Heaven Magnetism

Declaration

I declare that this thesis is my own work, that every round described in it was played by imp, and that I don't rotate to B, when they go ramp.

Abstract

A large share of attacking players who take Ramp on de_nuke will, when the route to B is completely open, turn away from it and climb to Heaven to attack A instead. This thesis names the phenomenon Heaven Magnetism, measures its distribution across ranks, and demonstrates that it can be induced on demand. The method, referred to throughout as the ladder fake, consists of producing audio consistent with a defender rotating from A to B through the vents, then holding a position in front of the yellow panel on A site, which falls outside the standard clear performed by players entering from Heaven. Across several thousand rounds, the baseline rate of Heaven Magnetism follows a non-monotonic curve by rank: high at low ranks, reduced in the middle ranks, and high again at the professional level. Under the ladder fake, the curve flattens: the method succeeds in approximately 95% of attempts at every rank up to Master Guardian Elite (MGE), at which point it stops working almost entirely. The effect appears when a full attacking team is alive and in 1v1 situations alike, and it supports a passive variant in which the defender performs no fake at all and simply stands under Heaven. The thesis argues that this boundary, referred to as the MGE ceiling, marks the rank at which players stop merely reacting to information and begin evaluating it. Below the ceiling, players try to be smart but do not yet know what smart would be.

Keywords: de_nuke, Ramp, Heaven, route selection, vent audio, Master Guardian Elite

1. Introduction

On de_nuke, Ramp connects the attacking side of the map to bombsite B. It is a metal incline, it is the primary route to that site, and it runs downhill the entire way. Its function is not ambiguous, and neither, acoustically, is anything that happens on it.

This thesis documents a stable finding: attackers who fight for Ramp, win it, and face no resistance below will, at high rates, turn around and climb to Heaven to attack A, where the defense is concentrated. The behavior persists when B is observably clear. It persists when B has been announced as clear. The author has watched it persist for ten years and has stopped being surprised, which freed up the attention required to study it.

The thesis asks two questions. First, why do players decline an open bombsite that the map's architecture is actively delivering them toward. Second, whether the tendency can be produced deliberately, at a chosen time, by a single defender.

It makes three contributions. It characterizes the baseline rate of Heaven Magnetism across ranks and identifies the shape of its distribution. It introduces a reproducible method, the ladder fake, for inducing the behavior on demand. And it identifies the rank at which the method fails, approximately Master Guardian Elite, and offers an explanation for why the failure begins exactly there.

2. Prior Work

Community understanding of Ramp holds that the position can be denied for an extended period by one defender and one incendiary (Larsen, 2021), a fact that shapes the fears examined in Section 6. Work on entry clearing from elevated positions documents systematic coverage gaps when the clear is performed at speed (Tan, 2019). Measurement studies confirm that ladder audio in the A vent is clearly legible from Ramp (Okafor and Reyes, 2020), a finding this thesis depends on entirely. On decision quality, Novak (2022) distinguishes between reacting to information and evaluating it, and reports that the second skill arrives considerably later in rank than the first.

Professional play constitutes a separate literature. There, the Ramp take followed by an attack on A through Heaven is not an anomaly but standard theory, executed with utility and timed support from Hut and Squeaky. The professional case is addressed in Section 6.3.

3. Theory

Heaven Magnetism is modeled as a pull, F, acting on any player standing at the top of Ramp:

• A = the player's altitude preference
• H = the number of teammates already on the Heaven ladder
• B_c = the perceived safety of entering B
• μ = a map constant that takes an unusually high value on de_nuke

Two properties of the model matter. First, B_c falls as B becomes quieter. An open site does not read as an opportunity. It reads as a prepared one, so silence from B increases the pull toward Heaven rather than reducing it. Second, H compounds. Heaven is reached by ladder, the ladder admits one player at a time, and the first player to touch it sets H = 1 for the remaining four. Nuke is in practice played by one person and four trolls. The constant 1 in the numerator is the player's own pull, which Section 5.3 shows is sufficient on its own.

4. Method

Data collection took place over several thousand rounds of matchmaking and Premier. All trials were performed by imp while anchoring A.

4.1 Primary procedure: the ladder fake

The audio produced in steps 2 and 3 is consistent with a defender rotating from A to B through the vents, which is the correct play for a defender who has lost Ramp. This is the load-bearing property of the method and is discussed in Section 6.1.

The predicted enemy behavior is as follows. The attacking side, hearing the rotation, concludes that A is now open and that B has been reinforced. Both conclusions favor Heaven. The attackers ascend the Heaven ladder, which admits one player at a time and is fully audible, supplying the defender with both a warning and a head count. They perform a quick clear of A site from above, a clear which does not include the yellow panel, drop down, and move to plant in the far right corner under Heaven on the back of the site, facing the corner, taking cover from a retake that is not coming.

The remaining steps of the round are straightforward and are left to the reader.

4.2 Passive variant: under Heaven

A second position exists for rounds in which no fake is performed, or main and squeaky are still fought for. The defender takes position directly under Heaven and relies on the baseline magnetism documented in Figure 1. Players clearing from above do not check the space directly beneath themselves, and they resolve the omission by dropping into it, at which point the engagement begins at a range of approximately zero.

The yellow panel at the front of the site nevertheless remains the recommended position, for one reason: it allows the rotate quicker in the 5% chance they do not go heaven. A defender under Heaven has committed to the read almost completely. A defender at the panel retains the option of leaving if the attack develops elsewhere. One position assumes the read is right. The other allows for the possibility that it is not, and Section 6.2 argues that this is precisely the distinction the rest of the thesis measures.

5. Results

5.1 Baseline distribution

Figure 1 shows the baseline rate at which attackers choose Heaven over an open B, by rank, in rounds where no fake was performed.

5.2 Effect of the ladder fake

Table 1 reports success rates of the ladder fake by rank band. Success is defined as the attacking side entering A via Heaven and planting under it while imp holds the panel.

Table 1. The fake flattens the curve. The approximately 95% figure held across rank bands that differ in every other measurable respect, then collapsed at Master Guardian Elite. This boundary is referred to as the MGE ceiling.

⚠️ One observation was excluded from the dataset: a round in which the attacking side simply walked down to B. It was removed as an outlier and is not discussed further.

5.3 Team size

The phenomenon appears both when a full attacking team is alive and in 1v1 situations. It is very common in both cases and slightly more common in the 1v1. The 1v1 case carries theoretical weight because it sets H = 0 and isolates the intrinsic term of the model: a lone attacker, with no teammates to follow, an open B below, and the entire round resting on the decision, still takes the ladder. The pull does not require an audience. In the passive variant, such rounds end with the last member of the enemy team dropping from Heaven directly in front of the defender they did not check for.

6. Discussion

6.1 Why the fake works

The ladder fake succeeds because it is shaped like intelligence. Rotating through the vents after losing Ramp is the correct play for a defender, and the audio depicts that correct play precisely. Players who have learned to listen for information receive the sound as a reward for listening. The fake does not deceive the enemy so much as agree with them: it supplies a reason to do the thing the magnetism already wanted them to do. Information that licenses a desired behavior is accepted without inspection.

6.2 The central claim

Below MGE, players try to be smart but do not know what smart would be.

They complete the first step of good play, which is noticing information, and skip the second, which is asking where the information came from. The ladder fake functions as an instrument that separates these two steps and measures the gap between them. The gap closes at approximately Master Guardian Elite, which is why the method dies there and not one rank earlier or later.

6.3 The structure of the curve

At the lowest ranks the fake is unnecessary but harmless. Heaven Magnetism alone is sufficient, and the ladder audio is either not processed or processed as ambient texture. In the middle ranks the fake is decisive: these are players whose training has begun to suppress the magnetism, who know an open B should be entered, and who therefore need a reason to take the ladder instead. The fake provides one, and the suppressed behavior returns at full strength. At the ceiling and above, reasons are no longer free, and the panel gets checked. Professionals go to Heaven regardless, but from theory rather than pull: Ramp control threatens both sites, the attack from above produces elevation and trade geometry, and walking five players down a single incline into a set defense is understood to be a poor trade. The behavior at both ends of Figure 1 is identical. The reasoning is not. The low ranks are drawn by the magnet. The professionals use it.

6.4 The plant

In successful trials the attacking side planted in the far right corner under Heaven, facing the wall, in the belief that the nearest defender was two floors below them on B. The corner is chosen for cover against a retake from Heaven, which is reasonable, and it requires turning one's back on the yellow panel, which is not.

7. Limitations

All data are self-recorded, and imp is a known variable in every trial. Several classifications of "B was clear" were disputed at the time by teammates who were dead and watching. The professional band in Table 1 is untested because imp has not played against professionals and, having read Section 6.3, does not intend to. Finally, a small number of repeat opponents have begun pre-aiming the yellow panel on entry. Under the norms of the field this constitutes impact, and it is reported here as such.

8. Conclusion and Future Work

Heaven Magnetism is real, measurable, rank-dependent, and inducible on demand for the cost of one sprint and one touch of a ladder. Its exploitation fails only at the rank where players begin to interrogate their information rather than obey it.

The metal carries bodies downhill.
The soul takes the ladder.

Future work will extend the framework to other maps, beginning with the question of why every AWP purchased on Nuke is pointed at Outside within ten seconds of the buy time, whether it can be afforded or not.

References

• Larsen, M. (2021). Single-incendiary denial of Ramp: duration and morale effects. Journal of Utility Studies, 12(3).
• Novak, S. (2022). Reacting to information versus evaluating it: rank differences in audio trust. Annual Review of Matchmaking Behavior, 8.
• Okafor, D., and Reyes, P. (2020). Audibility of the A vent ladder from Ramp: a measurement study. Proceedings of the Workshop on In-Game Acoustics.
• Tan, W. (2019). The quick clear: coverage gaps in entry sweeps from elevated positions. Tactical Review Quarterly, 4(1).
• The author (2016 to 2026). Field notes. Unpublished, extensive.

Acknowledgments

The author thanks the teammates who anchored and died on Ramp during data collection, several of whom were told what was happening (afterwards). Thanks are due to sabo, who in 2019 advised imp to just go B, advice which is addressed in Section 5.1. Finally, imp thanks every player who has taken Ramp, heard their footsteps, and climbed up to heaven, then proceeded to die. This work would not exist without you.