The Stress Test

Applying the queue model to parks you already know

The previous chapter built a capacity engineering model from scratch: design days, simultaneous occupancy, attraction throughput, expected queues during peak hours. All on paper, all applied to the project itself.

The obvious next question: does the model hold up when applied to real parks, with real data?

This chapter is that test. We take the same model and apply it to nine parks — from PortAventura to Europa-Park, from Terra Mítica to Efteling — using available data, real attractions and real visitor numbers. Some results are exactly what any regular visitor would have suspected. Others are counterintuitive. And a few are genuinely uncomfortable for the parks in question.

Before the numbers, two necessary warnings.

First: the model describes peak days. The two- and three-hour queues that appear in these pages are not the average experience at these parks — they are what happens on the worst day of the year. A park with two million annual visitors might have fifteen or twenty days per year at that level of pressure; for the rest of the year, queues are far shorter. This does not excuse structural throughput problems, but it puts the numbers in context.

Second: the CPH figures published by manufacturers assume all trains or vehicles running simultaneously. If a park takes a train out of service for maintenance or a breakdown — something that happens regularly, especially on wooden coasters and long-cycle attractions — real throughput drops by 25 to 50 percent. That difference appears in no published data. It is invisible, but it determines the actual queue experience.

With that said, let’s begin.

The Spanish parks

PortAventura World — the anchor that works

PortAventura is Spain’s largest theme park by attendance and the only one that appears in the TEA/AECOM ranking of Europe’s 25 most-visited parks. In 2024 it received approximately 3.8 million visits, giving a daily average of nearly 14,000 visitors across its roughly 276 operating days.

The design peak — the busiest day of the year — is estimated at around 35,000 visitors. Applying the 65% simultaneous occupancy factor, there are up to 22,750 people in the park at the same time during peak pressure hours.

What does PortAventura do with 22,750 simultaneous people? The answer lies in its main attraction portfolio:

The first three — Shambhala, Dragon Khan and Furius Baco — are high-throughput anchors doing the heavy lifting. The problems appear at the edges of the portfolio: Uncharted at the top (new attraction, high demand) and El Diablo at the bottom (a 1995 format in a 2025 park).

Applying the model to a 35,000-visitor day:

Estimated total ride experiences = 35,000 × 6 = 210,000
Peak-hour aggregate demand       = 210,000 × 0.60 / 6h ≈ 21,000 PPH

Empirical validation (queue-times.com, 2024 season): Shambhala recorded a 2024 season average of 37 minutes and a peak-day average of 64 minutes. The model predicts 75 minutes for the most crowded day of the year — marginally high but within the correct range for the worst Saturday in August. For Uncharted, empirical data shows a peak-day average of around 90 minutes, with ceilings of 120 minutes in August. The model predicts 160 minutes — significantly overstated. The explanation: PortAventura manages Uncharted with an active Virtual Queue system that suppresses the standby queue. The model predicts the queue that would exist without demand management; the park is already applying exactly the solution the model recommends.

Shambhala at 60–75 minutes on a peak day is the cost of being the most-demanded attraction in a 35,000-visitor park. The truly revealing finding is Uncharted and its active Virtual Queue.

Uncharted (Intamin + Sally Corporation, opened 2024) was PortAventura’s most significant investment in years: Spain’s first dark coaster, a genuinely original concept in the European market, with top-tier theming. The cost exceeded €25 million. And it has 900 CPH.

That is not a criticism of the design — it is a consequence of the system’s architecture. A long-cycle dark coaster with individual vehicle loading and multiple effects rooms cannot dispatch like a B&M train running on a chain lift. The problem is not that Uncharted is a bad ride; it is that a 900 CPH attraction in a 35,000-visitor park needs active demand management from day one. Without a Virtual Queue or time-slot booking system, Uncharted’s queue would be among the longest in Spain in August.

The lesson: an attraction’s cost and its throughput have no proportional relationship. You can spend €25 million and get 900 CPH. You can spend the same amount and get 1,500 CPH. The difference lies in the format chosen, not the budget.

Parque Warner Madrid — the most expensive investment with the worst queues

Parque Warner Madrid (San Martín de la Vega, near Madrid) is the second park in the Spanish sample by attendance: 2.25 million visitors in 2024, ranked 13th in the TEA European index. It operates approximately 190 days per year — fewer than PortAventura — which compresses its annual demand into a shorter window and raises peak-day pressure.

The design peak is estimated at 18,000 visitors. The main portfolio:

Peak-hour aggregate demand (18,000 visitors) ≈ 10,800 PPH

Empirical validation (queue-times.com, 2024 season): The empirical data for Warner reveals the most striking discrepancy in the entire validation. Batman: Gotham City Escape records a peak-day average of 40–53 minutes — the model predicts 204. Superman reaches 35–50 minutes on peak days — the model predicts 139. Stunt Fall sits at 20–35 minutes — the model predicts 168.

The model overestimates by a factor of three to five across all Warner attractions.

Why? Two concurrent reasons. First: the model distributes demand assuming each attraction receives its share of total visitor flow as if all visitors had identical preferences at equal intensity. In reality, not every Warner visitor attempts Batman GCE, Superman and Stunt Fall on the same day. Actual per-attraction demand is lower than proportional distribution suggests. Second: Warner operates its 2.25 million annual visitors across roughly 190 days with operating windows of up to twelve hours in summer — this spreads demand across far more time than the model allocates to the “peak hour.”

The D/D/1 model is correct as a tool — but it is sensitive to the per-attraction demand inputs. An overestimate of the percentage of visitors choosing each attraction produces predictions that diverge exponentially from reality.

What does not change: Batman: Gotham City Escape remains the attraction with the worst queue in Warner’s portfolio in relative terms, even at 40–53 real minutes, exactly as the model predicted comparatively. The ranking between attractions is valid even when the absolute values are incorrectly calibrated.

Here is the most revealing finding in the Spanish benchmark.

Batman: Gotham City Escape opened in 2023. It is the park’s newest attraction, the largest recent investment, Spain’s first multi-launch coaster. The park positioned it as a generational leap. And it has an estimated CPH of ~600 — the lowest of any major attraction in this study.

With 3 trains of 12 seats and a cycle time of approximately 111 seconds, the arithmetic is unforgiving: regardless of how spectacular the ride is, the system cannot process more than 600 people per hour. On an 18,000-visitor day, when that attraction receives the highest demand in the park simply by being the newest thing, the queue does not drop at any point during the day.

Stunt Fall (Vekoma Giant Inverted Boomerang) is a different but equally structural case: one train, a forward-and-back cycle of ~3.5 minutes, 450 CPH. The boomerang format has a throughput ceiling that cannot be raised without redesigning the block system — which would require a second train and a full operational rebuild. In practice, on a high-attendance day, Stunt Fall is an attraction that completes four or five useful cycles during the peak hour. It cannot operate as an open queue without Virtual Queue management.

Adding Batman: Gotham City Escape gave Warner’s portfolio a spectacular new headline. But by concentrating demand on a 600 CPH resource, it may have worsened the average guest experience on the park’s busiest days.

Terra Mítica — the park that solved its queue problem by losing visitors

Terra Mítica (Benidorm, Alicante) is the most singular case in this benchmark. The park opened in 2000 with an investment of approximately €360 million and projections of 3 million annual visitors that were never reached. Today it receives around 800,000 visitors per year, across approximately 120 operating days. It does not appear in the TEA/AECOM ranking — the publication threshold is one million visitors.

The single most important fact in its recent history: Magnus Colossus, the only wooden roller coaster ever built inside a quarry, closed in 2015. The park has spent ten years operating without its signature attraction, and has not replaced it.

With current data:

Estimated peak: 14,000 visitors
Peak-hour aggregate demand: ~7,000 PPH

Forty-two minutes on the main roller coaster, eleven on the second. Perfectly manageable queues for a high-attendance day.

Why? Not because Terra Mítica is well-designed for capacity. But because the park has lost enough visitors that its available throughput now covers the demand. Equilibrium was reached downward — by bringing demand down to the level of available CPH, not by raising CPH to meet the demand the park should have.

It is the case of a patient who no longer has a fever because they stopped eating. The queues are short because there are fewer people, not because the park is more capable.

The design lesson: losing a flagship attraction is not just an experience problem. It is a capacity problem. Magnus Colossus generated 1,280 CPH — 17 to 20 percent of the park’s total estimated throughput. That loss was recovered by no other investment. And ten years on, the park still operates without its wooden coaster.

Isla Mágica — one thrill coaster since 1997

Isla Mágica (Seville) operates approximately 125 days per year, receives around 800,000 visitors and has a declared maximum capacity of 15,500 people. Its footprint — 36 hectares — is nearly identical to that of the project in Asturias.

The entire thrill coaster portfolio consists of one element: Jaguar, a Vekoma SLC from 1997 with 1,010 CPH. It is the only thrill roller coaster the park has ever had. There is a log flume (Anaconda, Mack Rides), a shoot-the-chutes (Iguazú, Intamin) and a river rapids ride (Orinoco, Intamin). But the only coaster drawing thrill-seeking demand is Jaguar, now 29 years old.

Estimated peak: 12,000 visitors
Jaguar — estimated wait on peak day: ~94 minutes

Ninety-four minutes for the only coaster in the park. That is not a throughput problem — the Vekoma SLC at 1,010 CPH is a reasonably efficient model. It is a monopoly problem: when there is only one thrill attraction, all thrill-seeking demand in the park converges on that single point.

In October 2024 the park announced the construction of a GCI Titan Track wooden coaster: 25 metres tall, 595 metres long, a €2.7 million investment. With an estimated CPH of ~1,000–1,200, it would have given the park its second anchor attraction and redistributed about half the demand currently pressing on Jaguar.

In January 2026, the project was cancelled due to “early construction difficulties.” The park has no second thrill coaster. Isla Mágica’s 30th anniversary in 2027 arrives with Jaguar as the only thrill ride option.

Parque de Atracciones de Madrid — the park that can not invest

The Parque de Atracciones de Madrid (Casa de Campo) is the only park in this benchmark with no verified visitor data in the past five years. It is operated by Parques Reunidos under a concession from the Madrid city council (Ayuntamiento de Madrid), and Parques Reunidos does not disaggregate data by individual park. The last figure published by independent sources was 2019: 1,067,585 visitors. Revenue data from 2024 (€57.3 million, +5%) suggests a recovery approaching that level.

The park has 20 hectares — the smallest in the Spanish sample. With an estimated peak of 12,000 visitors, density at peak pressure is around 12 m² per person: manageable, but at the limit.

The active thrill attraction portfolio:

A note on Tornado: RCDB publishes a CPH of 1,500 that is implausible when calculated against the known configuration (2 trains × 24 seats, ~2-minute cycle). The realistic operational figure is around 700–800 CPH. The discrepancy matters because Tornado is the park’s primary anchor.

More important than any single number: Vértigo has been out of service since July 2025. The attraction — a Mack wild mouse with 1,120 CPH — was the park’s largest throughput generator per unit of ground space. Its loss, with no replacement announced, leaves the park’s aggregate CPH materially below where it needs to be for its visitor volume.

Estimated wait for Tornado (peak day, without Vértigo): ~120 min
Estimated wait for Abismo: ~76 min

The underlying business model explains why this is not easily resolved. The concession with the Ayuntamiento de Madrid expires in 2027. Parques Reunidos has no structural incentive to make major capital investments in an asset it may no longer operate in two years. The park has spent decades in a cycle of managing existing inventory rather than building its next chapter. Vértigo’s closure is a symptom of that dynamic.

The Parque de Atracciones de Madrid has the weakest recent investment ratio in the entire sample. Not through incompetence — through the logic of concession ownership.

The European references

So far, parks sharing the same problem at different degrees of severity: demand exceeding capacity on peak days. Now, the parks that have solved the problem differently.

Phantasialand — 28 hectares, the shortest queues

Phantasialand (Brühl, Germany) is the park with the highest theming density per square metre in Europe. It operates across 28 hectares — less than Terra Mítica, less than Isla Mágica, four times smaller than Europa-Park — and receives around 2.1 million visitors per year.

What makes Phantasialand different is not its size or its visitor count. It is the portfolio:

Average CPH across the four main anchors: 1,395.

Compared with Warner (average anchor CPH ~1,010, excluding Stunt Fall and Batman GCE) or with the Parque de Atracciones de Madrid (~750), the difference is nearly 40 percent in throughput. And that difference translates directly into shorter queues.

Estimated peak: 15,000 visitors

Empirical validation (queue-times.com and wartezeiten.app, August 2024): Taron records a peak-day average of 53 minutes in aggregate season data, with ranges of 60–90 minutes on the most crowded August Saturdays — the park operates at 89–99% of capacity on those days. The model predicts 45 minutes: here, the model understates real pressure by 15–45 minutes.

F.L.Y. records a peak-day average of 48 minutes, with ranges of 45–60 minutes in August. The model predicts 18 minutes: understated by a factor of three. The difference is explained by real operational loading friction — a flying coaster’s over-the-shoulder harnesses take systematically longer to secure than those of a conventional coaster, and that time does not appear in the declared CPH.

The lesson here is the inverse of Warner: the model understates when the declared CPH assumes a faster loading process than actually occurs in operation. F.L.Y.’s 1,440 CPH figure assumes efficient dispatch — but flying coaster harnesses have a systematically slower loading curve than any theoretical figure captures.

F.L.Y. deserves its own paragraph. It is the world’s largest flying coaster — a ride in which guests travel face-down, launched by LSM motors, through a circuit of unique design inside the Rookburgh themed area. And it has 1,440 CPH. That combination — world record for its format, first-tier throughput — is precisely what capacity engineering asks of a flagship attraction: be as spectacular as possible without sacrificing throughput.

The park also controls attendance through an online-only ticketing system introduced during the pandemic and retained since. On the highest-demand days, the daily allocation sells out before the park opens. This is not an administratively imposed restriction — it is a management decision to protect the quality of the experience. Phantasialand has turned a spatial limitation into a marketing argument: if you have a ticket, your visit is guaranteed without overcrowding.

Europa-Park — what happens when the manufacturer owns the park

Europa-Park (Rust, Germany) closed 2025 with more than 7 million visitors — the highest year in its history, in its 50th anniversary season. It is the second most-visited theme park in continental Europe, behind only Disneyland Paris.

The anomaly of Europa-Park that exists at no other park in this study: the Mack family, who founded the park in 1975, also owns Mack Rides — one of the world’s leading ride manufacturers. Of the park’s main identified attractions, nine of eleven are Mack Rides installations.

That is not a portfolio coincidence. It is vertical integration. And it has direct consequences for capacity engineering:

Blue Fire with 5 trains is the standout figure. A launched coaster designed from the outset to run five trains simultaneously can dispatch at a frequency that most attractions of this format never reach. With a declared CPH of 1,720, Blue Fire has higher throughput than PortAventura’s Shambhala (1,680), despite being a smaller-scale coaster.

Voltron Nevera (opened April 2024) was the largest investment in the park’s history: 1,385 metres of track, four LSM launches up to 100 km/h, seven inversions, the world’s first 105° launch angle. And 1,600 CPH — achieved through 7 trains and a moving loading platform that enables dispatches every ~36 seconds.

Compared with Batman: Gotham City Escape at Warner (similar investment, ~600 CPH) or Uncharted at PortAventura (~€25M, 900 CPH), Voltron’s figure demonstrates that with dispatch engineering designed into the ride from the beginning, a world-class attraction can also have world-class throughput.

The other Europa-Park anomaly: the park operates ~299 days per year, including a winter season with an integrated Christmas market event. That extended calendar is one of the reasons it can absorb 7 million visitors without peak days being more extreme than they would be with 200 operating days.

Efteling — the only European park with a legal visitor cap

Efteling (Kaatsheuvel, Netherlands) has something no other park in this study has: a legal ceiling of 5 million unique visitors per year, imposed by the Province of Noord-Brabant as a condition of its nature permit.

The reason has nothing to do with park engineering. The park borders the Loonse en Drunense Duinen National Park — a Natura 2000 protected area — and visitor traffic contributes to nitrogen emissions that threaten the ecosystem. The cap was established in 2016, the park exceeded it in 2017, 2018 and 2019, and in 2022 a district court ordered the Province to enforce it.

In 2025, Efteling received 5.78 million visits but only 4.98 million unique visitors — the park argues that the cap applies to unique visitors, not total visits, so a guest who comes twice counts once. The legal question is partially unresolved. The park’s CEO is hoping for a permit revision.

From a capacity engineering perspective, Efteling is notable for a different reason. Its attraction portfolio includes some of the highest CPH figures in Europe:

The Max & Moritz and Joris en de Draak figures need a clarification: these are two parallel tracks each. Max & Moritz are two Mack motor-driven trains running in competition; Joris en de Draak are two GCI wooden coasters (one water, one fire) racing against each other. The combined CPH of each pair is the sum of both tracks — meaning each attraction is, in throughput terms, the equivalent of two conventional coasters sharing the same ground footprint.

This is not a design accident. It is a deliberate throughput strategy: doubling capacity without doubling space, and adding an experiential dimension (the race, the duel) that improves the product at no additional land cost.

Symbolica (2017) illustrates a third strategy. It is an ETF trackless dark ride with 34 autonomous vehicles distributed across three different, randomly assigned routes within the same building. With 6 seats per vehicle and 34 units circulating, the system can have over 200 guests inside the building simultaneously. The estimated CPH of ~1,400 reflects a new vehicle entering the system every ~87 seconds — impossible to achieve with a single-track system and one vehicle per circuit.

Efteling operates 365 days per year without exception. It has no closing season. The resulting ADA — roughly 15,836 visits per day — is higher than any Spanish park in this study, and exceeded in Europe only by the large Paris, Orlando and Shanghai resort clusters. The park needs that aggregate throughput precisely because it never closes to recover.

Siam Park — a methodological aside

Siam Park (Tenerife), voted the world’s best water park by TripAdvisor for ten consecutive years, deserves a specific note because water park metrics are different from dry-park metrics.

In a theme park, CPH is the central metric: how many people can an attraction process per hour. In a water park, the primary bottleneck is the single-lane slide — not the wave pool, not the lazy river, not the swimming areas — because slides have sequential dispatch like a discrete queue.

The only publicly declared throughput figure at Siam Park is Saifa (ProSlide, opened 2023): 720 riders per hour, announced at the IAAPA trade show. Saifa is a dueling water coaster with two lanes — exactly the same logic as Joris en de Draak at Efteling: two parallel lanes that double throughput without doubling the queue space.

The most striking contrast is Tower of Power: the 28-metre freefall drop through a shark tank is the park’s absolute icon, the photograph shared across millions of social media posts, the element most responsible for Siam Park’s global reputation. And it has approximately 100–150 riders per hour as a single-lane body slide.

The park’s icon has the lowest throughput of any attraction in the park.

That is not a design error — it is a deliberate choice. Tower of Power does not exist to absorb demand; it exists to generate desire to visit. It does not move many people, but it makes everyone want to go to the park to see it. That is a completely different kind of value from throughput, and it is equally legitimate in portfolio design.

The distinction matters for the project we are building: some attractions exist to move people (throughput), and some exist to move conversations (iconicity). Confusing the two functions is a design error. An iconic low-throughput attraction without high-CPH support creates the park’s worst queue. An iconic low-throughput attraction with Virtual Queue management and surrounded by high-CPH alternatives creates the park’s best experience.

What the empirical validation teaches about the model

Before the cross-park conclusions, it is worth being explicit about what the comparison against real queue-times.com data reveals about the model itself.

The pattern is readable. The model performs well at PortAventura — the largest park in the sample, with the highest visitor volume and the lowest variance in per-attraction demand distribution. Where it systematically overestimates (Warner) is where the demand distribution across attractions is more heterogeneous: not every Warner visitor attempts Batman GCE, Superman and Stunt Fall on the same day, at the same intensity. The model treats them as if they do. Where it underestimates (Phantasialand) is where the declared CPH assumes a faster loading process than actually occurs — F.L.Y.’s harnesses take longer to secure than any theoretical figure captures.

This does not invalidate the model — it calibrates it. Three useful conclusions:

First: the model is more reliable for ranking attractions against each other (which has the worst relative queue) than for predicting absolute wait times. Batman: Gotham City Escape remains the worst-performing attraction in Warner’s portfolio in relative terms, even at 40 real minutes rather than 200.

Second: when the model predicts an inoperable queue (more than 120 minutes), the right response is not to trust the exact number — it is to treat that signal as an indicator that the attraction requires active demand management. That Uncharted has an active Virtual Queue and that observed queues are 90–120 minutes rather than 160 is not a refutation of the model: it is the model in action.

Third: declared CPH overstates real throughput whenever there is loading friction that the manufacturer’s figure does not reflect. Flying coaster harnesses, individual drop tower lap belts, dark ride safety briefings — all of that consumes cycle time without appearing in the technical specification.

A model that produces exact results is not a good engineering model — it is a model overfit to known cases. A model that produces correct orders of magnitude, identifies the real bottlenecks and fails in a systematic and understandable way is exactly what is needed for design decisions.

What the model says about all of them together

Applying the same lens to all nine parks, five patterns emerge consistently.

1. The newest attraction always has the worst queues

Without exception in this study. Batman: Gotham City Escape at Warner (2023): ~204 model-predicted minutes, ~40–53 observed. Danse Macabre at Efteling (October 2024): ~157 model-predicted minutes. Voltron Nevera at Europa-Park (April 2024): ~218 model-predicted minutes. Uncharted at PortAventura: ~160 model-predicted, 90–120 observed with VQ.

These attractions are not necessarily bad — novelty concentrates demand above equilibrium for one to three seasons, regardless of CPH. The only effective response is an active Virtual Queue system from the first day of operation.

2. Investment cost and CPH have no proportional relationship

Danse Macabre costs more than Batman GCE and has twice the CPH. Voltron Nevera is the group’s largest investment and also the best cost-to-throughput ratio among the large anchors. The attraction format — not the budget — determines CPH. Choosing a low-CPH format with a high budget is possible. That is what happened with Batman: Gotham City Escape.

3. A park with one thrill anchor is a mono-dependent park

Isla Mágica (Jaguar, ~94 estimated minutes at peak) and Terra Mítica before 2015 (Magnus Colossus as the only significant wooden coaster) both demonstrate this. When there is a single high-draw attraction, all demand from that visitor profile converges on one point. If the attraction has a breakdown on an August day, the park has no alternative for the thrill-seeking visitor. There is no plan B.

4. Park footprint is not the limiting factor — CPH is

Isla Mágica has 36 hectares and a peak density of 22 m² per person: space to spare. But it has 94-minute estimated queues on Jaguar. Phantasialand has 28 hectares — less space — and 60–90 minute real queues on its main anchor. The difference is not in the available land. It is in attraction CPH.

5. The two best guest-experience ratios have opposite solutions

Phantasialand: high CPH plus controlled daily caps. Queues are short because there are fewer people and high throughput — both simultaneously.

Terra Mítica: short queues because there are fewer people. Throughput is low, but demand is even lower. Equilibrium was reached downward.

The first is a replicable model. The second is what happens when a park loses its value proposition.

What this means for the project

The model in the previous chapter was not speculation — it was the same engineering that applies to parks that already exist. Validation against real data confirms the key parameters.

Three direct implications:

The first anchor attraction’s CPH cannot be below 1,200. Parks with 900–1,000 CPH anchors in the context of a 15,000-visitor peak day produce 90–120 minute queues. That does not destroy the park, but it defines its reputation in the first seasons. An anchor at 1,400–1,500 CPH, with well-executed operations, can stay below 60 minutes even on the worst day of the year.

Virtual Queue is not optional. Every new attraction — regardless of its CPH — will concentrate demand above equilibrium for the first one to three seasons. Without an active VQ system, the opening day attraction’s queue will define the park’s perceived quality. This cannot be remedied once the park is open.

Land ownership and concession horizon determine CAPEX for ten years. The Parque de Atracciones de Madrid and Isla Mágica illustrate what happens when the investment horizon is uncertain. A park that does not control its land, or that operates under a concession nearing expiration, has structural disincentives to invest. Decisions about the project’s legal and ownership structure are not a problem for the legal team to resolve later — they are a business design decision with direct operational consequences ten years out.

The next chapter develops the financial model: what investment this project requires, what revenue it can generate, and the point at which the model moves from hypothesis to viable case.

This series examines the feasibility of a theme park in Asturias from scratch, using the same tools that would apply to a real project. The models are estimates; the conclusions are for the reader to draw.

Published chapters:

• Chapter 1 — Could Asturias Have the Best Theme Park in Spain?
• Chapter 2 — Where to Build It
• Chapter 3 — What Kind of Park Should This Be
• Chapter 4 — The Worst Day of the Year
• Chapter 5 — The Stress Test (this article)