Padel Levels & Ratings, Explained Honestly

Why padel has levels at all

Padel's whole social model depends on strangers. A court holds four. Most people have two reliable padel friends, not three, so the fourth spot gets filled by an app, a club WhatsApp group, or someone's cousin.

Fill that spot blind and you get the two matches nobody enjoys: the 6-0 where one pair spends 40 minutes apologizing, and the mismatch where a beginner stands in the back corner while three better players hit around them. Both waste a booking; one of them makes a new player quit.

So platforms need a number — not a perfect one, just one good enough to keep four people on court inside about half a level of each other. That is the entire design brief, and it explains almost every decision that later feels unfair. A booking platform's level is a matchmaking instrument. It was never built to settle who is best among friends, and it behaves strangely when you use it for that.

Levels do a second job too, quietly: they tell you where you stand in the sport. That is why "what level am I?" is the most-searched question in amateur padel, and why the answer deserves more care than a number pulled out of a signup form. If you want to work yours out from scratch, an honest self-assessment beats any quiz that flatters you, and the padel level test walks the same observable markers in twelve questions and hands you a band rather than a decimal.

The four kinds of number you will meet

Padel has no single governing rating. It has four different species of number that all get called "your level," and the differences matter more than the digits.

1. Platform levels, seeded by questionnaire and then driven by results. Playtomic is the one most players have. Its scale runs 0 to 7, seeded by a short questionnaire you fill in once when you join — how that starting level is decided, and why it is so often off — and from there it moves automatically on competitive match results. Playtomic's own level system documentation lists exactly four inputs to each change: the match result, the opponents' levels, your partner's level, and your reliability score. Its manager help center publishes the band names too — the 2.5 to 3.4 range is labelled "Intermediate," 3.5 to 4.4 "Intermediate High," 4.5 to 5.4 "Intermediate Advanced" (Playtomic Levels & Algorithm, page updated 16 July 2026). The full mechanism, including where it stops explaining itself, is in Playtomic level explained.

2. Self-assessed scales. MATCHi is the clearest example. Its levels are set by you: "Levels are self-assessed, and when you set yours you'll see a description of each step to help you choose the right one," says the MATCHi levels help page. Note the scale, because a lot of older conversion charts on the web have it wrong: MATCHi levels now run 1 to 10 for all sports and markets, with padel in Denmark the single exception at 1 to 7. A self-assessed number is honest about what it is — a claim, not a measurement — and it is only as good as the person typing it.

3. Official federation ratings, which are usually two things wearing one name. In the UK the LTA publishes a padel rating from 1.0 to 7.0 that works by picking the statement that describes you. At 3.0 you are "pretty consistent and can use tactics like playing off the wall"; 4.0 is "quite controlled on the court, can cover it well, and have a basic understanding of padel tactics"; 5.0 is "an experienced tournament player"; 5.5 is "a nationally-ranked padel player"; and the top, 7.0, is a professional "inside the top 150 in the men's professional rankings or top 50 in the women's" (What is my padel rating?). One row of that scale rewards a close read: 3.5 is not a better 3.0, it is a different route in — "I lack technique but do have some control on the court. Competent tennis and squash players who are new to padel come under this category." A scale built from statements can do that, and a scale built from arithmetic cannot. Separately, the LTA runs padel rankings, which are tournament points — "a ranking is made up of the best 6 results within the previous 52 weeks" (how padel rankings are calculated). Rating and ranking are different objects. People quote one and mean the other constantly, which is half of what the LTA padel rating guide exists to untangle.

4. Club letters and local shorthand. A to D, or "the Tuesday intermediate group," or a coach's opinion. No scale, no math, and frequently the most accurate assessment in the building: a coach has watched you play for a year, the algorithm has watched four scorelines.

Four species, four definitions of the word "level." So "which system is most accurate?" is a category question: accurate at what? Each of them is good at its own job, and none of them was built to rank your friends.

How an algorithmic level actually moves

Here is the part worth understanding properly, because once you see it, most "the algorithm is broken" moments stop being mysterious.

Any results-driven rating runs the same three-step loop.

Step one: predict. Before the match, the system converts the four ratings into an expected result. A pair rated 3.4 and 3.0 against a pair rated 3.2 and 3.1 is close to a coin flip. A 4.0 pair against a 2.5 pair should win comfortably, and the system knows roughly how comfortably.

Step two: compare. After the match it compares what happened with what it predicted. This is the whole mechanism, and it is why the question is never "did you win?" SquashLevels, one of the most transparent rating operators in any racket sport, puts it plainly: "If you're better than your opponent then you're expected to win so the big question is not 'did you win?' but how did your result compare with what was expected?" (I won my match but my level went down).

It also answers the question people ask instead: is my win rate the better number? Usually not. A 70% win rate against opponents you outclass is worth less than 45% against the best pair in the club, and pricing that difference is the entire job of a rating. Win rate becomes the honest number only when the opposition is fixed — which is exactly what a standing crew is.

Step three: move, but only as far as the system trusts itself. The size of the change scales with how confident the system is in your current number. New account, three matches in, wide uncertainty — the first-ten-matches rollercoaster, and it is by design. Two hundred matches in — small, boring, correct adjustments. Playtomic calls this reliability and states it is "the most important factor affecting your level changes" (The Playtomic Levels: ups & downs). DUPR calls it a Reliability Score and considers 60% the threshold for a dependable rating. Chess's Glicko systems call it rating deviation and do the same arithmetic. Different names, same idea, explained properly in reliability, and why it matters more than your level.

Doubles adds one complication that padel players feel constantly. A rating engine sees one result for two people, so it needs a rule for splitting credit. The common shortcut is to treat each pair as its average: Playtomic states its algorithm "calculates the average level per team, compares the expected outcome with the actual result, and adjusts each player's level accordingly." That shortcut is defensible and it is also the direct cause of the single most common grievance in padel — your partner played badly, the pair lost, and the number blamed you for it. Why your level dropped walks through the four honest reasons that happens, one of which is not "the algorithm is rigged." To test a specific night rather than read about the mechanism, the level change simulator takes your level and reliability, the four numbers and the score, and returns a likely direction and range with its reasoning shown.

If you want to watch the loop run rather than read about it, how Elo ratings work in racket sports does the arithmetic from zero, and the padel Elo calculator will do a real one for you: put in four ratings and a score, and it shows the expected win percentage, the point change, and a sentence explaining why the number moved as far as it did.

Why your numbers disagree

You are a 3.2 on one app, a 5 on another, and your coach says LTA 3.5. All three can be right at the same time, for four reasons.

Different scales with different tops. Playtomic's 7 is a professional, and so is the LTA's 7.0 — but the LTA reaches it through world-ranking statements while MATCHi's 10 is simply the top rung of a ladder you climb by typing. Two scales can share a maximum and still stretch the middle completely differently, which is why "I'm a 4 on both" means almost nothing.

Different populations. A level is implicitly relative to the people you are compared against. The same player will read differently in Madrid, where the club average is high and deep, than in a city where padel arrived in 2024 — why a 3.5 travels badly is a market-maturity story, not a stereotype. It is also why nobody can honestly tell you what percentile a "real" 4.0 sits in: no operator publishes the distribution, so every figure you see is reasoning from band descriptions. The pyramid logic holds — each half-band above the middle thins fast — but "am I good?" is answered by the people you play, not by a percentile.

Different definitions of "level." A self-assessed scale measures your opinion. A questionnaire-seeded algorithm measures your results, starting from your opinion. A federation rating measures which sentence you picked. Converting between them is translating between languages that do not have the same words.

No official conversion exists. Not between Playtomic and the LTA, not between Playtomic and MATCHi, not anywhere. Every conversion chart on the internet — including any you find here — is somebody's approximation. Directional maps are still useful for picking the right session, and the padel level conversion guide gives you the most careful one available along with the caveats each row needs. Treat it as a translation aid, never as a transfer.

The practical version of all this: your level is not a property of you, it is a property of a system's opinion of you. When two systems disagree, neither is lying. And when you want to know what a specific number actually looks like on court — what breaks down at 3.0 that holds at 3.5 — read the behaviors rather than the digits. Padel level 2.5 vs 3.0 vs 3.5 is written entirely in observable behavior for that reason, and what a 0.5 gap does to a 2v2 does the arithmetic on why half a unit decides matches.

What the numbers do to people

Every rating system with public stakes produces the same four behaviors. They are not a padel problem or a Playtomic problem; they show up in chess, in squash, in pickleball, in every ranked video game ever shipped.

Protecting the number. You start choosing matches by risk instead of by fun. The one substantial neutral critique of Playtomic's system on the English web makes exactly this observation — that inaccurate matchmaking means "many players end up picking and choosing their open matches more and more carefully" (Asa Jay Ackley, Is Playtomic's rating system flawed?, September 2025). A rating that makes you play less has failed at its actual purpose.

Sandbagging. Under-representing your level to win easier matches, or keeping a number low because a tournament category caps at 3.5. It only appears where the number gates something valuable, which is the useful clue about how to prevent it — see sandbagging in padel. Platforms make their own decision about how far to indulge it: Playtomic lets you lower your own level by hand and then closes the door behind you — "once you lower your level manually in the app, you can't change it back" (How to change your level). One-way levers are an unusually honest piece of design. They price the escape hatch instead of pretending it does not exist.

Partner-farming. Pairing with a much weaker player because the expectation is low and the upside is high. Structural, not moral: if the math rewards it, someone will do it.

Ducking your friends. The worst one, and the quietest. When a friendly Tuesday game can cost you rating points, some people just stop turning up for it.

Notice what all four have in common. They are responses to stakes, not to arithmetic. A number nobody outside your group can see does not create any of them, because there is nothing to protect and nobody to impress. That is not a marketing claim, it is the mechanism — and pushed all the way, it becomes the case against one global number.

And there is a deeper objection that survives even a perfectly-built system: a single number cannot know that you were carrying an injury, that the sun was behind the glass at 7pm, or that you and Mara have played eleven matches together and never once talked about who covers the middle. Why ratings feel unfair takes that seriously rather than explaining it away, and lands on the framing that fixes most of the frustration: a rating is a weather forecast, not a verdict.

What honest rating math looks like

Rating systems are not all equally good, and the differences are specific enough to name. Four properties separate the trustworthy ones.

It uses the margin, not just the winner. A 7-5 and a 6-0 are different pieces of evidence and a system that treats them identically is discarding most of what it saw. SquashLevels uses "a combination of points scores and games scores to assess the result" and states the goal directly: "if you are twice as good as your opponent then your level will be double theirs" (What are Levels?). Pickleball's DUPR compares the actual score to an expected score, with the consequence that "DUPR ratings can now go up with a loss if you outperform expectations, or down with a win if you underperform" (Cracking the DUPR code). Margin awareness is also the best available antidote to the partner-drag complaint: if a narrow loss with a struggling partner can still raise you, the grievance largely evaporates. Padel has more margin signal than almost any sport — games, sets, and every deuce that decided a game — and most padel systems throw it away.

It shows its uncertainty instead of hiding it. Every serious system computes a confidence figure. The honest ones display it. Glicko's original documentation recommends reporting a player's strength as an interval rather than a single number: the 95% interval runs from the rating minus twice the rating deviation to the rating plus twice it, so a 1850 with an RD of 50 is really "somewhere between 1750 and 1950" (Mark E. Glickman, The Glicko system). DUPR publishes a Reliability Score as a percentage and names 60% as the point where a rating is dependable. Showing uncertainty costs nothing, and it prevents the worst rating experience there is: a confident-looking number handed over after three matches.

It converges in a stated number of results, and says so. The honest answer to "when is my rating real?" is a count, not a vibe. Glickman's own guidance for Glicko-2 is that the system "works best when the number of games in a rating period is moderate to large, say an average of at least 10-15 games per player" (Example of the Glicko-2 system, 2022). That is the right order of magnitude for padel too, and how many matches until your rating is accurate works through why margin data gets you there faster.

It cannot be farmed, and it explains every move. SquashLevels handles the runaround problem by expecting the better player to ease off: within a modelled band of behavior there is "no change for either player," and beyond a 3:1 level gap heavy extra damping applies (how to give a lesser opponent a runaround). And every change should decompose into one sentence a human can check: "we expected you to win 40% of the games, you won 55%." A number that moves without explanation is not a game, it is a grievance.

Pickleball is further along on most of this than padel is, and it also demonstrates the failure mode of getting it right: a rating so trusted that people start policing each other with it. What padel can learn from DUPR covers both halves.

Where a crew-local number fits

Now the honest part about us, and it is a boundary, not a pitch. The general version of the argument — should a friend group track its own rating at all? — is worth reading before any product is involved.

Rivals is a scoreboard for a fixed group of friends who play 2v2 most weeks. Its rating is deliberately crew-local: it exists inside your group and nowhere else. There is no global table, no cross-crew ladder, no number a stranger can look up. It is margin-aware, so a 7-5 and a 6-0 move it differently and a narrow loss against a strong pair can still move you up. And it explains every change in a sentence you can tap — expected share of games, actual share, resulting move.

That design buys three things a global rating structurally cannot. Nothing to protect, so nobody ducks a Tuesday. Rotating partners, which is exactly the data that makes individual ratings from doubles results mathematically defensible — the same four people in different pairings decorrelate quickly. And no policing, because the only people who can see the number already know how you play.

It also lets the smaller, better numbers matter. The record a crew actually quotes is never the rating — it is the head-to-head. "I'm 7-3 against Marco" is a complete sentence that needs no arithmetic explained, and it arrives with a villain, a streak and a revenge fixture. A global level cannot produce one, because it never plays the same person twice.

What it does not do is the thing a global number is for. It will not match you with strangers at a club, enter you in a category, or tell you where you stand in your city. For that you need Playtomic's number or your federation's, and this whole guide exists to help you read them properly. But "who is really best out of the six of us" was never a question a global rating could answer — and at crew scale it is the smaller, easier, more accurate problem.

Free covers one crew of up to four players with nothing inside it metered. Use it or use a notebook — the principle holds either way: results your group cares about deserve a longer memory than the four people in the car home.

Start anywhere in the chapters above. If you only read one, read why your level dropped — it is the question that brought most people here, and the answer is more interesting than "the algorithm hates you."

FAQ

What is a good padel level for an amateur player?

Most regular club players sit in the middle of whatever scale they are on — roughly 2.5 to 3.5 on a 0 to 7 platform scale, which Playtomic's own published bands label "Intermediate" and "Intermediate High." Higher than that usually means competitive play. There is no single answer because scales differ; compare yourself to the people you play with rather than to a number you read online.

Which padel rating system is the most accurate?

They are accurate at different things. A results-driven platform level like Playtomic's is best at predicting whether tonight's match will be close, because that is what it was built for. A self-assessed level like MATCHi's is best at getting you into the right club session. A coach's opinion is often the most accurate description of your actual game. None of them is designed to rank a fixed group of friends against each other.

Is there an official conversion between Playtomic, LTA and MATCHi levels?

No. No governing body or platform publishes an official mapping, and the scales are built differently — Playtomic 0 to 7 driven by results, LTA 1.0 to 7.0 by self-selected statements, MATCHi 1 to 10 self-assessed. Any conversion table you find, including ours, is an approximation useful for picking a session and useless as a transfer.

Why does my padel level change when my play hasn't?

Because the number is an estimate that is still being refined, and because it responds to the gap between expected and actual results rather than to your absolute standard. Early on, when the system's confidence in you is low, ordinary results move it a long way. Once you have a substantial match history the same results barely move it at all.

Can a padel rating tell who the best player in my group is?

Only if it has seen enough of your group's matches, with partners rotating. A global platform level cannot, because it is diluted by matches against strangers and, for most groups, does not count friendly games at all. A rating computed from your group's own results, with everyone partnering everyone, can — and needs roughly 10 to 15 results per player before the ordering is stable.