What Actually Wins In Rocket League 2v2: What 2,234 Replay Reports Actually Show
For years, one piece of advice has dominated Rocket League coaching, Play faster.
Stop hesitating. Stay closer to the play. Apply more pressure. Keep up with the lobby.
On the surface, that advice appears completely reasonable. Higher-ranked Rocket League players genuinely are faster. They travel at higher average speeds, spend more time supersonic and operate inside a more compact game.
But that leads to a far more important question:
Within the same rank, does the faster team actually win?
This is probably the most important piece of Rocket League research I have released in the past six years. Instead of relying on selected clips, ranked anecdotes or whichever opinion receives the most upvotes, Pacifist Labs compared the winning and losing teams from thousands of real Rocket League 2v2 matches.
The results challenge some of the most common assumptions about improvement, and provide the strongest evidence so far for one of the central principles behind the Pacifist System®.
Better players are faster. But faster teams within the same rank do not consistently win.
What was analysed?
The complete export contained 2,234 paired replay reports.
Not all 2,234 reports were placed into the same rank graph. The main current-rank analysis used 1,080 precisely classified games from 2026:
| 2026 cohort | Replays |
|---|---|
| Platinum 1 | 242 |
| Diamond 1 | 195 |
| Champion 1 | 168 |
| Grand Champion 1 | 186 |
| Supersonic Legend | 183 |
| Professional 2v2 | 106 |
| Total | 1,080 |
Another 203 reports came from 2024 SSL and Professional play and were used for the historical comparison later in this article.
The remaining 951 reports were historical, unranked or more broadly classified. They remained useful for pooled analysis, including the overall Pacifist Score result, but were not quietly added to precise individual-rank claims.
Every match was treated as one paired observation, the winning team compared directly with the losing team from that exact replay.
This matters because the four players inside one match are not four independent experiments. They experienced the same opponents, game duration, score state, overtime and sequence of events.
How statistical significance was tested
A statistical test produces a p-value, often described as a probability value.
The analysis begins by assuming there is no genuine difference between the winning and losing teams. The p-value then asks:
If there genuinely were no difference, how often could normal random variation still produce a pattern this convincing?
A p-value of:
- 0.05 means 5%, or approximately one in 20.
- 0.001 means 0.1%, or approximately one in 1,000.
- 0.0001 means 0.01%, or approximately one in 10,000.
This is not the probability that the conclusion is wrong. It is also not proof that one behaviour directly caused the result. It tells us how difficult the observed pattern would be to explain through ordinary sampling variation under a no-difference model.
The primary test was the Wilcoxon signed-rank test. It examined the paired difference between the winner and loser in every match and tested whether those differences consistently leaned in one direction.
A paired t-test was also used as a sensitivity check. This tests whether the average winner-minus-loser difference is meaningfully different from zero.
Wilcoxon was chosen as the primary test because replay statistics can be skewed and influenced by unusual matches. The paired t-test provided a second opinion based on the mean.
Because dozens of metrics were tested, the raw p-values were then adjusted using the Benjamini–Hochberg false-discovery-rate correction. These adjusted values are reported as q-values.
Throughout this article:
A result is described as statistically significant only when its corrected q-value remained below 0.05.
The tables generally display medians because a few unusual replays can drag the mean away from the typical match. This does not mean the mean was ignored; it was examined through the paired t-test. As the dataset grows, future reports will display both more frequently.
Missing metrics were excluded from that specific comparison. They were never treated as zero.
Finding one: Better players really are faster
The conventional advice initially appears correct. Median team speed increased steadily from Platinum through SSL.
| Rank | Median average speed | Median supersonic time |
|---|---|---|
| Platinum 1 | 1,353.5 | 8.50% |
| Diamond 1 | 1,397.5 | 9.83% |
| Champion 1 | 1,438.8 | 11.73% |
| Grand Champion 1 | 1,480.8 | 14.70% |
| Supersonic Legend | 1,524.6 | 17.95% |
| Professional | 1,515.7 | 17.85% |
Supersonic time more than doubled between Platinum 1 and SSL.
Higher-level players read situations earlier, recover more efficiently and execute actions at a greater pace. Speed is clearly one of the capacities required to play high-level Rocket League.
However, SSL players are also better at mechanics, positioning, challenge selection, ball control and boost management. Comparing an SSL player with a Platinum player cannot isolate whether speed itself is what makes one team win.
To answer that question, winners and losers must be compared within the same level.
Finding two: The faster team did not consistently win
The percentage below shows how often the team with the higher average speed won, excluding exact ties.
| Rank | Faster team won | Raw p-value | Corrected q-value |
|---|---|---|---|
| Platinum 1 | 55.8% | 0.0749 | 0.1498 |
| Diamond 1 | 47.7% | 0.7227 | 0.8399 |
| Champion 1 | 48.8% | 0.9678 | 0.9908 |
| Grand Champion 1 | 45.2% | 0.1304 | 0.3070 |
| Supersonic Legend | 46.4% | 0.4563 | 0.5607 |
| Professional | 50.9% | 0.4851 | 0.5617 |
Average speed was statistically significant at none of the six levels tested.
That does not prove speed is useless. It means the current data did not provide sufficient evidence that simply being the faster team consistently separated winners from losers within these ranks.
These are two different conclusions:
- Higher-ranked players are faster.
- Faster teams within the same rank do not consistently win.
Confusing those findings is how sensible advice such as “you must become capable of playing faster” turns into the far less useful instruction to “do everything faster.”
Speed is a capacity. It is not a universal decision rule.
Pacifist Score was the strongest non-scoreboard separator
Before examining the individual behaviours, I tested the thing I had actually built, Pacifist Score.
Across 2,156 replay pairs with complete composite data, the team with the higher Pacifist Score won 67.76% of non-tied matches.
| Pacifist Score result | Value |
|---|---|
| Valid paired reports | 2,156 |
| Higher Pacifist Score team won | 67.76% |
| Winner median | 74.91 |
| Loser median | 71.23 |
| Raw p-value | 2.06 × 10⁻⁷³ |
| Corrected q-value | 1.81 × 10⁻⁷² |
Outside the obvious statistics directly connected to the scoreboard, goals, normal scoreboard points, assists and shots, Pacifist Score was the strongest single winner separator in the complete export.
| Metric | Higher team won |
|---|---|
| Goals | 99.8% |
| Normal scoreboard points | 88.1% |
| Assists | 77.4% |
| Shots | 68.3% |
| Pacifist Score | 67.8% |
| First-Man Score | 64.4% |
| Behind Ball | 64.3% |
That is extremely encouraging, but Pacifist Score is a composite. It combines several behaviours and is not an independent experiment proving that every component directly causes victories.
I therefore removed the composite from the central argument.
The more important question was, which raw behaviours survived underneath it?
Finding three: Behind Ball was significant at every rank
One metric repeatedly separated winning and losing teams:
Behind Ball percentage.
Behind Ball measures how often a player remains on the safer side of the ball relative to their own goal.
It does not mean sitting in the net, refusing to attack or automatically retreating whenever the ball moves forward. It describes the player’s underlying positional relationship with the ball and the next state of the match.
The higher Behind Ball team won:
| Rank | Higher Behind Ball team won | Winner median | Loser median | Raw p-value | Corrected q-value |
|---|---|---|---|---|---|
| Platinum 1 | 64.9% | 72.08% | 69.75% | 0.0000000976 | 0.000000656 |
| Diamond 1 | 65.1% | 72.30% | 69.70% | 0.0000000395 | 0.000000294 |
| Champion 1 | 57.7% | 71.90% | 70.28% | 0.00901 | 0.03893 |
| Grand Champion 1 | 65.4% | 72.60% | 70.45% | 0.00000174 | 0.00000849 |
| Supersonic Legend | 63.2% | 74.20% | 72.25% | 0.0000105 | 0.0000509 |
| Professional | 64.2% | 75.05% | 72.83% | 0.000555 | 0.00256 |
The relationship appeared in the same direction at all six levels.
Every result remained statistically significant after correcting for the many metrics tested.
Average speed: Zero of six ranks significant.
Behind Ball: Six of six ranks significant.
Champion 1 produced the weakest result, but even there, the relationship survived the corrected significance threshold.
The SSL result did not merely scrape through
For Behind Ball at SSL, the raw Wilcoxon p-value was:
0.0000105
That equals 0.00105%.
Under the assumption that there genuinely was no systematic Behind Ball difference between winning and losing SSL teams, a pattern at least this convincing would appear approximately once in 95,000 comparable samples through ordinary random variation.
After correcting for all the other metrics tested, the q-value was still:
0.0000509
The significance threshold was 0.05. The SSL result was not close to that line.
This does not prove that artificially raising Behind Ball percentage will directly cause a player to win. But it makes the position that Behind Ball “does not matter” substantially harder to defend.
Correlation is not causation – but it is not an eraser
Statistical significance does not prove causation.
That sentence is important, but it is frequently misunderstood.
Consider children who are regularly read to. They generally develop stronger reading and language skills. However, one observational dataset cannot prove that reading alone caused the entire difference.
Those families might also have more books, more available time or different educational backgrounds. The children may already have a greater interest in language.
You cannot create thousands of otherwise identical childhoods, change only whether the child is read to, and hold every other experience constant for years.
That does not mean we conclude that reading probably has no effect.
We look for repeated relationships, plausible mechanisms, intervention evidence and different forms of evidence pointing in the same direction.
The same principle applies here.
Several mechanisms could contribute to the Behind Ball relationship:
- Remaining behind the ball may preserve defensive coverage and reduce overcommits.
- Better decision-makers may naturally produce both higher Behind Ball percentages and more victories.
- Teams protecting a lead may spend more time in safer positions.
- Behind Ball may capture several related behaviours, including recoveries, challenge selection and second-man discipline.
This observational analysis cannot perfectly isolate those possibilities.
But “correlation is not causation” is the beginning of the investigation. It is not a phrase that makes repeated evidence disappear.
Finding four: The 2024 data was correct – but incomplete
The Professional results provide a possible example of how competitive metas evolve.
In the 2024 Professional cohort, the faster team won 68.6% of non-tied matches. Speed was highly significant after correction.
| 2024 Professional metric | Higher team won | Winner median | Loser median | Raw p-value | Corrected q-value |
|---|---|---|---|---|---|
| Average speed | 68.6% | 1,533.2 | 1,503.3 | 0.0000551 | 0.000136 |
| Supersonic time | 66.7% | 19.1% | 16.4% | 0.0000257 | 0.0000674 |
| Average boost | 74.4% | 55.25% | 50.90% | 0.0000227 | 0.0000637 |
| Behind Ball | 69.5% | 75.20% | 72.80% | 0.0000105 | 0.0000478 |
If Pacifist Labs had existed in 2024 and this was the only evidence available, the obvious conclusion would have been Play faster.
That conclusion would not have been stupid. It would have been supported by the data available at the time.
However, in the 2026 Professional cohort, the faster team won only 50.9% of non-tied matches. Its corrected q-value was 0.5617, providing no significant winner-versus-loser relationship.
| Professional cohort | Faster team won | Corrected q-value |
|---|---|---|
| 2024 | 68.6% | 0.000136 |
| 2026 | 50.9% | 0.5617 |
This comparison does not prove that the professional meta changed. The players, competitions, replay sources and sample composition may all contribute.
But there is a plausible competitive explanation.
When very few players possess a skill, that skill can be a differentiator. Once everybody learns it, the skill becomes an entry requirement.
Professional players did not become slow between 2024 and 2026. They remain extraordinarily fast. Speed may simply have stopped separating them because it became widely adopted.
Yesterday’s competitive advantage becomes tomorrow’s minimum requirement.
The 2024 result was not necessarily wrong.
It was incomplete.
Available speed may matter more than permanent speed
The 2026 Professional boost results provide another clue.
| 2026 Professional metric | Winner median | Loser median | Corrected q-value |
|---|---|---|---|
| Time at zero boost | 12.78% | 13.90% | 0.02693 |
| Time completely full | 13.83% | 12.43% | 0.00843 |
| Time at 75–100 boost | 30.20% | 28.30% | 0.02693 |
| Average boost | 53.73% | 51.83% | 0.05166 |
Professional winners spent significantly less time completely empty, more time completely full and more time in the 75-to-100 boost range.
Average boost narrowly missed the corrected threshold.
The advantage may not be using more boost to remain permanently fast. It may be retaining the ability to become fast when the match actually demands it.
Constant speed and available speed are not the same thing.
Central positioning: Current result or future advantage?
The broad Central Lane metric did not consistently separate winners across ranks.
| Rank | Higher-central team won | Raw p-value | Corrected q-value |
|---|---|---|---|
| Platinum 1 | 55.4% | 0.00127 | 0.00329 |
| Diamond 1 | 50.8% | 0.6225 | 0.8224 |
| Champion 1 | 57.5% | 0.0769 | 0.1837 |
| Grand Champion 1 | 45.9% | 0.7771 | 0.8141 |
| Supersonic Legend | 51.9% | 0.6791 | 0.7300 |
| Professional | 52.8% | 0.7158 | 0.7325 |
Only Platinum remained statistically significant after correction.
The honest current conclusion is therefore:
Whole-match central positioning is not yet a replicated, universal winner separator.
That does not mean every principle of central support was disproved.
The current metric measures positioning across the entire match. The Pacifist System does not teach players to maximise centrality at all times.
Sometimes wide is correct. Sometimes central is correct.
The tactical question is whether the second man moves centrally at the useful moment:
- while remaining safely behind the play;
- while opening a passing lane;
- while covering both possible directions;
- while accessing useful small pads;
- and without crowding the first man.
A player can be central and useful. They can also be central and directly in their teammate’s way. Both situations can increase a broad whole-match central number.
Situational central support therefore requires a more precise future metric.
What if central support is earlier in the meta cycle?
This is where the 2024 speed result becomes important.
Had we stopped with the 2024 data, speed would have looked like the answer. Two years later, speed may have become so widely adopted that it no longer separates Professional winners in the same way.
So what if useful central support is earlier in that cycle?
What if a small number of exceptional players are already exploiting it before the broader population has learned to use it consistently?
That is not a confirmed result. It is a hypothesis.
But “not significant today” does not mean “can never matter.”
A behaviour can appear in the future of elite play before it becomes visible in average winner-versus-loser statistics.
Real-world cases also justify further study. Sparrow, for example, moved from Champion 2 to high Grand Champion 2 Division 3 in approximately three to four months through solo queue while applying the complete system.
That case does not prove centrality caused the improvement. The Pacifist System also changes roles, challenge selection, recoveries, spacing and decision-making. It does show why the complete intervention deserves longitudinal before-and-after analysis rather than being reduced to one broad metric.
What does Zen actually do differently?
Population averages can hide what the best individual players are doing.
Perhaps Zen separates himself by being dramatically faster, staying permanently closer to the ball or applying constant pressure.
The current 2026 export contains 45 Zen matches:
- 26 ranked games;
- 19 private or tournament games.
His speed and distance profile is surprisingly normal for an elite player.
| Metric | Zen | Non-Zen SSL | Non-Zen Professional |
|---|---|---|---|
| Average speed | 1,535.9 | 1,526.5 | 1,518.7 |
| Supersonic time | 17.9% | 17.9% | 17.6% |
| Distance to ball | 2,413.6 | 2,408.8 | 2,394.8 |
Zen is fast, but not separated from the elite baseline by an extraordinary speed gap.
His supersonic time is identical to the non-Zen SSL median.
His average distance from the ball is almost identical to the elite populations around him.
Zen is not simply driving faster than everyone else, and he is not glued to the ball.
Behind Ball produces a very different profile.
| Cohort | Median Behind Ball |
|---|---|
| Platinum 1 | 71.0% |
| Diamond 1 | 71.1% |
| Champion 1 | 71.1% |
| Grand Champion 1 | 71.8% |
| Non-Zen SSL | 73.5% |
| Non-Zen Professional | 73.7% |
| Zen | 75.8% |
The positional metric that separated winners at every rank reached its highest level in Zen’s current profile.
Zen was:
- approximately 3.1% higher than the non-Zen SSL baseline;
- approximately 2.9% higher than the non-Zen Professional baseline.
He was not simply the fastest or closest player.
He was exceptionally disciplined in his positional relationship with the ball.
Zen’s wider profile
The rest of Zen’s profile strengthens that interpretation.
| Metric | Zen | Non-Zen SSL | Non-Zen Professional |
|---|---|---|---|
| Behind Ball | 75.8% | 73.5% | 73.7% |
| Central Lane | 45.0 | 42.9 | 43.1 |
| Second-Man Score | 64.3 | 59.5 | 60.2 |
| Average boost | 57.7% | 52.1% | 52.0% |
| Teammate distance | 3,118.4 | 3,001.8 | 2,986.7 |
| Defensive Goal Side | 57.8% | 56.6% | 57.4% |
Compared with the non-Zen SSL baseline, Zen was approximately:
- 3.1% higher in Behind Ball;
- 4.9% higher in Central Lane;
- 8.1% higher in Second-Man Score;
- 10.7% higher in average boost;
- and 3.9% farther from his teammate.
His greater teammate distance is particularly important.
Zen does not appear to achieve structure by remaining permanently close to his teammate. He maintains slightly more separation while preserving an exceptional relationship with the ball and the defensive shape of the play.
Structure is not the same as proximity.
Centrality must still be interpreted cautiously. Champion 1 recorded a broad Central Lane median of 45.6, slightly above Zen’s 45.0, so this is not a simple “higher number equals higher rank” ladder.
However, Zen was noticeably more central than the elite SSL and Professional baselines directly surrounding him.
Behind Ball and central positioning therefore belong in different categories of evidence:
Behind Ball
- Present at an elite level in Zen;
- associated with winning at every rank;
- significant after correction at all six levels;
- strong current evidence.
Central support
- present in Zen’s elite profile;
- supported by a plausible tactical mechanism;
- not yet a consistent population-wide winner separator;
- an important future research question.
Behind Ball may be showing us what already works.
Zen’s central support may be showing us what needs to be tested next.
Limitations
This is a large observational replay analysis, not a randomised experiment.
The findings should be interpreted with several limitations in mind:
- Association is not causation. Higher Behind Ball may contribute to winning, result partly from protecting a lead, or reflect better decision-making underneath the metric.
- The 2024 and 2026 Professional cohorts are not identical. Different players, events and replay sources may contribute to the apparent change in speed.
- Zen is one player. His 45-match 2026 profile is an individual case series, not a population estimate. It is useful for identifying elite patterns, but it cannot prove those patterns cause his success.
- Whole-match metrics can hide context. Central positioning may be useful in specific second-man states without producing a stronger whole-match central average.
- Pacifist Score is a composite. Its strong winner association does not prove every component independently causes victories.
- The higher-metric win percentages exclude exact ties. Statistical testing used the full paired differences according to the relevant test procedure.
- Not significant does not mean no relationship exists. It means the evidence in this sample did not clear the corrected significance threshold.
- The study has not yet undergone independent peer review. The methods, definitions and limitations are being published so the work can be challenged and improved.
These limitations do not make the data meaningless. They define what the current evidence can and cannot support.
What should Rocket League players do with this?
The results do not suggest playing slowly. However, it could be argued that playing slowly, will allow a greater depth and understanding of the game and allow you to add speed in, whilst keeping the important things in check.
Players moving from Platinum towards SSL will almost certainly need to read, recover and execute at a greater pace. Their average speed and supersonic time will naturally increase as their overall ability improves.
The mistake is treating speed as the decision itself.
“Go faster” does not tell a player:
- whether they should challenge;
- whether they should rotate out;
- whether their teammate is ready;
- whether the play is already dying;
- or whether accelerating will turn a recoverable position into a 2v1.
Behind Ball tells a stronger and more consistent story.
Winning teams repeatedly preserved a safer positional relationship with the ball. That may reflect better recoveries, stronger second-man discipline, fewer unnecessary cuts and a greater ability to attack without surrendering the next phase of the match.
The clearest practical conclusion is not:
Play slowly.
It is:
Play fast enough. Position better. Don’t rush.
Pacifist Labs will continue expanding this dataset, building more situational metrics and testing the Pacifist System® against the evidence.
If the next 10,000 replays show that one of my ideas is wrong, the idea will change.
The goal is not to find statistics that agree with me.
The goal is to get Rocket League right.
Want to understand your own Rocket League 2v2 positioning?
Digital Pacifist System® owners can analyse their replays with Pacifist Score and compare their decision-making, positioning and match structure in greater depth. Get the book to access Pacifist Score and understand your decisions in greater depth today!
