A tweet on X, a five-game sample, and the maths behind a start that’s running well ahead of the performances underneath it — time the rebuild shouldn’t waste.

Twelve of sixteen. That’s how many “big chances” Manchester City’s opponents have missed against them this season — the club sitting top of the Premier League, conceding chances at a rate that would alarm most defensive coordinators, saved from the numbers by finishing so wasteful it borders on charitable. It’s a useful place to start, because the exact same mechanism is sitting underneath the numbers at Newcastle right now — just working in the other direction. That’s not a reason to panic. It’s a reason to be honest about what the next few months are actually buying the club, and what needs to happen while the buying lasts.

The claim

On Monday night, Ross (@deeprossed) replied to a thread from the analytics account Statz (@Statz_Ai) that had ranked Premier League sides by big chances conceded across their first five matches. Newcastle didn’t crack the top ten. Ross’s conclusion: the volume of shots Newcastle are conceding doesn’t matter, because the shots aren’t good ones, and a defence facing low-quality efforts is a defence that’s fine.

It’s a reasonable-sounding instinct — the sort of thing that gets nodded along to by ten thousand people before anyone runs a number on it. It’s also wrong, in a way that’s provable rather than debatable, and it’s worth taking apart properly rather than dismissing with a shrug. Ross clearly cares about the numbers enough to go looking for them, which is more than can be said for most of Twitter. He’s just reached for the wrong one.

Here’s the leaderboard he was working from:

Two things jump out immediately, and Statz himself flagged the second one in the very next line of his own thread: Manchester City — top of the league, unbeaten, five straight wins — sit fifth on this list, and twelve of the sixteen big chances they’ve conceded have been missed. That’s a 25 per cent conversion rate against opponents in the highest-quality bracket of shot, roughly half what you’d expect a competent Premier League forward line to manage from situations Opta’s own analysts have judged good enough to be expected to score from. City aren’t controlling that outcome. They’re benefiting from it.

Which means the stat Ross is leaning on for reassurance is the same stat currently explaining why the best team in the country isn’t getting punished the way its underlying defending suggests it should be. If it can flatter Manchester City by twelve wayward finishes, it can just as easily flatter anyone sitting further down that list — including a side that isn’t on it at all.

Where “big chance” breaks down

Start with what the category actually measures. A big chance, in Opta’s framework, is a shot from a position where a player would reasonably be expected to score — broadly, the far right tail of shot quality, roughly the efforts carrying something north of a 30–40 per cent chance of going in. It’s a deliberately narrow bucket, built to flag the sitters and the one-on-ones. It was never built to describe the totality of a team’s defensive exposure, and using it that way is the first error in Ross’s reasoning.

Expected goals exists precisely because a binary “big chance or not” filter throws away almost everything that actually happens in a football match. Every shot faced carries some probability of becoming a goal — a cut-back turned first-time from the edge of the box, a corner that falls kindly for a half-hit volley, a defensive header that only half-clears — and none of those routinely qualify as big chances, while each of them carries a real, non-trivial chance of finding the net. A defence can concede zero big chances by Opta’s definition and still be bleeding an expected-goals total north of two a match, purely by accumulating enough of that middle tier. Counting only the tail and drawing conclusions about the whole distribution is the mistake sitting underneath the tweet.

The maths nobody tweets

Treat each shot faced as an independent trial with some probability of success, p, and the probability of conceding at least once across n shots follows: P(at least one goal) = 1 − (1 − p)ⁿ. Shots that fall outside the big-chance bracket still convert at something in the region of one in twelve to one in fourteen — nowhere near zero. Run fifteen of those through the formula at p = 0.07 and the probability of conceding climbs past 65 per cent. Push it to twenty shots and it clears 75 per cent. None of those shots needs to be a big chance for that outcome to arrive. The danger was never sitting in any single effort — it’s compounding across the volume, the same way a casino’s edge does across enough hands, regardless of how unthreatening any individual one looks.

Goals themselves don’t arrive smoothly, either — they arrive as counts, which is where a Poisson model earns its keep. The tool takes its name from Siméon Denis Poisson, the French mathematician who introduced it in an 1837 study of criminal and civil jury verdicts, working out how to predict the number of rare, independent events that would occur within a fixed stretch — given nothing more than their average rate — from a starting point in probability theory rather than sport. It’s since become the standard way to model exactly that kind of count in football: not how likely any single shot is to score, but how many goals a team is likely to concede across ninety minutes given its overall rate of chances allowed. Treat goals conceded in a match as Poisson-distributed with mean λ equal to cumulative expected goals faced, and the relationship between shot volume and disaster is convex, not linear: nudging λ from 1.0 to 2.0 doesn’t just double the expected goals conceded, it disproportionately inflates the probability sitting in the two-and-three-goal outcomes — the collapses, not the routine defeats.

And there’s a reason a five-game count of big chances conceded is such an unreliable guide in the first place. Rare-event counts behave like Poisson processes, and their relative variance — the coefficient of variation — scales with 1 ÷ √λ. A stat averaging one or two events a game wobbles enormously over five matches; a stat averaging twelve to fifteen events a game, like total shots faced, doesn’t. Put a number on it: a defence with a genuine long-run rate of 1.5 big chances conceded per game would, over five matches, have a Poisson mean of 7.5 and a standard deviation of roughly 2.7 — meaning a perfectly average defence could produce anywhere from two to thirteen in this exact sample purely by chance, before a single tactical or personnel factor gets involved. The leaderboard Ross quoted, running from 20 down to single figures inside five fixtures, is a metric demonstrating its own unreliability in real time.

This is also the terrain of PDO — a name that looks like an acronym but isn’t one; it comes from the online handle of a commenter on an old Edmonton Oilers hockey blog who first noticed the pattern it now describes. The stat itself is simple: a team’s save percentage plus its shooting percentage, added together and used as a rough proxy for how much of a team’s results are being carried by finishing variance rather than repeatable process. PDO across a division tends to sit close to a league-average baseline over a full season and drift back toward it wherever it’s strayed. City sitting on a big-chances-conceded record that would normally trouble a mid-table side, while their opponents miss three-quarters of their clearest sights of goal, is a PDO story. Newcastle’s own numbers, below, show exactly where that same dynamic currently sits for them.

Newcastle, by the numbers

Here’s where the theory stops being abstract. Through the first four gameweeks of 2026-27 — the last point at which Opta’s underlying data has been fully reconciled into the season totals — Newcastle have created 3.54 expected goals. That’s the second-lowest tally of created chances in the entire Premier League, ahead of only Aston Villa. At the other end, they’ve conceded 7.38 — the second-highest in the division, behind only Crystal Palace. Put those together and the expected-goals difference is −3.84. Not bottom-half. Not relegation-form. The single worst expected-goals difference of any club in the Premier League, marginally worse even than bottom-placed Aston Villa’s −3.73.

None of that should land as a surprise dressed up as one. This is a squad four months into a genuine rebuild — Bruno Guimarães, Sandro Tonali and Anthony Gordon all sold across the last two windows, a reshaped spine still learning each other, and Matthias Jaissle only a handful of matches into installing a way of playing that was never going to arrive fully formed by September. A dip in the underlying numbers during a squad and identity change is normal. It’s what the numbers below actually mean that matters.

Run those same four gameweeks through Squawka’s rounded xG model — converting each match’s underlying numbers into the scoreline they actually deserved rather than the one that went on the board — and Newcastle sit 19th, with a record of nought wins, two draws and two defeats. Seven places worse than where they actually stood in the real table at that exact moment. There is no bigger gap between deserved position and actual position anywhere in the division.

The finishing explains the whole of it. Newcastle have scored seven goals from that 3.54 xG — an overperformance of +3.46, the third-highest in the Premier League, behind only Brighton and Chelsea. It’s carrying the entire points tally. And it’s worth being precise about where the gap actually sits, because the instinctive read — a defence riding its luck — isn’t quite what the numbers show. Newcastle’s expected goals against through four games works out to 1.85 a match. Their actual goals conceded, across five completed fixtures as of 20 September, sit at nine — 1.8 a match. The defence isn’t overperforming its process at all. It’s delivering almost exactly what a genuinely leaky process predicts, goal for goal. The only thing standing between Newcastle and a considerably tougher league table is finishing form at the other end of the pitch, and finishing form is, by some distance, the least repeatable metric in the sport — the first thing to regress once a sample grows.

One result already hints at what happens when the wrapping comes off. At Elland Road, Newcastle’s underlying play merited something close to a 2-1 defeat — Leeds’ 1.79 xG against Newcastle’s 0.68 rounds cleanly to that scoreline. The actual result was 4-1. The process was already bad enough to lose comfortably on its own merits; on the day, it lost considerably worse than that.

The very next match, still outside this reconciled four-game window but worth naming on its own terms, put the other half of the argument on the pitch. Hull City had 21 shots at St James’ Park on 19 September — nearly double Newcastle’s own tally of 12, eighteen of them arriving in a second-half siege that had the hosts men behind the ball for long spells. None of that volume amounted to a big chance by Opta’s count; Newcastle actually finished with the better expected-goals total on the day, 1.41 to 1.24, which means Hull’s 21 attempts averaged out to barely a tenth of a goal apiece in quality. Run twenty-one shots at that kind of average probability through the binomial formula above and conceding at least once was odds-on regardless — and Newcastle did, in the 67th minute. What kept the afternoon from unravelling further was two clinical finishes inside the first seven minutes at the other end, and a goalkeeper who made the saves that were there to be made. It’s the whole piece in miniature, inside ninety minutes: a defensive picture that’s genuinely stretched, an attack currently bailing it out, and a final scoreline that gave no hint of either.

FotMob’s own model, running independently of Opta’s and refreshed through that Hull match, tells much the same story from a different angle: an expected-goals-against total of 9.1 across five games — barely different from the nine goals Newcastle have actually conceded — with only two sides in the entire division rated worse defensively. Different provider, same verdict, one gameweek further on.

Three clubs, one lesson

Newcastle aren’t operating in isolation here, and looking at where they sit against two other clubs makes the pattern harder to wave away.

Manchester City’s process matches their position — elite underlying numbers, elite results, with a layer of finishing luck against them sitting on top of genuine quality rather than substituting for it. Crystal Palace and Aston Villa’s processes also match their positions — poor underlying numbers, poor results, the table doing exactly what you’d expect it to do. Newcastle are the anomaly in this set. Their process is worse than Palace’s. It is marginally worse than Villa’s. And yet they sit fifth in this four-club comparison on points while owning the worst underlying number of the lot — which is exactly why the next block of fixtures matters more than the points column currently suggests, and why the work is worth starting now rather than once the table catches up and says so for them.

Tyneside has been here before

None of this is hypothetical for this football club specifically, and it’s worth naming honestly rather than skating past it. In 2011-12, Alan Pardew’s Newcastle finished fifth — an eight-year contract handed out on the strength of it. Buried inside that campaign was a defence recording the worst shots-against totals of Pardew’s tenure, carried by the best save percentage and the kindest finishing luck he ever had at his disposal. None of it repeated. Newcastle spent most of 2012-13 in a real relegation fight, escaping on the second-to-last weekend in 16th.

The lesson from that season isn’t that the collapse was inevitable — it’s that the window bought by good fortune went unused. Pardew’s Newcastle didn’t reinforce the areas the underlying numbers were flagging; a hot finishing spell got mistaken for the finished article, and the rebuild that should have happened during the grace period simply didn’t. That’s the part worth this side getting right instead, and there’s real reason to think the club is better placed to do it: considerably more resource behind the project than Pardew ever had, a sporting director’s office running recruitment with a plan rather than reacting to results, and a manager who is early enough in his own tenure that these numbers are a diagnosis rather than a verdict. History rhyming isn’t the same as history repeating. What happens next depends on what gets done with the time currently being bought — and that part hasn’t been decided yet.

The short version

Facing a lot of shots is dangerous even when hardly any of them are clear-cut, because the risk adds up across the many moderate chances a defence concedes, not just the handful of obvious ones a stats account counts by hand. A ranking built on just five matches of a rare event like “big chances conceded” is too small a sample to trust — it swings wildly by chance alone, which is exactly why Manchester City can sit mid-table on that particular list while topping the real one. Right now, Newcastle’s actual underlying numbers — chances created and chances conceded combined — are the worst in the Premier League, and the only thing keeping the league table from reflecting that is a finishing return in front of goal that ranks among the best in the division — and shooting form is, on the weight of everything football’s underlying numbers show about how it behaves, the least repeatable output in the sport, which makes the current rate far likelier to fall away than to hold.

None of that is cause for alarm, and it shouldn’t be read as one. It’s a side four months into a genuine rebuild, playing for a manager who is still installing his own way of doing things, currently being carried by a finishing spell that’s buying real, usable time. That’s not a small thing — points banked now are points a harder patch later won’t have to find from scratch. The job in front of Jaissle and his staff is simple to state even if it isn’t easy to do: tighten the process producing the shots faced, keep building the chance creation that’s currently running short, and treat the table’s current generosity as working capital rather than proof the job’s already done. Newcastle have lived a version of this story before, under Pardew in 2011-12, and what went wrong wasn’t the lucky start — it was that the club spent the following year finding that out the hard way instead of getting ahead of it while there was still time. That’s the difference that actually matters here, and unlike the stats above, it’s the one still open to being written the right way.

What do you think of our start to the season? Will Jaissle be able to tighten things at the back? Join the conversation on X @mytoonarmy #NUFC