Plat Teams Leave Nobody Back 60% More Often
What 22,462 ranked 2v2 games across three years reveal about the moments when neither player is between the ball and their own net.
In Study 005 we found Plats keep one back as often as SSLs at the moment they go for the ball. So we asked a wider question. Across the whole game, how often is nobody back at all? That’s where Plat loses ground, at every rank on the way up.
more often at Plat than at SSL,
both players past the ball
Ranked 2v2, all 22,462 games
rank and year groups
Nobody back falls at every rank, while both players behind the ball takes about half the game everywhere.
Same answer in sealed games, and in a year split frozen before it ran. In every year with enough games, no rank moved more than half a point.
Proof of cause. Some nobody back time is a clear you couldn’t stop, and a replay can’t tell that from a choice.
How much comes from double commits and how much from creeping up. And close games at Diamond, where it’s too close to call.
You keep one back when you go. What about the rest of the game?
Study 005 asked where your teammate is at the exact moment you go for the ball. When a Plat went, their teammate was in a covering position 29.8% of the time. SSL, 29.4%. Basically identical.
That answers one moment. A game is full of other moments. Clears, rebounds, 50s that go wrong, teammates pushing up. So Study 006 asks something wider. Across the whole game, how often is nobody back at all?
22,462 ranked games. Three states, no role labels.
Pacifist Labs analysed 22,462 ranked 2v2 games from 2024, 2025 and 2026, Platinum to Supersonic Legend. We also looked at 10,161 games with a professional player in the lobby, 3,171 of which are already among the 22,462. Every game had at least two minutes of live play.
We didn’t ask who was first man or second man. At every moment we asked where both players were compared to the ball, measured along the length of the pitch. There are only three possibilities.
| Rank | 2024 | 2025 | 2026 |
|---|---|---|---|
| Platinum | No data | 2,043 | 1,126 |
| Diamond | No data | 2,038 | 2,041 |
| Champion | No data | 2,018 | 2,020 |
| Grand Champion | 878 | 80 | 1,515 |
| Supersonic Legend | 2,035 | 4,055 | 2,613 |
Every rank up, nobody back happens less.
At Platinum, both players are past the ball 11.0% of the time. At SSL it’s 6.9%. Every rank sits lower than the one before it. Teams with a pro in them sit at 6.6%, though that one is descriptive only.
Nobody back, share of the game
Plat teams have nobody back about 60% more often than SSL teams. Put it into time. Over five minutes, a Plat team has nobody back for about 33 seconds. An SSL team, about 21. That’s 12 extra seconds every five minutes with nothing between the ball and your net.
Seconds with nobody back, per team, every five minutes
SSLs aren’t sitting back more. They just don’t both get caught.
So are Plats just too aggressive? Put Plat next to SSL across the three states. Both players behind the ball takes up about half the game at every rank, 52.2% at Plat and 53.8% at SSL. One behind and one at or past the ball is actually more common at SSL, 39.3% against 36.7%. The difference is the third state. Both at or past the ball, nobody back, 11.0% at Plat and 6.9% at SSL.
The three states, Plat against SSL
| Rank | Both behind | One behind, one at or past | Both at or past |
|---|---|---|---|
| Platinum | 52.2% | 36.7% | 11.0% |
| Diamond | 52.0% | 38.2% | 9.9% |
| Champion | 51.5% | 39.5% | 9.0% |
| Grand Champion | 51.9% | 40.0% | 8.1% |
| Supersonic Legend | 53.8% | 39.3% | 6.9% |
What “at or past” means. Every frame, the scorer asks whether a car is further up the pitch than the ball’s centre, by any distance. So it includes pressuring the ball from level with it, not just being out of the play. More than half of that time sits in the scorer’s shallowest depth band, on average about three car lengths past the ball. The player closest to the ball is at or past it 24.1% of the time they’re closest at SSL, and 29.6% at Plat.
One caveat. Inside a game, the team that had someone at or past the ball less often also won more often, 61.0% to 64.4% by rank. Nobody back is part of that time, so the win test can’t separate the two. The rank comparison can, and that’s where nobody back stands out.
Same game, same lobby. Nearly two in three at every rank.
SSLs are better at everything, so of course they look different. That doesn’t mean this one wins games. So we compared winners against losers inside the same game, which takes the lobby, the rank and the opponents out of it.
Games won by the team with nobody back less
And it scales. When the two teams were within a point of each other, it was close to a coin flip. As the gap grows, so does the edge.
Win rate for the team with nobody back less, by size of the gap
On its own that doesn’t prove cause. A bigger edge with a bigger gap is what you’d expect if it did, but it’s also what you’d see if being outplayed leaves you with nobody back. This can’t separate the two.
Where do you sit?
Upload one replay to Pro Match. The free mini report shows your behind ball and ball-relative depth next to the pros, which pro you play most like, and your biggest opportunity.
Find my Pro MatchDifferent question, different answer. No contradiction.
| Study | What it measured | Plat | SSL | Verdict |
|---|---|---|---|---|
| 005 | Teammate in a covering position at the moment you go for the ball | 29.8% | 29.4% | Same |
| 006 | Nobody back, across the whole game | 11.0% | 6.9% | 60% more at Plat |
The moment you commit isn’t where Plats lose ground. It’s the rest of the game. One way the two connect is Study 005’s main finding. Plat teams put two players on the same ball about two and a half times as often as SSL teams, and when both go and both get beaten, that’s nobody back.
Study 003 caught a close cousin of that. When your teammate is already past the ball and it’s coming at you fast, diving in as last man left nobody goal side 55% of the time, against 35% when the last man didn’t dive. That’s essentially the state this study counts, caused by one decision. Study 003 didn’t find that teams who dived more often lost more games, and it was exploratory, so read its figures as a first estimate. Read Study 003.
Study 001 fits too. It measured behind the ball across all of a player’s time and found the team behind the ball more won 58% to 65% of games, depending on rank. On that same measure, this data gives 73.4% for non-pro SSL players and 73.7% for pros, against the 73.5% and 73.7% Study 001 reported.
A fingerprint for your rank, under a frozen test.
Before splitting nobody back by year, we wrote down three tests and what would count as a pass, and froze them with a timestamp. All three passed.
| Rank | 2024 | 2025 | 2026 | Largest change |
|---|---|---|---|---|
| Platinum | No data | 11.1% | 10.9% | 0.21 |
| Diamond | No data | 9.9% | 9.9% | 0.01 |
| Champion | No data | 8.9% | 9.1% | 0.14 |
| Grand Champion | 8.2% | 7.0%, 80 games, not tested | 8.2% | 0.01 |
| Supersonic Legend | 7.1% | 6.9% | 6.8% | 0.35 |
| Test | Pass if | Result | Verdict |
|---|---|---|---|
| The ladder repeats inside each year | Lowest rank minus SSL, 95% interval above zero, every year | GC minus SSL 1.07 in 2024 · Plat minus SSL 4.20 in 2025 and 4.14 in 2026 | Passed |
| The win test repeats | Lower bound above 50% in every rank and year with 300 or more games | 11 of 11 · 61.1% to 65.5% | Passed |
| Rank matters more than year | No rank moves more than half the Plat to SSL gap | Largest change 0.35, limit 2.06 | Passed |
This one wasn’t the original plan. Here’s exactly what happened.
At the start of the analysis, one match in five was sealed away by a fixed hash of its match ID, 4,447 of these games. A different test was frozen for them first, on the scorer’s first man and second man (see Methods for why we don’t lean on those). Nobody back was spotted in the other four fifths.
After those frozen tests were reported, we checked nobody back in the sealed games. It came back the same. Because that check wasn’t frozen in advance, the year split was frozen before it was run.
| Check | Plat | Diamond | Champ | GC | SSL |
|---|---|---|---|---|---|
| Nobody back | 11.1% | 9.9% | 9.1% | 8.3% | 6.9% |
| Win test, all sealed games | 63.2% | 61.3% | 65.1% | 65.5% | 64.6% |
| Win test, no forfeits | 62.1% | 60.4% | 64.7% | 63.7% | 62.1% |
| Win test, one goal games | 58.2% | 53.4% | 60.1% | 60.1% | 60.2% |
A replay can’t read minds.
Some nobody back time is a clear you couldn’t stop. The data can’t tell that apart from a choice, and the team that’s being outplayed will spend more time with nobody back too. So this study doesn’t prove that keeping someone back causes wins.
It also can’t say how each moment happened. Two routes are common. A double commit, where both players go for the same ball and both get beaten. Or creeping up, where the second player pushes past the ball while their teammate attacks and the clear goes over both. These numbers don’t separate them.
One rule. Never both past the ball.
The rest of this section is my coaching from The Pacifist System, not data. The book already has a rule for both routes.
The 2nd man is not trying to force their way in.
The Pacifist System · Offence, 2nd man
Do not dive in as last man.
The Pacifist System · fundamental 17
- Be the anchor. When your teammate is on the ball, stay central, about two to three boost pads behind them.
- Keep the string tight. Imagine a piece of string between you. Too close and it goes slack. Too far back, it snaps. That’s string theory, from the book.
- Hard 50 coming? Drop back. When the opponents are about to have a hard 50 with your teammate, be further back, still central.
- Teammate beaten? Delay. Don’t dive in as last man. Drop back, delay the attack, and give your teammate time to get back involved.
Before you push up, check one thing.
| Question | What to look for |
|---|---|
| Is my teammate already past the ball? | If yes, you stay goal side. Never both past it. |
| Am I forcing my way in? | If you’re creeping up looking for a touch, wait for your turn instead. |
| Is the string tight? | Two to three pads behind and central. Not on top of them, not miles back. |
| Is a hard 50 coming? | Drop further back before it happens, still central. |
| Has my teammate been beaten? | Don’t dive in. Drop back, delay, and let them get back involved. |
The drills from the book. Drill 8A, Counterattack two man, where you hit it round the corner and back off for the counter. And Drill 18A, Trailer, behind and central, applying string theory. Ten minutes on each.
Large, stable and replicated across years. Still observational.
| Measure | Result | Read it with this |
|---|---|---|
| Nobody back | Plat 11.0% · Diamond 9.9% · Champ 9.0% · GC 8.1% · SSL 6.9% · pro teams 6.6% | Includes clears nobody could stop |
| Plat minus SSL | 4.12 points (4.03 to 4.22) · Plat 60% higher | Intervals bootstrapped over games |
| Three states, Plat against SSL | Both behind 52.2% · 53.8% · one behind, one at or past 36.7% · 39.3% | At or past means any distance, including level with the ball |
| Win test, all games | Plat 64.8% · Diamond 64.3% · Champ 63.7% · GC 61.8% · SSL 63.7% | Same game comparison, not proof of cause |
| Win test by size of the gap | 52.8% · 56.1% · 61.6% · 69.1% · 78.8% | Descriptive |
| Frozen year test | 3 of 3 passed · win test 11 of 11 · largest change 0.35 points | Sources differ a little by year |
| Sealed games | Plat 11.1% · SSL 6.9% · win test 61.3% to 65.5% | Checked afterwards, not frozen in advance |
| One goal games, sealed | Plat 58.2% · Diamond 53.4% · Champ 60.1% · GC 60.1% · SSL 60.2% | Too close to call at Diamond |
The strongest evidence is how consistent it is. Every rank in order, a bigger edge with a bigger gap, the same answer in sealed games, and a frozen test across three years where, in every year with enough games, no rank moved more than half a point. What it isn’t is proof of cause. And it isn’t peer reviewed.
The detail behind the numbers.
The sample
Ranked 2v2 replays from 2024, 2025 and 2026, scored by the Pacifist Labs scorer. Games needed at least two minutes of live play. Rank is the replay’s rank bracket, Platinum to Supersonic Legend. 2024 covers GC and SSL only. 2025 has only 80 GC games, so that one cell is reported but not tested. Teams with a pro are teams with a pro-flagged player in a ranked 2v2 lobby.
Nobody back and the three states
For each team the scorer reports the share of time with both players behind the ball, one behind and one at or past it, and both at or past it, measured along the length of the pitch. The three add up to 100%. At or past means the car’s centre is further up the pitch than the ball’s centre, by any distance. Nobody back is the share with both at or past the ball. The scorer counts every frame of live play, about 30 a second. Countdowns and goal replays are left out, and the play straight after a kickoff counts the same as any other. There’s no separate level with the ball category, so every frame is either behind or past. By year, the share with at least one player at or past the ball stayed between 45.9% and 48.6% for every rank and year with enough games. Gold appears only in 2026, outside the main sample, at 12.4% nobody back across 125 games.
Why we don’t use first man and second man here
The scorer also labels each player first man or second man, but it assigns those roles by distance to the ball, not by depth. Its second man is behind its first man only about 66% of the time, so it can’t answer whether the last man is back. An early draft of this study used those labels and drew the wrong conclusion from them. The depth states above need no labels.
The win test
For every decided game we compared the two teams in that game. The win test asks how often the team with less nobody back time won. Games where both teams had exactly the same figure were dropped. Intervals come from 2,000 bootstrap resamples of games.
The sealed holdout and the frozen plans
One match in five was set aside by a fixed hash of its match ID at the
start of the analysis. A first plan, on the scorer’s first and second
man, was frozen before those games were opened, and its tests passed.
Nobody back was then checked in the sealed games after the fact. A
second plan, the year split for nobody back, was frozen before any
nobody back result was split by year, and all three of its tests
passed. Both plans are timestamped. Their sha256 fingerprints, so the files can be checked, are 74ed3c5d82da286e943cc5600693b490af63f3f9a1b371f3141c1eec369f12f0 for the first plan and 80de3f0c04fb9759b094479a4badaabff871291eb6950e937b5a806ea03d10e1 for the year plan. The sealed game list is efffaba2b8adcbf6664ad148e69ed09ee269ae6ac3a130236126ccba2a808f99.
Sources and remaining limits
Figures were checked against the Study 006 result files, the frozen plans and the holdout list. These identify the internal evidence. They are not public download links.
The study is observational. Being outplayed can leave nobody back as well as choosing to push up, and the data can’t separate the two. No result here establishes a causal effect on wins or rank.
What the findings show and what’s still open.
Doesn’t this contradict Study 005?
No. Study 005 measured cover at the moment you go for the ball, and Plats matched SSLs. This measures the whole game. Both can be true. 005’s coaching said keeping one back at that moment isn’t the fix, and that still holds. This rule is about the rest of the game.
So should I stop pushing up?
No. SSLs have one player up at the ball about as often as Plats, so that isn’t what separates the ranks. Keep pressuring as first man. Just don’t both end up at or past the ball.
Is this just double commits again?
Double commits are one way it happens, and Study 003 measured a close cousin, diving in as last man after your teammate is already beaten. Creeping up and getting cleared over is another. These numbers can’t tell which it was each time, and the fix covers both.
Isn’t this just better players getting caught out less?
Partly, probably. Comparing teams in the same game takes out rank and lobby, but not which team was better on the day, and that team will both win more and get caught out less. That’s why we don’t claim cause. The ladder did hold in every year with enough games to test.
Is the coaching proven by this study?
No. It’s my coaching from the book, aimed at what the data found. It hasn’t been tested as an intervention.
Can I see my own number?
The free Pro Match mini report shows your behind ball and ball-relative depth next to the pros from one replay. The full replay analysis goes further.
Is your teammate already past the ball?
You keep one back when you go for the ball. Across the whole game, Plat teams end up with nobody back 60% more often than SSL teams, and the team that does it less wins nearly two games in three at every rank.
Never both past the ball.
Get your free Pro Match mini report.
Upload one replay and see your behind ball and ball-relative depth next to the pros, which pro you play most like, and your biggest opportunity. Or explore the Pacifist System’s Digital Edition for the full 2nd man chapter and the drills. Study 006 informs that coaching. It is not a validation of every part of the system.
PACIFIST LABS Shaping the Future of Rocket League Decision-Making.