011
    How we know

    I started working on Minecraft in the summer of 2009, starting with a top-down isometric mining game. The elevator pitch for the game was “Dwarf Fortress meets Rollercoaster Tycoon meets Dungeon Keeper”. … but then I played a game called Infiniminer and realized the first person perspective with blocky graphics worked really well and started prototyping it.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

43. ^Sawyer calls RollerCoaster Tycoon 2 his favorite, but thinks the original and its expansion packs may have been better games: the sequel's free-form play took away some of the challenge, and the original's scenarios were better. RollerCoaster Tycoon Classic, which recreates the sequel, brought back the original's scenario structure, for 95 scenarios in all.InterviewArcade Attack, Chris Sawyer interview (2018)revised
    How we know

    RollerCoaster Tycoon 2 is my personal favourite

    The park scenarios I designed for the original game and its expansion packs were better too … Looking back though I think the original game and its expansion packs might have actually been better “games” … what this added in terms of features and flexibility I think it also lost in terms of taking away the challenge and sense of achievement of the original game. … although it recreates the entire RollerCoaster Tycoon 2 game on the more modern platforms I also included the park scenario challenge structure from the original game … It was amazing seeing that in total this meant 95 different park scenarios to work through

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger

44. ^Sawyer said the RollerCoaster Tycoon engine was then adapted for a new transport game, Chris Sawyer's Locomotion (2004), so the code went full circle from the abandoned Transport Tycoon 2.InterviewRetro Gamer 138, Chris Sawyer interview · Chris Sawyer, site biography
    How we know

    RollerCoaster Tycoon was developed from the ashes of the abandoned Transport Tycoon 2 project, and now the well-honed RollerCoaster Tycoon game engine was then being adapted to create a new transport game.

    The long-awaited sequel to Transport Tycoon, Chris Sawyer's Locomotion, was finally completed and released through Atari in October 2004.

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Checked by a second AI model, 2026-10-26.

    In the claim ledger

45. ^PC Gamer's Ellen McGrody likened Parkitect in 2018 to the early RollerCoaster Tycoon games, as isometric and management-focused, and wrote that it had grown from a recreation of the isometric rollercoaster sim into an evolution of it. Its developers, quoted there, speak of "older games" without naming one.WebPC Gamer, Parkitect (2018)revised
    How we know

    Like the early RollerCoaster Tycoon games, it's isometric and focused on management.

    the game has grown from a simple recreation of the isometric rollercoaster sim to an evolution of it.

    "People were unsure whether we were only trying to recreate older games or if we were going to add anything meaningful and new," they said

    Checked against the evidence by its author, an AI model, 2026-10-26.

    Revised after a check by a second AI model, 2026-10-26.

    In the claim ledger