There's a moment in one of Ted Knutson's episodes of The Transfer Flow — I couldn't tell you exactly which one if my life depended on it — where he says something that stuck with me. The gist of it was that someone should actually try to measure which leagues produce more success when you recruit from them. I thought: that's interesting, can I do it?
Well, I can try at least.
Before I get into what it found, I want to be clear about what I mean by "success". The naive version is: did the player play well at their new club? The problem is what "played well" actually means. A striker who scores 8 goals at a mid-table Premier League side might be a massive success story or a significant underperformance depending on what they were doing at their previous club. Without a baseline you're not measuring success — you're measuring output in a vacuum.
The version I built tries to measure something more precise: did the player maintain what they were doing before, in a new environment? That framing matters because when a club pays money for a player, they're doing it on the basis of past performance — the implicit bet is that the player will at least reproduce that in the new context. So that's the bar.
What success actually means here
The pipeline is built on atomic VAEP data — a framework that assigns a value to every individual action based on how it changes the probability of scoring or conceding. I'm not going to go deep into the theory here, but the short version is that it lets you measure player contribution across all actions on the pitch, giving value much like xG does for shots — just for everything.
For each transfer in the dataset — six seasons, 2020/21 through 2025/26, across the major European leagues plus Brasileirão, Argentine Liga Profesional and MLS, up to the 17th of March 2026 — the pipeline computes two things at origin and destination.
The first is VAEP rate: how much value the player was generating per minute of game time. The second is exposure: what share of their team's available minutes they were actually playing. Both matter. A player who performs brilliantly in 400 minutes at their new club while sitting on the bench for 2,600 is not a success story — the club isn't playing them, which is its own verdict, even if in some cases it represents more of a market malfunction than a genuine failure. We just have to work with what the data can tell us.
The baseline at origin uses a backward time decay, so that games closer to the departure date carry more weight than games from 18 months earlier. The evaluation window at destination runs forward from the first appearance, up to 18 months, with an early-game discount — because adaptation takes time, and penalising a player for their first few games before they've settled in produces unfair assessments. The 18-month ceiling is also deliberate: in Ted's words, if you're wrong you need to cut your losses as fast as possible, and a tool that takes three seasons to return a verdict isn't particularly useful for that.
Transfers land in one of four buckets: success (exposure maintained, VAEP maintained), performance failure (playing but underperforming), selection failure (not being picked — the club's verdict), or unscored (not enough data). Both the exposure and VAEP thresholds have a 10% tolerance built in, and the VAEP threshold is adjusted for the difficulty gap between the origin and destination league.
What the data covers
Six seasons. Cross-league transfers only — intra-league moves are excluded. Cup competitions and European club football are also excluded, because mixing Champions League and domestic league data creates sample and context problems I didn't want to deal with. Calendar-year leagues (Brazil, Argentina, MLS) are included but with lower confidence — thinner samples, sparser data.
The output is a matrix: origin league × destination league × position group. For each cell, you get a success rate and a sample size.
I'd encourage you to spend some time with that before reading my interpretation of it.
What it shows
The individual cases are a good place to start, because they illustrate both what the pipeline gets right and where it runs into honest limits.
Evan Ferguson at Roma is publicly talked about as a failure — and from a footballing narrative standpoint, Gasperini seemingly didn't fancy him. But the pipeline flags him as a success. His share of available minutes increased by over 1,300% compared to his West Ham baseline, and his VAEP rate is nearly 280% higher. The data isn't wrong — he genuinely performed well when he played. The pipeline just can't see the coaching relationship, and that gap is real.
Achraf Hakimi, on the other hand, is classified as a performance failure for his first 18 months in the data. His VAEP rate dropped by 17% at the destination relative to his origin baseline. Today you probably wouldn't describe that move that way — but that's partly the point. The 18-month window is a deliberate constraint, not a complete picture of a career. The pipeline is making a judgment call about an evaluation period, not a verdict on a player.
The corridor-level findings are where this becomes genuinely useful for recruitment thinking rather than just interesting as an exercise.
The most counterintuitive number in the whole matrix is the Jupiler Pro League recruiting from the Championship — 83% success rate on 12 transfers. That's a real sample, not a quirk of two or three cases. And before you ask: no, this isn't a Union Saint-Gilloise / Tony Bloom effect, because none of those transfers went to USG. It seems to be a genuine corridor story — something about the profile of Championship players fitting Belgian Pro League football well enough that the adaptation rate is unusually high.
A few other things are consistent enough to be worth stating. The Bundesliga corridor into Serie A sits at 50% across all positions with 30 transfers — a meaningful sample, and part of why it featured heavily in my Juventus piece. The Turkish Süper Lig corridor into Serie A is quietly interesting: 50% overall, limited volume, but consistently above what you'd expect given how rarely that corridor gets taken seriously.
The corridors that don't work are almost as informative as the ones that do. When a league shows a low success rate to a particular destination — especially with a decent sample — it's telling you something about the gap in tactical context, physical demands, or competitive level that makes adaptation genuinely difficult. That's worth knowing before you start the detailed work on a player, not after.
What this is and isn't
This is a base rate tool. It tells you, historically, how often players from League A have succeeded when they've moved to League B — and I think it does that well enough to be useful, at least at the level I actually operate.
What it doesn't do is tell you how any individual player will adapt. A corridor with 40% success still produces successes — you're just working against a harder prior. It also can't capture everything that makes a transfer succeed or fail: injury, personal circumstances, tactical fit within a specific system, the quality of the coaching environment at the destination. Those things matter enormously and they're invisible in this data.
The sample size problem is real and I want to be honest about it. Six seasons sounds like a lot until you start slicing by position group — suddenly you're looking at cells with four or five CDM transfers from a given league, and the percentage becomes more noise than signal.
The pipeline also treats loans and permanent transfers identically, which is a simplification worth flagging. A player on a six-month loan who returns to their parent club generates a transfer record that gets scored the same way as a permanent move — even though the context, motivation and time horizon are completely different. That's something worth fixing in a future version. So is the absence of any age filter at the point of transfer — a 22-year-old moving leagues and a 29-year-old making the same move are not the same problem, and collapsing them into the same success rate obscures something real.
Why I built it this way
Part of the honest answer is that Ted Knutson said someone should, and I happened to have the curiosity to try. The more considered answer is that I think it would genuinely help me in the transfer evaluations I do and that’s what I have available — and the Juventus piece is a decent example of that already, where the corridor data was doing real work rather than just sitting in the background. If you happen to have better data at your use and what to give me access, I’d be glad to have fun with it.
This newsletter is, among other things and at the end of the day, a public version of the kind of work I'd like to do professionally.
If you have thoughts on the methodology, or you've seen something similar done differently that's worth knowing about, I'd genuinely like to hear it.
See you soon.
