The Statistical Danger of Diving as Last Man in Rocket League 2v2.
Pacifist Labs analysed 1,142 ranked 2v2 replays to test a familiar piece of advice: do not dive as last man. Teams that dived more often did not clearly lose more matches. A closer look at individual decisions found a more specific pattern. Diving was linked to a greater risk of losing defensive cover, particularly when the ball was moving quickly toward the defenders’ own goal.
What we found, and what still needs checking.
When the teams had different dive counts, the team that dived more lost 52.3% of matches (p = 0.288). That was not a clear departure from an even split.
In similar situations with a fast ball moving toward goal, the adjusted estimate for losing cover was 20.2 percentage points higher after an early dive.
The 65% speed marker helps describe that pattern. It comes from these replays and needs checking on new matches before we treat it as a general rule.
The analysis uses fixed rules to identify dives. Coaches still need to check whether those rules consistently recognise the situations we mean by “diving as last man”. Until then, the findings apply to the situations those rules picked out.
What happens when the last man dives?
If your teammate is already ahead of the ball and you are the final defender, the familiar advice is to stay patient and shadow the play. A rushed challenge can leave nobody covering the goal.
A dive can also work. The last man might win the ball, force a fifty, disrupt the attack or create a chance at the other end. The challenge is understanding when that commitment helps and when it leaves the defence exposed.
The team that dived more did not clearly lose more often.
We began with a simple question. In matches played to completion, when one team had more recorded last-man dives than the other, did that team lose more often?
The team that dived more lost 52.3% of those matches. That was not clear evidence of a link between dive count and defeat. The 95% uncertainty interval ran from 48.4% to 56.2%, allowing for players who appeared in more than one match. The exact two-sided p-value was 0.288.
The results were close to an even split
| Match-level result | Value |
|---|---|
| Completed matches with unequal dive counts | 599 |
| Losses by the higher-dive team | 313 |
| Loss rate | 52.3% |
| 95% interval, allowing for repeat players | 48.4% to 56.2% |
| Exact two-sided p-value | 0.288 |
A full match includes successful challenges, mistakes, recoveries and many other decisions. We therefore looked more closely at what happened immediately after individual dives. Those later tests explore a different, narrower question. They do not turn this first result into evidence that more dives mean more defeats.
We identified the commitment before looking at the result.
The analysis looked for a specific situation. The first defender was already beyond the ball, leaving their teammate as the only player still on the goal side. An opponent was close to the ball, and the remaining defender made a sustained forward commitment toward it.
The last man did not have to touch the ball for the move to count. The decision was identified when the commitment began, so successful and unsuccessful dives were both included.
Where the players were before the challenge
The first defender is already beyond the ball
First man had to be at least 300 UU beyond the ball, roughly 2.5 Octane lengths farther forward. Last man had to remain at least 300 UU goal-side, roughly 2.5 Octane lengths behind it. UU means Unreal Units, the game’s distance measurement.
The opponent is closer and has recently touched the ball
An opponent had to be within 650 UU of the ball, roughly 5.5 Octane lengths. They also had to be at least 250 UU closer than the last man, roughly 2.1 Octane lengths, and have touched the ball in the previous two seconds. These rules suggest the opponent could contest the ball, but cannot tell us exactly how much control they had.
The spacing used to identify these situations
One Octane hitbox is approximately 118 UU long. The car lengths below help you picture each distance; the analysis used the exact UU values. Pitch percentages refer to the distance from midfield to the standard back wall.
| Distance used | What that looks like | What it means here |
|---|---|---|
| 100 UU | ≈ 0.8 Octane lengths≈ 2.0% of midfield → back wall | Last man must be over halfway |
| 200 UU | ≈ 1.7 Octane lengths≈ 3.9% of midfield → back wall | Ball must be over halfway |
| 250 UU | ≈ 2.1 Octane lengths≈ 4.9% of midfield → back wall | How much closer the opponent had to be |
| 300 UU | ≈ 2.5 Octane lengths≈ 5.9% of midfield → back wall | First man beyond the ball and last man behind it |
| 650 UU | ≈ 5.5 Octane lengths≈ 12.7% of midfield → back wall | Maximum opponent distance from the ball |
| 3,000 UU | ≈ 25 Octane lengths≈ 58.6% of midfield → back wall | Maximum last-man distance from the ball when the dive began |
| 5,120 UU | ≈ 43 Octane lengths100% of midfield → back wall | Midfield to the standard back wall |
The car had to meet each of these movement rules
| Movement rule | Exact value | What that looks like |
|---|---|---|
| Car speed | ≥ 1,000 UU/s | ≥ 43% of max car speed · about 36 km/h / 22 mph |
| Speed toward the opponent’s goal | ≥ 500 UU/s | ≥ 22% of max car speed in that direction · about 18 km/h / 11 mph |
| Closing speed on moving ball | ≥ 700 UU/s | Closing at ≥ 30% of max car speed · about 25 km/h / 16 mph |
| Movement alignment | Cosine similarity ≥ 0.55 | Travel direction within about 57° of the ball |
| Distance from ball | ≤ 3,000 UU | Within ≈ 25 Octane lengths≈ 58.6% of midfield → back wall |
Coaches still need to check the situations picked out by the analysis.
The rules identify the commitment before looking at what happened next. The remaining check is whether coaches agree that these were genuine last-man dives, with the teammate out of the play and an opponent in control. Reviewers will assess the lead-up without seeing the outcome, including whether shadowing was available. Until that review is complete, the results apply to the situations identified by the automated rules.
From completed, non-forfeit ranked 2v2 matches.
Across Platinum 1, Diamond 1, Champion 1, Grand Champion 1 and SSL.
Defensive cover was lost more often after a dive.
We first compared decisions across all ball speeds, allowing for differences in the situation before the challenge. A recorded dive was linked to a higher chance of losing defensive cover within two seconds. Here, cover means having at least one player on the goal side of the ball. “No early dive” means the analysis did not identify a dive in the first 0.8 seconds; it does not tell us exactly what the defender did instead.
Adjusted estimate within two seconds.
A difference of +14.8 percentage points.
What changed in the seconds after the decision
| Outcome | Dive | No early dive | Difference | Risk ratio | Adjusted p-value |
|---|---|---|---|---|---|
| Last man beaten within 2 s | 37.3% | 22.5% | +14.8 pp | 1.66 | 0.00022 |
| No goal-side cover within 2 s | 40.7% | 25.9% | +14.8 pp | 1.57 | 0.00070 |
| No cover while ball travels goalward | 31.2% | 19.6% | +11.6 pp | 1.59 | 0.00041 |
| Opponent touch during an uncovered attack toward goal | 8.3% | 2.7% | +5.5 pp | 3.02 | 0.01572 |
| Team touch within 3 s | 60.6% | 53.8% | +6.8 pp | 1.13 | 0.15261 |
| Goal against within 10 s | 8.0% | 8.4% | −0.4 pp | 0.95 | 1.00000 |
“Lost cover” is not shorthand for “conceded a goal.”
Moving forward can leave the defence exposed, so some loss of cover is expected when a player dives. The comparison measures how often that happened, including when the ball was moving toward goal and the opponent could touch it. A team can still recover or avoid conceding. These results do not establish whether diving increased goals conceded within ten seconds.
A dive can still win a touch.
The estimated chance of a team touch was slightly higher after a dive. However, that difference was not clear enough after allowing for the number of tests carried out. A possible touch-winning benefit could help explain why losing cover did not translate into a clear link with match defeats.
The faster the ball moved toward goal, the more often dives were followed by lost cover.
Ball speed showed the clearest pattern. We measured how quickly the ball was moving toward the defenders’ own goal when the dive began. This is only the part of its movement in that direction. A ball moving diagonally or rising into the air can have a higher total speed.
Rocket League’s normal maximum car speed is 2,300 UU/s. We use percentages of that speed to make the ball’s movement easier to picture. The tables also include the exact measurements in Unreal Units per second.
Ball not travelling goalward
Cover was sometimes lost even when the ball was not moving toward goal, but much less often than when it was travelling quickly toward goal.
Cover loss increased across the speed bands
| Ball movement when the dive began | Dives | No cover in 2 s | No cover + goalward | First touch in 2 s | Goal against in 10 s |
|---|---|---|---|---|---|
| Not travelling goalward | 197 | 28.9% | 17.8% | 40.1% | 5.7% |
| Up to 22% max speed 0–500 UU/s · 0–18 km/h / 0–11 mph | 350 | 30.9% | 19.7% | 47.1% | 6.2% |
| 22–43% 500–1,000 UU/s · 18–36 km/h / 11–22 mph | 519 | 43.7% | 33.9% | 48.9% | 7.3% |
| 43–65% 1,000–1,500 UU/s · 36–54 km/h / 22–34 mph | 436 | 49.5% | 39.4% | 46.8% | 8.6% |
| 65–87% 1,500–2,000 UU/s · 54–72 km/h / 34–45 mph | 212 | 59.4% | 52.4% | 40.6% | 16.5% |
| Above 87% >2,000 UU/s · >72 km/h / >45 mph | 93 | 83.9% | 78.5% | 20.4% | 25.9% |
Only 19 of those 93 dives produced the team’s first touch within two seconds. Among the 85 events with a valid ten-second follow-up, 22 were followed by a goal against and only one by a goal for.
About two thirds of these dives were followed by a complete loss of cover.
There were 305 dives with the ball moving toward goal at or above roughly 65% of maximum car speed. In 204 of them, neither defender remained on the goal side of the ball within two seconds.
That is 66.9%, with a 95% uncertainty interval of 61.3% to 72.4% after allowing for repeat players. Even the lower end of that interval was above 60% in this set of replays.
Fast balls compared with slower balls
| Outcome after a recorded dive | 65%+ toward goal | Slower / away | Risk ratio | 95% interval for ratio |
|---|---|---|---|---|
| No goal-side cover within 2 s | 66.9% | 40.5% | 1.65 | 1.49 to 1.82 |
| No cover while ball remains goalward | 60.3% | 30.1% | 2.00 | 1.78 to 2.24 |
| Team wins first touch within 2 s | 34.4% | 46.7% | 0.74 | 0.61 to 0.87 |
| Goal against within 10 s | 19.2% | 7.2% | 2.66 | 1.97 to 3.56 |
| Goal for within 10 s | 6.2% | 10.5% | 0.59 | 0.36 to 0.88 |
“Goal against within 10 seconds” records what happened after the dive. It does not show that the dive caused the goal. A fast ball moving toward goal is already dangerous before the defender acts.
Both groups in this table dived. The difference is how the ball was moving at the start. To learn more about the decision itself, the next section compares dives with other responses in similar situations.
Was the ball simply dangerous anyway?
Yes. A fast ball travelling toward your goal is already a threat. The 66.9% versus 40.5% comparison alone cannot tell us how much of the difference is linked to the defender’s decision.
We therefore compared situations where the last man dived early with similar fast-ball situations where no early dive was recorded.
The gap remained after allowing for the situation before the dive.
We excluded situations where the defender had already committed, then checked whether a dive began during the next 0.8 seconds. The statistical comparison allowed for information recorded before that decision, including rank, score, time, ball and player positions, movement, boost, distance to the opponent and recent touches.
Some situations almost always led to a dive, while others almost never did. We limited the comparison to situations where the model estimated a dive probability between 0.05 and 0.95, so there were examples of both responses to compare. This kept 1,913 of the 6,784 situations meeting the study’s rules. The results therefore apply to this smaller group, not every possible last-man decision.
Adjusted estimate for similar fast-ball situations.
Adjusted estimate, roughly 55 in every 100 similar situations.
For these similar situations, the estimated chance of losing cover while the ball moved toward goal was 55.4% after a dive and 35.3% with no early dive. The risk ratio was 1.57, meaning the estimated rate was 1.57 times as high after a dive. Its 95% uncertainty interval was 1.20 to 2.03, allowing for repeat players. The p-value was 0.00153, rising to 0.00763 after the Holm adjustment for the five outcomes tested. The link remained statistically clear, but players chose their own actions, so differences we could not measure may still help explain it.
Estimated outcomes from similar starting situations
| Outcome | Dive | No early dive | Adjusted difference | Risk ratio | 95% interval for ratio |
|---|---|---|---|---|---|
| No goal-side cover within 2 s | 57.2% | 41.0% | +16.2 pp | 1.40 | 1.07 to 1.76 |
| No cover while ball travels goalward | 55.4% | 35.3% | +20.2 pp | 1.57 | 1.20 to 2.03 |
| Opponent touch during an uncovered attack toward goal | 9.2% | 3.0% | +6.3 pp | 3.09 | 0.59 to 11.01 |
| Team touch within 3 s | 51.5% | 60.5% | −9.0 pp | 0.85 | 0.58 to 1.11 |
| Goal against within 10 s | 13.6% | 9.6% | +4.0 pp | 1.42 | 0.51 to 3.03 |
The goal result is still uncertain.
The estimated rate of conceding within ten seconds was higher after a dive, but the uncertainty interval was too wide to establish a reliable difference. The clearest finding concerns the loss of defensive cover immediately after the decision.
Diving may offer less chance of a touch when the ball is moving quickly toward goal.
Below the 65% speed marker, the estimated chance of a team touch within three seconds was 10.1 percentage points higher after a dive than with no early dive. In the faster group, the estimated difference was −9.0 points.
The difference between those two estimates was −19.2 points. The initial p-value was 0.022, but the result was no longer statistically clear after allowing for all five tests of how the outcomes changed with speed. This is a pattern worth checking on new replays, rather than firm evidence of a change at one particular speed.
The touch comparison is less certain than the cover-loss finding. It is a reason for further investigation, not a proven rule that the benefit of diving disappears at 65% speed.
Defenders with plenty of boost still lost cover often.
What happened when defenders had more boost
| Starting boost | Dives | No cover within 2 s | First touch within 2 s | Goal against within 10 s |
|---|---|---|---|---|
| Below 50 | 60 | 75.0% | 30.0% | 22.2% |
| At least 50 · half a tank | 245 | 64.9% | 35.5% | 18.6% |
| At least 80 · four fifths of a tank | 164 | 61.6% | 37.2% | 13.2% |
Players began with different amounts of boost for many reasons, so these figures do not isolate the effect of boost itself. They show that losing cover was still common among defenders with plenty available. The 80+ group is included within the 50+ group.
Pacifist aims to draw out the first defender and put the last man under pressure.
The attacking idea is to draw the first defender into a corner challenge, move past them, and send the ball toward goal before they can recover. That leaves the remaining defender temporarily alone.
This study looks at the last part of that sequence, when the first defender is already beyond the ball. It does not establish how often Pacifist corner play creates the situation in the first place. That needs a separate test.
Draw out the first defender
Use controlled possession to encourage the nearest opponent to challenge in the corner.
First man goes beyond the ball
Once that defender is beyond the ball, they need time to get back into the play.
Move toward goal quickly
Move the ball through the gap before the first defender can recover.
Attack the isolated defender
The last man has to decide whether to hold their position or commit before their teammate gets back.
This group includes several responses, such as shadowing, retreating, fake challenging or hesitating. It is not a group of confirmed successful shadow defenders.
Estimated loss of cover while the ball moved toward goal, compared with 35.3% with no early dive in similar fast-ball situations.
The Pacifist System aims to get past the defenders one at a time through controlled pressure. These results help explain the difficult decision facing the last defender. They do not yet show how reliably the full attacking sequence creates that opportunity.
Judge the challenge by the cover behind you and the ball coming toward you.
During play, look for three things. Is my teammate already beyond the ball? Does the opponent have close control? Is the ball travelling quickly toward our goal?
When all three are true, be more cautious about committing. The coaching recommendation is to shadow, delay and protect the centre while your teammate recovers. That advice draws on the cover-loss finding, but this study did not directly compare the effectiveness of those alternatives.
A guide to choosing your response
| Situation | What to consider | Coaching recommendation |
|---|---|---|
| First man still goal-side | Your teammate may still be able to cover | Coordinate your challenge with your teammate’s position |
| First man beyond ball, but ball slow or moving away | There may be a touch to win, but committing can still leave no cover | Challenge selectively when the touch is clearly available |
| First man beyond ball + ball moving rapidly toward goal | Lost cover was more common; the chance of winning a touch may also fall | Shadow, delay and protect the central route |
| Ball toward goal above 87% max car speed>2,000 UU/s · >72 km/h / >45 mph | 83.9% of recorded dives were followed by no cover | Keep the dive as a last resort unless you have a clear read |
When attacking, try to create that pressure for the opposition. Get past the first defender and move toward goal before they can recover.
The cover-loss finding is clear, but there are limits to what it can tell us.
Players chose their own actions. This was a study of existing replays, not an experiment assigning players different decisions. Skill, communication, what players could see and how they read the bounce may still help explain the results.
Only some situations could be compared fairly. The model’s 0.05–0.95 dive-probability range kept 1,913 of 6,784 situations. The adjusted results apply to that group, not every last-man decision.
No early dive does not necessarily mean shadowing. It means no dive was identified within 0.8 seconds. The defender could have retreated, fake challenged, hesitated or chosen another response.
Manual checks are still pending. A sample of 200 short replay sequences has been prepared for coaches to review without seeing the outcomes. They will check the opponent’s control, each defender’s position, whether shadowing was available and whether a genuine dive occurred.
The 65% marker needs further testing. Risk rose across the speed bands. This particular marker was chosen from the same dataset, so it should be checked on new replays before being used as a general rule.
There were too few goals for a clear conclusion. Losing cover was strongly linked to diving, but we could not establish a reliable difference in goals conceded within ten seconds.
The closer analysis was exploratory. The first test did not show that more dives meant more defeats. Later tests identified patterns around individual decisions. Those patterns need checking on new replays with the tests set out in advance.
The statistical correction has a specific scope. The Holm adjustment accounts for the five displayed high-speed outcomes together. It does not cover every model, speed band or subgroup explored during the study.
Some loss of cover is expected when a defender moves forward. That is part of what makes a dive a commitment. These figures describe the resulting exposure; they do not prove that the choice caused a goal.
What was tested and how.
Identify the decision first
Use the players’ positions and movement to identify a dive before looking at whether it worked.
Check the next two seconds
Measure whether defensive cover remained immediately after the decision, as well as checking later touches and goals.
Allow for repeat players
Shared players linked matches into 500 groups. These connections were kept together when checking the models and estimating uncertainty.
Allow for several tests
Testing several outcomes raises the chance of finding a pattern by chance. The Holm adjustment addresses this for the five displayed high-speed outcomes, not every analysis in the study.
Replays and decisions included
The dataset contained 1,143 ranked 2v2 replays, of which 1,142 could be analysed.
- Completed non-forfeit matches · 913
- Dives identified in completed matches · 1,807
- Situations meeting the study’s rules in non-forfeit matches · 6,784
- Situations with a dive in the early decision window · 556
- Situations retained for the main adjusted comparison · 1,913
- Ranks studied · Platinum 1, Diamond 1, Champion 1, Grand Champion 1 and SSL
How long we followed each decision
- The forward commitment had to last at least 0.12 seconds, roughly one eighth of a second or four analysed frames.
- An early dive had to begin within 0.8 seconds of the situation being identified.
- We checked for loss of cover within two seconds.
- We checked for a team touch within three seconds.
- We followed goals for ten seconds. Cases without enough usable replay footage were excluded from that goal calculation.
Speed measurement
We measured how fast the ball was moving along the pitch toward the defenders’ goal, rather than its total speed in every direction. The main reference point is roughly 65% of maximum car speed.
The main threshold is 1,500 UU/s = 65.2% of maximum car speed = 15 m/s = 54 km/h = 33.6 mph. Normal maximum car speed is 2,300 UU/s. At that goalward rate, the ball covers 3,000 UU in two seconds: approximately 25 Octane lengths or 59% of midfield-to-back-wall distance.
How we compared similar decisions
The data were divided into five groups, keeping matches linked by the same players together. Each group received predictions from models trained on the other four. This method, called cross-fitting, reduces the risk of judging a model on the same examples it learned from. Tree-based models using gradient boosting estimated the chance of a dive and each outcome from information recorded before the decision.
That information included rank, score, time, ball position and velocity, both teammates’ positions, boost, closing speed, movement direction, opponent distance and recent touches. We kept situations with an estimated dive probability of 0.05–0.95. The model distinguished between dive and no-dive situations well, with an AUC of 0.917. This showed how strongly the starting situation influenced the action, and why comparisons needed to be limited to situations where both responses occurred.
How we accounted for the same players appearing again
Players who appeared in more than one match linked the replays into 500 groups. We kept those groups together when checking the models and estimating uncertainty, because decisions involving the same players are not independent examples. For the speed comparison, we resampled the linked groups 5,000 times, a method called bootstrapping.
Measurement references
Rocket League’s normal maximum car speed is treated as 2,300 UU/s, maximum throttle-only driving speed as 1,410 UU/s, and the supersonic threshold as 2,200 UU/s. One Unreal Unit is one centimetre, and the standard back wall is 5,120 UU from midfield.
An Octane hitbox is approximately 118 UU long. Car lengths and pitch percentages are included to help readers picture the distances. The analysis used the exact UU measurements.
Which results still need confirmation
The first test did not establish a link between more dives and more defeats. The later analyses explored individual decisions, ball speeds and boost levels. These were not a separate pre-registered test, with a fixed plan for confirming the original claim.
The Holm adjustment accounts for testing several outcomes together within the groups reported. It does not cover every exploratory analysis carried out. The main high-speed cover-loss result remained statistically clear after this adjustment, but it still needs to be tested on new replays with the analysis planned in advance.
Why some situations were left out of the adjusted comparison
The model could often predict the defender’s action from the starting situation. Its AUC was 0.917; on this measure, a higher score means better separation between the recorded dive and no-dive cases. If almost everyone makes the same choice in a situation, there are too few examples of the alternative for a useful comparison.
We therefore kept cases with an estimated dive probability between 0.05 and 0.95. This retained 1,913 of the 6,784 situations meeting the study’s rules. The adjusted results apply to those comparable situations and should not be extended to all last-man decisions.
What “no early dive” includes
The comparison is a recorded early dive versus no recorded early dive. The second group can include shadowing, retreating, fake challenging, hesitating or another response. Those actions have not been individually checked and labelled by coaches, so we cannot treat the whole group as an example of good shadow defence.
The manual review still to be completed
A sample of 200 short replay sequences has been prepared for independent review. Coaches will assess whether the opponent had control, whether the first defender was out of the play, whether the remaining player was truly last man, whether shadowing was available and whether the movement was a genuine dive. They will not see the outcome while making those judgements.
The review has not yet been completed. Until we know how closely the automated rules agree with the coaches, the findings should be read as results for the situations those rules identified.
What the next study should test
The next study should follow the attack from the corner, rather than starting when the last defender is already alone.
Control the ball in the corner → get past the first defender → move toward goal quickly → track the last man’s response → check whether the defence is exposed.
Using a new set of replays would help test whether Pacifist reliably creates the opportunity described here, as well as what happens once that opportunity exists.
Common questions about the findings.
Does this prove that every last-man dive is wrong?
No. More dives did not clearly predict more match defeats. Diving may also help win a touch in some slower-ball situations. The clearest finding was a higher estimated rate of losing cover after an early dive in similar situations with a fast ball moving toward goal. That finding still needs testing on new replays.
Does 66.9% mean 66.9% of those dives caused a goal?
No. It means 66.9% were followed by a complete loss of goal-side cover within two seconds. The recorded rate of conceding within ten seconds in the same fast-ball group was 19.2%.
Did diving cause that 19.2% goal-against rate?
We cannot tell from that figure. The ball was already moving quickly toward goal. After allowing for differences in the starting situation, the estimated goal rate was higher after a dive, but too uncertain to establish a reliable difference. The clearer finding concerns the loss of cover.
Does high boost make the dive safe?
It does not guarantee safety. Among dives with a fast ball moving toward goal and at least 50 starting boost, 64.9% were followed by no goal-side cover within two seconds. At 80+ boost, the figure was 61.6%.
Why look at individual dives when the match result was unclear?
Counting dives and studying what follows a particular decision answer different questions. The first test did not support the claim that more dives mean more defeats. The later tests explored patterns within individual situations. Their strongest finding remained statistically clear after correction for the five high-speed outcomes, but exploring the same dataset can produce chance findings. Confirmation requires a new set of replays and tests planned in advance.
Does the danger suddenly change at 65% speed?
No. Cover loss increased across the speed bands. We chose 65% as a useful reference within these results, not as a dividing line between safe and unsafe. Both the pattern and the marker need checking on new replays.
Is “no early dive” the same thing as shadowing?
No. It can include shadowing, retreating, fake challenging, hesitating or another response. The analysis only tells us that no early dive was identified. We recommend shadowing in the relevant situations as coaching advice; the study did not specifically test a group of confirmed shadow defenders.
What should the last man do instead?
When your teammate is out of the play and the ball is coming back quickly, the coaching recommendation is to shadow, delay and protect the centre while they recover. The findings support being cautious about the forward commitment, but do not establish which alternative is best in every situation.
Be more cautious when you are the last defender and the ball is coming back quickly.
This study did not find clear evidence that teams which dive more often lose more matches. Looking at individual decisions revealed a more specific pattern that now needs checking on new replays.
When the first defender was beyond the ball and it was travelling quickly toward goal, a last-man dive was linked to a greater chance of losing the remaining cover. In similar situations, the adjusted estimate was 55.4% after a dive and 35.3% with no early dive, a difference of 20.2 percentage points. The separate 66.9% figure simply describes cover loss across all recorded high-speed dives. Neither figure tells us how many goals the decision caused.
Ball coming back quickly.
Protect the remaining cover.
Pacifist Labs tests coaching ideas against replay data. That means reporting unclear results as well as stronger findings, explaining the limits and updating the advice as the evidence improves.
Better coaching starts with an honest reading of the game.Put the ideas into practice in your own games.
The Digital Edition teaches the complete Pacifist System. Pacifist Score analyses positioning and decisions in your own Rocket League 2v2 replays, and the Dojo uses those findings to build a structured training programme.