Why Plats Stay Plat
What 836 ranked 2v2 games and 118,965 approaches to the ball reveal about two teammates ending up on the same ball.
Nearly every Plat I coach tells me the same thing. If my mechanics were better, I’d be Diamond. So I tested something I reckon matters more. It turns out SSLs go for the ball more than Plats do, and still end up on top of their teammate less than half as often.
how often Plat teams overlap
compared with SSL teams
Approach overlap, not validated double commits
Test written down and frozen first
SSL teams go for the ball more and still overlap far less.
The gap came back almost identical in games locked away before the study.
Messier games, keeping one back and bailing out faster were all tested. None of them explained it.
Whether these overlaps cost goals. That result was inconclusive, so we’re not claiming it.
Is it really your mechanics?
I’ve coached for over 1,500 hours. Ask a Plat why they’re stuck and you’ll nearly always get the same answer. Mechanics.
I’ve never fully bought it. When I watch Plat replays, the goals that hurt most usually don’t come from a missed flick. They come from two teammates arriving at the same ball, both committing, and leaving the net wide open behind them.
So Study 005 asks a simple question. When you go for the ball, how often is your teammate already going for it too? And does that change as you climb?
If it’s just ball chasing, the fix is easy. Go less. If it’s something else, we need to find out what, because “go less” isn’t what the best players are doing.
836 games. Every player tracked ten times a second.
Pacifist Labs analysed 836 ranked 2v2 games from February to April 2026. Five ranks, tier I of each. Platinum I, Diamond I, Champion I, Grand Champion I and Supersonic Legend, with known pros taken out of the SSL group.
A replay can’t tell you what a player was thinking. It can tell you what they did. So every time a car went at the ball for real, we logged it as an approach. Close, quick, pointed at the ball and on course to get there within about a second. Anything shorter than a fifth of a second got binned, so a twitch on the stick doesn’t count.
That gave us 118,965 approaches. Then we asked one question of every single one. Was your teammate going for the same ball at the same time?
We report most rates per five team-minutes. That’s one team in live play for five minutes, which keeps fast and slow games comparable.
It drops at every single rank.
At Platinum, 7.33% of approaches happen while your teammate is going for the same ball. That’s about one in fourteen. At SSL it’s 3.00%, or about one in thirty three.
Approaches where a teammate was going for the same ball
Every rank up, the number drops. Clear as day. Plats, we need to talk.
Now, you’d assume that’s because Plats ball chase. I thought the same. They don’t. SSL teams made 82.0 approaches every five team-minutes. Plat teams made 78.5.
Approaches per five team-minutes
I had three explanations. None of them survived.
Something is happening at SSL that just isn’t happening in Plat. I went in with three ideas about what it could be, and tested each one.
Theory one. Plat games are just messier. More bounces, more scrambles, cars upside down in the corner. If that’s true, you’d clash with the other team more often too. So we checked how often an opponent was going for the same ball as you. Plat, 10.1%. SSL, 9.8%. Same chaos. The difference only shows up between you and your own teammate.
Theory two. SSLs just keep one back more. It’s the oldest advice in Rocket League and I’ve leaned on it myself. But when a Plat goes for the ball, their teammate is in a covering position 29.8% of the time. SSL, 29.4%. Basically identical. Plats already keep one back as often as SSLs, so that’s not what separates them. It doesn’t mean keeping one back is wrong.
Theory three. SSLs make the mistake and peel off faster. Once a teammate joins an approach, Plats end theirs within three tenths of a second 47.7% of the time. SSLs, 46.6%. They don’t bail out any quicker.
| Theory | What we measured | Plat | SSL | Verdict |
|---|---|---|---|---|
| Plat games are messier | Opponent going for the same ball, matched game state | 10.1% | 9.8% | Ruled out |
| SSLs keep one back more | Teammate in a covering position when you go | 29.8% | 29.4% | Ruled out |
| SSLs peel off quicker | Approach ended within 0.3 seconds of a teammate joining | 47.7% | 46.6% | Ruled out |
Opponents are not a perfect control group. Cover is a geometric position, not the ability to make a save. The bail-out figures have wide intervals and no clear trend across ranks.
The rank told to rotate back post the most does it the least.
This one’s going to annoy a few people. We ran a smaller study inside this one, first man only. That’s 27,000 recoveries where we could see who was first man. Then we asked how often first man loops all the way round back post.
First man recoveries that loop round back post
It’s rare at every rank. But Plat does it less than any other rank we tested, and SSLs do it more. The rank that hears “rotate back post” more than anyone actually does it the least.
Tighten it right up and the gap gets bigger.
An overlapping approach only tells you so much. So we tightened the definition. Both teammates fully committed, and both on course to arrive within about a third of a second of each other. The kind you wince at in a replay.
Strict double commits per five team-minutes
Plat teams did that 3.23 times every five team-minutes. SSL teams, 1.23. That’s about 2.6 times as often, and the same direction as the looser overlap measure.
A computer calling it a double commit doesn’t make it one.
Sometimes two on the ball is right. A passing play, a follow up. So I sat down and watched 300 plays myself. The flagged ones were shuffled in with near misses and normal approaches.
I watched them blind. No rank, and no idea what the computer had decided. Just cars and a ball.
| Result | Figure |
|---|---|
| Clear flags judged genuine, unnecessary double commits | 79 of 146 · 54% |
| The rest | Justified two-player plays, or not double commits at all |
About half of what the detector flagged were real mistakes. The rest weren’t. And with a coach’s eyes on it, the gap between Plat and SSL was still there.
This is where most findings die. This one didn’t.
A sense check wasn’t enough for me. I wanted to know if this would actually hold up.
Before any of this started, 232 games were locked away. Nobody looked at them. When the main study was done, we wrote down exactly what we’d test and how we’d measure it, and timestamped it so we couldn’t move the goalposts. Only then did we open them.
210 of those games passed our quality checks. This is the point where a pattern from the first batch usually turns out to be noise.
Strict double commits per five team-minutes · sealed games
| Games | Platinum I | SSL |
|---|---|---|
| First batch | 3.23 | 1.23 |
| Sealed games | 3.30 | 1.26 |
Near enough identical, in games we’d never seen. The Plat to SSL gap in the sealed games was 2.04, with an interval running from 1.48 to 2.62. And all five ranks dropped in order again, Plat down to SSL.
That’s replication. And that’s when I stopped calling it a hunch.
We couldn’t show these overlaps cost goals.
The obvious next question is whether all this ends up in your net. We looked. After an early overlap, the goal rate was 5.3%. Without one, it was 5.7%. The interval runs in both directions, so the honest answer is that we don’t know yet.
That’s why you won’t hear me say double commits lose you games because of this study. I believe they do, from years of watching replays. This data doesn’t prove it either way.
It’s ownership. And that’s a decision you can change.
To every Plat who’s told me it’s their mechanics. Two players see the same chance at the same time, and both of them go, yep, that’s mine. At SSL, one says yes and the other’s already said no, before either of them has moved.
That’s a decision. Which means you can change it next game.
The rest of this section is my coaching, not data. It’s aimed straight at what the data found. I call it string theory.
This isn’t keep one back again. Plats already sit in cover as often as SSLs, so that’s not the fix. It’s about the moment your teammate goes for the ball. You don’t go with them. You move with them on the track instead.
Imagine a piece of string between you and your teammate, and you keep it tight. You’re on a train track down the middle of the pitch. When they go to the ball, you move with them, up and down the track. But you never leave the middle.
To find your spot, take your teammate’s position, slide it across to the centre, and sit two to three boost pads behind it. From there both posts are covered, and you can step onto anything that comes through.
Winning the challenge is their job. Yours, as last man, is making sure nothing happens if they lose it.
Before you commit, ask yourself one thing.
Is my teammate already going? If yes, stay on your track and keep the string tight. Let them work.
| Question | What to look for |
|---|---|
| Is my teammate already going? | If they’re closer, quicker or already committed, the ball is theirs. Your job changes. |
| Where’s my spot on the track? | Their position slid across to the centre, then two to three boost pads behind it. |
| Is the string tight? | When they move up the pitch, you move up the track with them. When they come back, so do you. |
| What happens if they lose it? | You should already be in line to step onto the next touch. If you’re not, you’re too far up or too wide. |
| Am I the second player on the ball? | If you’re both arriving within a moment of each other, one of you needs to have said no already. |
Watch one of your own replays with just that question in mind. You’ll probably catch yourself saying yes at the same time as your teammate more often than you’d like. I know I did.
A clear, replicated gap. An observational study.
118,965 approaches don’t mean 118,965 independent experiments. The same players appear again and again, and one passage of play can produce several approaches. That’s why the sealed games matter more than any single number from the first batch.
| Measure | Result | Read it with this |
|---|---|---|
| Teammate overlap by rank | Plat 7.33% · Diamond 5.75% · Champ 4.63% · GC 3.53% · SSL 3.00% | Approach overlap, not validated double commits |
| Approaches per five team-minutes | Plat 78.5 · SSL 82.0 | Detected approaches, not intent |
| Opponent overlap | Plat 10.1% · SSL 9.8% | Opponents are not a perfect control group |
| Teammate in cover | Plat 29.8% · SSL 29.4% | Geometric position, not ability to save |
| Bail out within 0.3 seconds | Plat 47.7% · SSL 46.6% | Wide intervals, no clear rank trend |
| Strict double commits, first batch | Plat 3.23 · SSL 1.23 | Detector segments, not distinct plays |
| Blind review | 79 of 146 clear flags genuine · 54% | One reviewer, the author |
| Strict double commits, sealed games | Plat 3.30 · Diamond 2.31 · Champ 2.14 · GC 1.67 · SSL 1.26 | 210 of 232 usable · Many players also in the first batch |
| Sealed Plat to SSL gap | 2.04 · interval 1.48 to 2.62 | Decline across ranks was not pre-tested |
| First man back post | Plat 4.1% · SSL 5.7% | Captured first-man returns only |
| Goals after an early overlap | 5.3% with · 5.7% without | Inconclusive · interval spans both directions |
The strongest evidence here is the replication. A gap that shows up in the first batch, then comes back at almost the same size in sealed games tested under a plan frozen in advance, is much harder to explain away as noise.
What it isn’t is proof of cause. And it isn’t peer reviewed.
The detail behind the numbers.
The sample
836 ranked 2v2 games from February to April 2026. Replays were grouped by source rank cohort, using tier I of each rank. That means Platinum I, Diamond I, Champion I, Grand Champion I and Supersonic Legend. Known professional players were excluded from the SSL group. A source cohort is not a verified rank for every player in every game.
Approaches and overlaps
Every player was tracked ten times a second. An approach is a spell where a car is close to the ball, moving quickly, pointed at it and on course to reach it within about a second. Spells shorter than a fifth of a second were discarded. This is a motion detector. It records what a car did, not what the player intended.
An overlap is an approach during which a teammate was also on a qualifying approach to the same ball. Opponent overlap uses the same rule with an opponent in place of a teammate, compared within matched game states.
Strict double commits
A strict double commit requires both teammates to be fully committed and on course to arrive within about a third of a second of each other. Rates are reported per five team-minutes, meaning one team in live play for five minutes. Counts are detector segments, so one messy passage of play can contribute more than one.
Cover and bail-out measures
A teammate counts as covering when they are in a geometric covering position as their teammate goes for the ball. It does not assess whether they could actually make the save. The bail-out measure asks whether a player ended their approach within 0.3 seconds of a teammate joining it. Its intervals are wide.
The first-man back post study
A smaller analysis of about 27,000 recoveries where the first man could be identified. It records how often that recovery looped round the back post. Only captured first-man returns are included.
The blind review
300 plays were reviewed. Detector flags were shuffled in with near misses and normal approaches. The reviewer saw no rank and no detector verdict. Of 146 clear flags, 79 were judged genuine, unnecessary double commits. There was one reviewer, the author, so this is a sense check rather than an independent accuracy study.
The sealed holdout
232 games were set aside before the study began and were not inspected. After the main analysis, the test and measurement were written down and frozen with a timestamp. Only then were the sealed games opened. 210 passed quality checks. The pre-specified test was the Platinum to SSL gap in strict double commits, which came back at 2.04 with an interval of 1.48 to 2.62. The decline across the ranks in between was not part of the frozen test. Many players in the sealed games also appear in the first batch.
Sources and remaining limits
The figures were checked against the Study 005 research pack. That covers the routes and goals report, the roles follow-up report, the corrected rank metrics, the expert review results and the holdout results. These identify the internal evidence. They are not public download links.
The study is observational. Players chose their own approaches and the sample is grouped by source cohort. Measurement error and detector selection can influence the comparisons. No result here establishes a causal effect on goals, wins or rank. The link between early overlaps and goals was inconclusive.
What the findings show and what’s still open.
So is it not mechanics at all?
Mechanics obviously matter. This study doesn’t measure them. What it shows is that something other than mechanics changes a lot between Plat and SSL, and it’s something you can work on without hours in free play.
Should I just go for the ball less?
No. SSLs actually go for the ball more than Plats. The difference is how rarely they end up on the same ball as their teammate. Go when it’s yours. Stay on your track when it isn’t.
Doesn’t theory two say keeping one back is pointless?
No. It says Plats already keep one back as often as SSLs, so it can’t explain the gap. It says nothing about whether keeping one back is good. String theory isn’t about sitting back more. It’s about what the last man does when their teammate goes. Plats are about two and a half times as likely to go as well.
Does this prove double commits lose games?
No. We looked at goals after early overlaps and the result was inconclusive. I believe they cost you, but this data doesn’t prove it either way.
How do you know SSLs decide earlier?
I don’t, directly. Messier games, keeping one back more and bailing out faster were all ruled out, which leaves an earlier decision as the explanation still standing. That’s my read of the data, not something the study measured.
Is string theory proven by this study?
No. String theory is my coaching. It’s aimed at what the data found, but it hasn’t been tested as an intervention.
Does this mean back post rotation is wrong?
It means back post rotation isn’t what separates Plat from SSL. First men at every rank rarely loop back post, and Plats do it least. Study 004 covers ball side rotation in much more depth.
Are you the second player on the ball?
Plat teams put two players on the same ball about two and a half times as often as SSL teams. SSLs go for the ball more, so it isn’t about how often you go. It held up in games we’d never seen.
On my read, it comes down to a decision. One you can fix.
Is my teammate already going? If yes, let them work.
Find out where your decisions sit.
Upload a replay to the free Which RL Pro Do You Play Like test, or explore the Pacifist System’s Digital Edition for the full coaching framework. Study 005 informs that discussion. It is not a validation of every part of the system.
PACIFIST LABS Shaping the Future of Rocket League Decision-Making.