What a season of 23 assists is actually worth

Last season Christos Tzolis got 23 league assists for Club Brugge. He's at Arsenal now. How much of that number comes with him?

This is the method from our earlier work, pointed at one player, in public, before anyone knows the answer. We'd rather be checkable than right.

The split

Our persistence research found a clean divide across 657 Premier League season-pairs. Volume counts such as passes, shots, dribbles and tackles keep 58 to 82% of their year-one gap twelve months later. Success rates on shooting and dribbling keep 12 to 32%.

Every attacking season splits along that line.

1. How many chances he created. That's a habit. It reflects his role and how he plays.
2. How many of those chances got finished. That's down to whoever was on the end of the pass.

An assist needs both, which makes an assist total an odd thing to buy a player on. Half of it is really a report on his teammates.

Tzolis, league football only

We've left out cups and internationals throughout. The opposition, the role and the rate are all different there.

The half that looks real

His chance creation has gone up every season: 1.74, then 2.45, then 3.92 key passes per 90. Key passes are one of the most persistent stats we measured, keeping 75% of their gap from one year to the next, and his are still climbing.

This cuts the other way to the rest of the piece: a rising line isn't a stable one. Our retention figures describe players whose level holds steady, not players who are still improving. For someone getting better this quickly, the average of his past seasons undersells him, and nobody knows yet where the line levels off.

The half that leans on other people

Now the success rate. Of the chances Tzolis created, how many turned into goals?

His teammates converted his chances at 1.41 times the rate they'd managed across his previous two seasons.

Replay last season at his own earlier rate and those 134 key passes give you roughly 16 assists, not 23. On those numbers, about seven of his assists came from the finishing rather than from him.

Why we can't prove this for him

One player's season is around 130 chances. That's a small sample, and small samples come with wide margins of error. We can't show that Tzolis himself was lucky, and anyone claiming that from one season of data is going too far.

What we do have is the population. Across 657 season-pairs, success rates behave like this as a rule. So the claim is actuarial rather than a prediction:

Players whose success rate jumps this far above their own usual level tend to give most of it back. By that standard, this much of the 23 looks like the spike rather than the player. That's a statement about the population, not a forecast for one man.

An insurer can price a house without knowing which one will burn down. Same idea, same limits.

A closed case, to check the method

Tzolis is still in progress. Here's one that's finished.

In 2023/24 Dominic Solanke scored 19 Premier League goals for Bournemouth. Tottenham paid a reported £65m for him that August, Bournemouth's record sale. Using only what could be seen beforehand:

The habit held almost exactly: 1.03 shots on target per 90 in the season that sold him, 1.06 in his first season after it. He kept getting into the same positions and getting his shot away just as often.

The success rate didn't hold. It was 50% in the season that sold him and 35% after the move, against 21% the year before.

The catch: that 35% comes from 26 shots on target, which gives a 95% interval of 19% to 54%. The 50% from the selling season sits inside that range. So even in a case that looks this clean, the drop isn't established for Solanke as an individual. It's consistent with the population finding, but it can't prove it on its own.

We've left his 2025/26 out entirely. He hurt his ankle in pre-season, needed surgery and made fifteen appearances. Injury cuts the sample, not the rate, and three goals from eleven shots on target measures nothing at all.

Not a transfer verdict. Fee, fit, role, minutes and injuries all sit outside this, and every one of them matters more than anything here. The only claim is that the number clubs reacted to was part habit and part spike, and that you could see the split in advance.

Our call, made in public

Tzolis has played 211 Premier League minutes over three appearances. He's creating 3.41 chances per 90, against 3.92 last season. His teammates have converted 12.5% of them, against 17.2% last season and 12.2% before that.

Three appearances off the bench prove nothing. But notice how it differs from Solanke's lost season: these minutes are few because it's September, not because anyone's injured. This season will add up, and by May it'll be a proper test.

Our position, written down now rather than after the fact: the chance creation should largely survive the step up, and the assist rate per chance should settle closer to 12% than 17%. If both hold, the method worked on a case we called in advance. If the creation collapses, we were wrong about the more important half.

Method

League football only. Retention figures come from 657 Premier League season-pairs across 2022/23 to 2025/26, measured within position. Player seasons are season totals (sources below), with no event or tracking data. The intervals are Wilson intervals on the underlying counts.

If this call goes against us, it'll be written up here just the same. That's the deal with calling things in advance.

Data sources

• API-Football: season statistics for Tzolis (Fortuna Düsseldorf, Club Brugge, Arsenal) and Solanke (Bournemouth, Tottenham), league football only.
• Solanke's fee and his 2025/26 ankle injury: press reports at the time.

Edited for style on 22 September 2026, and the persistence figures above corrected the same day after a data-labelling fix (see the persistence piece). Tzolis's own numbers and the call are unchanged.

Corrected on 25 September 2026. The persistence figures quoted above were updated after a data fix to how split seasons are counted: 657 season-pairs, volume counts 58 to 82%, success rates 12 to 32%, key passes 75%. Tzolis's own numbers, Solanke's and the call are unchanged.