How Far Apart Should You Be In Rocket League 2v2: What 10,488 Matches Actually Show
Rocket League Spacing Isn’t About Distance.
Pacifist Labs analysed 10,488 deduplicated 2026 Rocket League 2v2 matches to test a familiar coaching idea. How far should second man stay from first man? The answer is not a fixed number. The direction of the spacing mattered far more than the distance itself.
The published analysis uses the stricter, match-deduplicated dataset.
The accompanying Pacifist Labs video was produced using the broader replay-file export. For this published analysis, matches were additionally deduplicated using their Rocket League match GUID, preventing multiple uploads of the same game from being counted more than once.
The complete export contained 10,882 source replay files. Match-level deduplication reduced this to 10,698 unique matches, with 10,488 2026 matches forming the primary analysis across the six competitive cohorts.
This produces slightly different percentages and averages in some places compared with the video. Importantly, the central finding remains unchanged. Raw teammate distance becomes a weak predictor of winning at high rank, while greater longitudinal depth behind first man is consistently associated with winning across every level tested.
Two players can have the same total distance between them and still have completely different protection.
Second-man advice is often reduced to distance. Stay close enough to help. Do not get too close. Leave a few car lengths. Keep a fixed gap. Those rules sound measurable, but they collapse two very different shapes into one number.
A second man can be 3,000 Unreal Units from a teammate because they are safely behind the play, or because they are standing 3,000 units sideways from them. Both arrangements have the same straight-line distance. Only one necessarily preserves depth if first man is beaten.
Raw second-man distance stops predicting the winner at high rank.
The study first tested the simplest version of spacing. During stable second-man states, how far was second man from first man in a straight line?
At Platinum and Diamond there was a small advantage for greater distance. By Champion the effect had weakened. At Grand Champion, SSL and Professional level it disappeared completely.
How often the further team won
| Rank | Higher distance won | Winner difference | p-value | q-value |
|---|---|---|---|---|
| Platinum 1 | 54.2% | +51 uu | 0.000413 | 0.000516 |
| Diamond 1 | 53.1% | +41 uu | 0.000351 | 0.000438 |
| Champion 1 | 52.0% | +13 uu | 0.0666 | 0.0833 |
| Grand Champion 1 | 49.4% | −6 uu | 0.607 | 0.607 |
| Supersonic Legend | 50.1% | −11 uu | 0.297 | 0.297 |
| Professional | 48.1% | −13 uu | 0.170 | 0.170 |
SSL winners were actually 11 uu closer on average. Professional winners were 13 uu closer. Neither difference was statistically meaningful.
The same distance becomes powerful when we ask how much of it is behind first man.
Pacifist Labs then separated teammate spacing into geometry. Longitudinal gap measures second man’s depth behind first man along the team’s attacking axis. Positive values mean second man is behind. Lateral distance measures the sideways component.
This changes the result completely. The team with more longitudinal depth won more often at every tested level, with the strongest separation appearing in SSL and Professional play.
Distance alone cannot tell whether second man is protected
Depth-oriented spacing
Second man is separated mainly along the front-to-back attacking axis. If first man is beaten, another defensive layer still remains behind the play.
Lateral-heavy spacing
The players may have the same total distance between them, but much of that separation sits sideways. One forward action can still remove both layers.
Six levels. The same direction every time.
Longitudinal depth by rank
| Rank | Winner advantage | Higher depth won | p-value | q-value |
|---|---|---|---|---|
| Platinum 1 | +166 uu | 59.9% | 1.07 × 10−15 | 1.78 × 10−15 |
| Diamond 1 | +187 uu | 61.7% | 2.43 × 10−35 | 4.04 × 10−35 |
| Champion 1 | +151 uu | 58.5% | 1.53 × 10−24 | 2.54 × 10−24 |
| Grand Champion 1 | +154 uu | 60.6% | 2.28 × 10−23 | 3.80 × 10−23 |
| Supersonic Legend | +180 uu | 63.2% | 7.97 × 10−42 | 1.33 × 10−41 |
| Professional | +198 uu | 64.3% | 9.92 × 10−47 | 4.96 × 10−46 |
Higher straight-line spacing was essentially a coin flip. q = 0.297.
Higher depth behind first man strongly separated the winner. q = 1.33 × 10−41.
At high rank, winners were not only deeper. They were slightly less spread sideways.
Player-level V3 telemetry allowed the same second-man samples to be decomposed laterally. The pattern changes across skill level. Platinum winners showed slightly more lateral separation. By Grand Champion, SSL and Professional play, winners showed less.
The absolute lateral effect is smaller than the longitudinal effect. It should not be interpreted as an instruction to sit directly behind a teammate. The useful point is that elite winning geometry becomes increasingly depth-oriented rather than width-oriented.
The direction of the difference flips with rank
| Rank | Winner difference | p-value | q-value | Interpretation |
|---|---|---|---|---|
| Platinum 1 | +29.6 uu | 0.0231 | 0.0231 | Winners slightly wider |
| Diamond 1 | +7.8 uu | 0.327 | 0.327 | No clear difference |
| Champion 1 | −6.4 uu | 0.549 | 0.549 | No clear difference |
| Grand Champion 1 | −34.0 uu | 0.000315 | 0.000393 | Winners slightly narrower |
| Supersonic Legend | −37.6 uu | 3.58 × 10−7 | 4.48 × 10−7 | Winners slightly narrower |
| Professional | −24.3 uu | 0.0110 | 0.0138 | Winners slightly narrower |
The winner-minus-loser lateral difference moved downward as rank increased.
The exploratory linear rank trend returned p = 2.28 × 10−8. After correcting the three geometry trend tests together, q ≈ 3.41 × 10−8.
This does not establish one ideal lane. It supports the broader idea that high-level support gains value from depth rather than simply creating more sideways distance.
Elite players are not winning by stretching the string further.
The string does not simply get longer.
SSL and Professional teams were closer together in raw straight-line terms than the Diamond and Champion cohorts.
The useful part of the string points backwards.
Within those elite matches, the team maintaining more attack-axis depth behind first man was substantially more likely to win.
Zen, Nass and Vatira show why one spacing number cannot describe an elite style.
Population results tell us what separates winners. Named profiles show how different elite players can arrive at strong outcomes through different shapes. The current export contains 34 Zen profile matches, 35 Nass profile matches and 34 Vatira profile matches after match-level deduplication.
These profiles are descriptive examples for the public article. They are not used as named evidence in the peer-review analysis.
Second-man total spacing
Their total second-man spacing is remarkably similar. The shape inside that distance is not.
The explorer uses a metric-specific visual scale so elite differences remain visible. Exact values are always shown.
Zen was depth-oriented.
His mean longitudinal gap was 814 uu and he was behind first man in 69.1% of stable second-man samples.
Nass was more aggressive in this sample.
His total spacing was almost identical to Zen’s, but longitudinal depth was only 457 uu. Our earlier prediction that Nass would have the strongest second-man depth was not supported.
Vatira was not simply “too close.”
His total spacing was the largest of the three at 3,160 uu. The difference was that more of the separation sat laterally rather than behind first man.
One professional style can break a population rule and still be elite.
Nass and Vatira are not “wrong” because their geometry differs from Zen’s. Named profiles contain role, teammate, opponent and match-context effects. The population result tells us what was associated with winning across thousands of matches. The profiles show that elite players can distribute spacing differently.
As second man sat further behind the play, the win rate climbed sharply across the well-sampled range.
The rank-by-rank comparison asks which team was deeper inside the same match. A second analysis asks a different question. What happens when the average longitudinal gap itself is grouped into progressively deeper 200-uu bands?
Across the well-populated part of the distribution, the relationship is striking. The win rate rises from 41.4% at 400–599 uu to 72.4% at 1,600–1,799 uu. This is an observational relationship, not evidence that players should retreat indefinitely.
The well-sampled range
Every 200-uu band from 400–599 through 1,800–1,999 remained significantly different from a 50% result after correction.
The shallow bands are significantly below 50%, while the bands from 800–999 upward are significantly above 50%. The q-value tells us how strong the statistical evidence is, not how large the effect is. Because the number of observations falls sharply at deeper distances, q-values do not have to become smaller even while the observed win rate rises.
For each band, the significance test preserves the match pairing. It uses only discordant matches where exactly one of the two teams occupied that depth band, then asks whether that team won more or less often than expected by chance. Benjamini–Hochberg FDR correction was applied across all 15 displayed 200-uu bands.
Increase in the odds of winning for each additional 100 uu of longitudinal depth.
The association remained extremely statistically significant: p ≈ 2.0 × 10−161.
No. The depth association is strongest in SSL and the Professional-SSL subset.
A natural objection is that deeper second-man positioning might only work because lower-ranked players overcommit more often. The matched-game model does not support that explanation. The association between greater longitudinal depth and winning survives every level tested and is strongest at the top end of the ladder.
Winning odds per additional 100 uu of longitudinal depth. BH-FDR q = 8.8 × 10−47.
Winning odds per additional 100 uu of longitudinal depth. BH-FDR q = 1.7 × 10−30.
Win rate and statistical significance at every rank
| Depth | Platinum 1 | Diamond 1 | Champion 1 | Grand Champion 1 | SSL overall | Professional-SSL |
|---|---|---|---|---|---|---|
| 400–599 uu | 36.3%q 5.08×10−6 · n300 | 40.7%q 9.12×10−6 · n599 | 44.0%q .00148 · n727 | 43.1%q .00171 · n524 | 41.2%q 4.81×10−7 · n878 | 41.2%q .000226 · n478 |
| 600–799 uu | 51.7%q .582 · n391 · NS | 46.7%q .0743 · n733 · NS | 50.3%q .897 · n860 · NS | 50.1%q 1.000 · n665 · NS | 46.5%q .00822 · n1,240 | 46.5%q .0579 · n677 · NS |
| 800–999 uu | 49.1%q .777 · n436 · NS | 52.1%q .250 · n794 · NS | 52.8%q .131 · n767 · NS | 57.9%q .000148 · n570 | 56.8%q 2.56×10−6 · n1,117 | 56.0%q .00203 · n616 |
| 1,000–1,199 uu | 54.0%q .166 · n354 · NS | 56.9%q .000913 · n619 | 57.3%q .00148 · n527 | 61.1%q .000127 · n329 | 63.2%q 2.70×10−12 · n728 | 63.8%q 7.65×10−8 · n398 |
| 1,200–1,399 uu | 58.8%q .00668 · n262 | 60.6%q 4.64×10−5 · n406 | 62.1%q .000195 · n280 | 63.3%q .000963 · n180 | 73.4%q 1.82×10−17 · n357 | 73.4%q 5.67×10−11 · n218 |
| 1,400–1,599 uu | 60.7%q .0213 · n140 | 69.6%q 1.31×10−8 · n230 | 67.1%q .000108 · n149 | 71.7%q .00226 · n60 | 64.3%q .00620 · n112 | 68.9%q .00800 · n61 |
| 1,600–1,799 uu | 69.3%q .00226 · n75 | 72.2%q 5.81×10−5 · n97 | 74.5%q .00226 · n47 | 83.3%q .0579 · n12 · NS | 68.2%q .181 · n22 · NS | 69.2%q .327 · n13 · NS |
| 1,800–1,999 uu | 60.5%q .321 · n38 · NS | 53.8%q .809 · n39 · NS | 83.3%q .0579 · n12 · NS | 100%q .321 · n3 · NS | 71.4%q .532 · n7 · NS | 80.0%q .450 · n5 · NS |
| 2,000–2,199 uu | 88.9%q .00262 · n18 | 94.4%q .000460 · n18 | 100%q .0511 · n6 · NS | 100%q .574 · n2 · NS | 50.0%q 1.000 · n4 · NS | 50.0%q 1.000 · n2 · NS |
The rank-by-distance q-values are corrected across all 54 displayed comparisons.
Each cell preserves the winner-versus-loser match pairing. The q-value comes from an exact discordant-match test for that rank and depth band, followed by Benjamini–Hochberg FDR correction across all six competitive groups and all nine displayed distance bands. NS means the individual band did not remain below q = 0.05 after that correction.
The significant deeper-spacing signal survives into SSL and Professional play.
In SSL overall, the 800–999, 1,000–1,199, 1,200–1,399 and 1,400–1,599 uu bands all remain significant after the 54-test correction. The strongest SSL band is 1,200–1,399 uu at 73.4% wins, q = 1.82 × 10−17.
In the Professional-SSL subset, 800–999 through 1,400–1,599 uu also remain significant. The 1,200–1,399 uu band is 73.4% wins, q = 5.67 × 10−11. The observed win rates remain high at 1,600–1,999 uu, but those individual elite tail bands are no longer statistically significant because only 13 and 5 Professional-SSL team observations remain.
These are exploratory band tests, not fixed-distance prescriptions. The Professional-SSL line is a nested elite subset of the SSL-overall view for this depth-band analysis. High win percentages in tiny tail samples should not be interpreted as proof of a magic spacing distance. The continuous matched-game model remains the stronger test of the overall depth relationship.
No simple attack-versus-defence explanation removes the depth signal.
One obvious alternative explanation is that winning teams may simply attack more often. If attacking teams naturally spread out while defending teams are compressed near their own goal, greater second-man depth could appear to predict winning even if it was only a by-product of territorial position.
To examine that possibility, a separate state-level String Theory analysis used a stricter stable-role definition and compared winning and losing Professional teams within the same ball zones. The winning team still maintained more longitudinal depth in the defensive third, middle third and attacking third.
The winning team stayed deeper inside every third of the pitch
| Ball zone | Winner depth | Loser depth | Winner advantage | Higher-depth team won |
|---|---|---|---|---|
| Defensive third | 861 uu | 830 uu | +31 uu | 54.5% |
| Middle third | 1,932 uu | 1,845 uu | +88 uu | 58.1% |
| Attacking third | 2,568 uu | 2,444 uu | +124 uu | 58.5% |
Both teams are being compared inside the same broad territorial state. The result therefore cannot be explained simply by saying “the winner attacked more and the loser defended more”. The depth advantage remains visible after the play is separated into defensive, midfield and attacking zones.
Winner-minus-loser total teammate spacing in the defensive, middle and attacking thirds respectively.
Winner-minus-loser longitudinal depth across those same three ball zones.
Possession changes the correct spacing. It does not turn the finding into a fixed-distance rule.
The state-level analysis also estimated which team had the clearer ball-access advantage. When the opponent had the access edge, winning Professional teams positioned roughly 250–300 uu farther behind first man than when their own team had the access edge.
Relative to the normal ball-zone baseline, winning teams sat roughly 150–200 uu deeper under opponent access and roughly 75–125 uu shallower when their own team clearly owned the next touch. Good teams compressed when possession was secure and expanded their protective depth when the next state was less secure.
The target moves with the ball
| Ball zone | Central 50% longitudinal depth | Median | Median total spacing |
|---|---|---|---|
| Defensive third | 731–986 uu | 851 uu | 2,253 uu |
| Middle third | 1,699–2,147 uu | 1,919 uu | 2,867 uu |
| Attacking third | 2,242–2,817 uu | 2,516 uu | 3,311 uu |
The same adaptive pattern appears under pressure and during retreating play.
High nearest-opponent pressure shifted the winning Professional position roughly 100–150 uu deeper than the ball-zone baseline. When the ball was retreating, winning teams added roughly 100–200 uu of extra depth in the defensive and middle thirds. That retreating-ball adjustment largely disappeared in the attacking third.
Boost level did not produce a clean monotonic spacing rule, and ball height did not create a major consistent correction in this exploratory state analysis.
The result is a moving tether, not “sit as far back as possible”.
Field position, possession, pressure and the direction of the play all change how much depth is useful. The strongest interpretation is therefore not that one fixed distance causes winning. It is that protective longitudinal depth remains a winning signal after the obvious territorial explanation is separated out.
This remains observational evidence rather than proof of causation. A future multivariable state model can estimate the independent contribution of depth, possession, pressure and ball position simultaneously. But the current state-level results already make the simple explanation “winners only look deeper because they spend more time attacking” much harder to sustain.
What happens beyond 1,800 uu?
| Average longitudinal depth | Win rate | Team observations | BH-FDR q | Interpretation |
|---|---|---|---|---|
| 1,800–1,999 uu | 63.7% | 102 | 0.0135 | Significant, but tail is thinning |
| 2,000–2,199 uu | 89.6% | 48 | 2.93 × 10−8 | Significant, but very small sample |
| 2,200–2,399 uu | 63.6% | 11 | 0.748 | Too sparse for a target |
| 2,400–2,599 uu | 50.0% | 6 | 1.000 | Too sparse for a target |
| 2,600–2,799 uu | 50.0% | 2 | 1.000 | Extreme tail |
| 2,800–2,999 uu | 100% | 1 | 1.000 | Single observation |
| 3,000–3,199 uu | 100% | 2 | 0.748 | Extreme tail |
| 3,200–3,399 uu | 100% | 1 | 1.000 | Single observation |
The maximum recorded average longitudinal depth was approximately 3,304 uu.
That does not make 3,304 uu a coaching target. Above roughly 1,800 uu the number of observations falls rapidly, so individual bands become unstable and can swing sharply. The strongest evidence is the broad rise across the well-sampled range and the continuous matched-game association, not a single maximum-distance bucket.
Depth should still adapt to ball position, pressure, possession and the state of the attack.
The positional behaviour most often criticised as “too passive” is not showing up as a competitive weakness in the data.
The Pacifist System has long taught second man to preserve a protective layer behind first man rather than collapsing toward the play simply to remain close. One of the most common criticisms of that approach is that it can make players too hesitant or leave second man too far behind.
This study did not measure hesitation or passivity as psychological traits, so it cannot prove that a player “was not passive”. What it can test directly is the positional behaviour at the centre of that criticism. Across every competitive level tested, greater longitudinal depth behind first man was significantly associated with winning. The continuous association was strongest in SSL and the Professional-SSL subset rather than disappearing at elite level.
That matters because the evidence does not support treating deeper second-man positioning as inherently passive or inferior. At Professional level, winning support depth also expanded with ball position, with median longitudinal gaps of approximately 851 uu in the defensive third, 1,919 uu in the middle third and 2,516 uu in the attacking third.
For years, this principle came from coaching observation. Pacifist Labs now gives us a way to put that coaching claim under pressure with replay data. In this dataset, the result strengthens the principle rather than overturning it.
A directional relationship, not a tape measure.
Protect depth first.
When supporting first man, preserve enough front-to-back separation that one successful challenge or outplay cannot remove both teammates.
Let distance breathe.
The correct straight-line gap can contract or expand with ball speed, first-man control, pressure, boost and transition state. The data does not support one universal number.
Width needs a reason.
Lateral separation can create options, but at elite level more sideways distance was not the winning signal. Depth remained the stronger layer of protection.
Then adapt the exact distance to the state in front of you.
The spacing claim was tested inside the match, not inferred from rank averages.
Deduplicate the match
Replay files were collapsed by Rocket League match GUID so duplicate uploads of the same game could not count twice.
Identify stable roles
Second-man geometry was measured only during stable first-man and second-man states rather than every frame indiscriminately.
Pair winner and loser
Both teams from the same replay stayed linked, preserving the exact lobby, game state and match context.
Correct the tests
Paired Wilcoxon tests were followed by Benjamini–Hochberg false-discovery-rate correction.
Dataset composition
The complete V3 export contained 10,882 source replay files. Match-GUID deduplication reduced this to 10,698 unique actual matches. The primary String Theory analysis used 10,488 current 2026 matches across six mutually exclusive cohorts.
- Platinum 1 · 1,131 matches
- Diamond 1 · 2,050 matches
- Champion 1 · 2,033 matches
- Grand Champion 1 · 1,478 matches
- Supersonic Legend, recognised professionals excluded · 2,026 matches
- Professional · 1,770 matches
What longitudinal and lateral mean
Longitudinal gap is the second man’s front-to-back separation from first man along the team’s attacking axis. Positive values mean second man was behind first man. Lateral distance is the absolute sideways separation during the same stable second-man states. Straight-line spacing is the ordinary Euclidean teammate distance.
Statistical testing
The primary paired test was the Wilcoxon signed-rank test. The five geometry outcomes shown in this article were corrected together within each cohort using the Benjamini–Hochberg false-discovery-rate procedure. A corrected q-value below 0.05 was treated as statistically significant.
The reported “higher metric won” percentages exclude exact ties. Missing values were excluded only from the metric requiring them and were never converted to zero.
The exploratory 200-uu depth-band q-values use an exact paired discordant-match test. For a given band, only matches where exactly one team occupied that band contribute to the significance test; matches where both teams occupied the same band are uninformative for that contrast. The resulting 15 p-values, covering 400–599 through 3,200–3,399 uu, were corrected together using Benjamini–Hochberg FDR. These q-values test whether occupying a particular band was associated with winning or losing relative to the opposing team. They should be read alongside the continuous matched-game model rather than as evidence for a single optimal fixed distance.
The Professional state-level sensitivity analysis is separate from the match-level depth-band analysis. It uses one-second stable-role states with both teammates behind the ball, one stable first man and one stable second man, and a stricter involvement definition. Ball position is oriented by the team’s attacking direction and divided into defensive, middle and attacking thirds. Ball-access advantage, nearest-opponent pressure and retreating-ball state are then used as contextual modifiers. These analyses are exploratory and are presented to test whether obvious game-state explanations can account for the match-level association.
Named professional profiles
Zen, Nass and Vatira are used only as public-facing descriptive examples. Their profiles were aggregated by stable professional identity after match-level deduplication. No individual professional needs to be named in the anonymised manuscript analysis.
Why this is not proof of causation
This is an observational replay study. Greater longitudinal depth may help preserve defensive coverage, but it can also reflect game state, teammate behaviour, pressure, possession or other connected decisions. The repeated within-rank association identifies a strong competitive signal. It does not by itself prove that artificially increasing one number will cause a victory.
Large-scale exploratory analysis.
This public article uses named professional profiles for illustration. The peer-review version will keep professional players anonymous and treat the Professional cohort at population level.
The result changes the rule. It does not create a magic distance.
The analysis is observational. Directional depth is associated with winning and is not yet proven to cause it.
Whole-match averages can hide different states. Counterattacks, controlled possessions and emergency defence may require different spacing.
No corrected analysis supported one universal straight-line distance target. The current evidence favours adaptive geometry rather than a fixed number.
Named professional samples remain small compared with the population study and should be read as examples rather than rankings of who “rotates best.”
Lateral spacing showed a smaller effect than longitudinal depth. Less width is not automatically better, especially when width creates a passing or coverage option.
The study has not yet undergone independent peer review. Methods and limitations are published so the findings can be challenged and refined.
Stop asking for the perfect gap. Start asking whether you still have a layer behind the play.
Second man does not need to stand a fixed number of units from first man. Elite matches show why. SSL and Professional winners were not further away overall. They were more effectively separated along the direction that preserved depth.
String Theory therefore becomes less rigid, not more complicated. The string stretches and contracts with the game. What matters is whether its useful part still sits behind first man when the play breaks.
Adapt the gap.
Stay playable.
The original idea was that spacing should change with the state of the game. The new data makes the reason more precise.
A good theory should become more accurate when better evidence arrives.Learn the complete framework. Then measure it in your own replays.
The Digital Edition teaches the complete Pacifist System. Pacifist Score then measures how the same positioning, role and decision-making principles appear in your own Rocket League 2v2 matches.
