A few weeks ago I was doing what I usually do when I find interesting takes on Juventus — I did try to understand the point of view of a professional about which midfielder my team should target in the summer. The specific exchange was with @CharlesOnwuakpa a professional in recruitment, and his point was a good one: Juventus need a midfielder who operates on the right side of the pitch. Thuram pulls left. Koopmeiners pulls left. Locatelli pulls left in his opinion.
That got me thinking about two things. First, is he actually right about Locatelli — or is that a general impression that doesn't hold up when you look at the data? And second, if the analysis supports it, can I actually find someone with my own tools? That became this piece, which turned out to be considerably longer than I expected.
I've split it into two halves. The results come first — names, brief reasoning, recommendations — for people who want the payoff without the journey. The process follows — methodology, filters, decisions — for people who want to follow the logic and tell me where I went wrong. Both halves assume some familiarity with the metrics I use, but I've tried to explain the less obvious ones as I go.
What Juventus actually need
Before getting to candidates, it's worth being precise about the problem. Locatelli is good enough and well-rounded enough that he can pair with almost anyone Juventus put next to him — he's the captain, the foundation, the player the whole midfield is built around. The constraint isn't Locatelli. The constraint is that Thuram seems to be staying and is appreciated, and Thuram not only operates on the left but has to. That leaves Juventus with two routes: either they sign a new defensive midfielder and let Locatelli shift slightly higher and more to the right, or they sign a centre midfielder with a natural right-side tendency who provides the positional balance the current squad lacks.
The recommendations
Eduardo Camavinga — the big name you can't ignore
Camavinga's radar tells you what he is and isn't quite clearly. The defensive metrics are strong — tackles and interceptions, defensive actions, tackles true win rate all well above average. That's the profile of a midfielder who can sit and protect, which would let Locatelli operate higher and wider to the right, while also create some interesting dynamics with Cambiaso given the frenchman ability to cover at left-back. The creative output is another story: his passes aVAEP and final third progression are both modest, and the under-pressure passing proxy at -3.88 is the worst of any player in this group by a significant margin.
For whatever reason Real Madrid seem open to selling him — reports put the figure at 45 to 50 million euros, which against the roughly 60 million Juventus spent on Koopmeiners represents a meaningful market opportunity. The problems are real too: every Premier League club needing a midfielder will be in that queue, Juventus might not be in the Champions League next season, and Capology has him at 7.5 million net per year — would be top three earner at Juventus right now still considering the absurd Vlahovic contract. The data wouldn't rank him first on pure metrics, but you have to at least model this scenario. These windows don't reopen.
Angelo Stiller — probably the best pick
Stiller's radar is the most coherent of the group for what Juventus need. Passes aVAEP in the 90th percentile, passes and carries into the final third well above average, while the opponents' aVAEP responsibility of 1.30 tells you he's doing this against quality opposition. The defensive numbers are not solid, but in the past he had good numbers. The under-pressure proxy at -2.07 is negative, which is worth watching and possibly evaluate with better dara, but it's better than Camavinga's.
A year older than Camavinga at 25 and paid significantly less at Stuttgart. By Transfermarkt's reckoning they're valued similarly, which already tells you something about where the value is. I consider him the best pick because he does most of what Locatelli does while bringing more forward creativity — a player who could complement Locatelli in a double pivot or, further down the road, alternate with him as the primary midfield anchor. There's a slight left lean in his heatmap — we'll get to that — but as I'll show later, looking at his actual passing behaviour after receiving sideways changes the picture considerably.
Mateus Fernandes — less of a gamble than the name recognition suggests
That overall aVAEP of 0.69 is the highest in this entire group. The defensive profile is exceptional — tackles and interceptions, defensive actions, defensive aVAEP all well above average — and the under-pressure passing proxy at +1.29 is the only positive number among the finalists. The progression numbers (passes into final third, take-on and carries aVAEP) are more moderate, which is probably the main limitation if Juventus need their midfielder to actively build from deep to give more freedom to Locatelli.
West Ham's likely relegation creates the window. The fee could end up below what his profile merits if they're forced to sell, and he's young enough (born 2004, same as Avdullahu) that you're buying his ceiling as much as his present form. He plays centrally, which solves part of the positional problem. The race will be with Premier League clubs who are already watching, and Juventus would need to move with some urgency.
Leon Avdullahu — a real gamble with a real profile
The shape is distinctive — strong on passes aVAEP and the passing metrics generally, with visible gaps in creative output (npxG assisted very low, take-on and carries aVAEP weak). For a CDM operating in a controlled, possession-maintaining role, that gap is manageable. He's playing against strong opposition — opponents' aVAEP responsibility of 1.18 — and getting meaningful minutes at Hoffenheim at 22. The radar is a bet on the floor being higher than the shape suggests.
This is scouting-as-projection. He's at a top European league, playing real minutes, and his passing skill is confidently above average in the Bayesian model. He's also probably a cleaner passer than Fernandes, even if Fernandes' overall profile is more complete. At the right price and in the right market moment, this is a player worth tracking closely — not a guaranteed hit, but the kind of profile that tends to look cheap in hindsight if the development goes right.
Jordan Holsgrove — hear me out
This radar is better than it has any right to be for a 26-year-old at Estoril. Passes and carries into final third in the 90th percentile, npxG assisted solid, defensive actions and tackles decent. The under-pressure proxy at -2.33 is the main concern — along with the obvious question of competition level — but given the context you still have a player whose shape in several key metrics sits at Stiller's level. The opponents' aVAEP responsibility of 0.65 reflects the Liga Portugal context, and the wide credible intervals in the passing skill model reflect the sample quality issue.
He's probably not coming to Juventus. He's 26, he's at Estoril, the Portuguese league corridor to Serie A has thin sample in my data. But Spalletti won a Scudetto with Lobotka, who was at Celta Vigo before Napoli took a chance on him. These things don't happen often or predictably — but they happen. Worth keeping on the list purely as a market surprise option.
Didn't make the cut — worth a mention
Jerdy Schouten, Joey Veerman, Vitaly Janelt, and Orkun Kökçü all surfaced during this process and were cut at different stages for different reasons. Kökçü is probably the most interesting omission. He'd make Locatelli's life easier by operating further forward, his numbers are good, and the Turkish Super Lig has a 55% transfer success rate to Serie A in my data — higher than every league except the Premier League in that corridor. But his heatmap makes the problem immediate:
Almost all of Kökçü's activity sits in the left-attacking half. He's an excellent footballer operating in a very specific zone of the pitch, and that zone is exactly where Juventus are already congested. He'd stack on the same side as Thuram rather than balance it.
There are other bigger names in the cuts too, and you'll understand why once you've read the process section.
The process
For those who want to follow how this was built — and ideally point out what I've missed.
Starting point: team similarity and weighted rankings
I began the same way I started my piece on Juventus' summer transfer window — by running my team style similarity model against Spalletti's Juventus to find the clubs playing most similarly to them, then using those similarity scores to weight my player rankings. The rankings themselves are Z-scores across a set of metrics I selected for each position group, combined into a single score. They're a simplification — a radar compressed into a number — and should be treated as a starting point rather than a conclusion.
From there I took the top 150 CMs and top 150 CDMs. Then I applied one specific filter to each group: for CDMs, I searched for volume of progressive actions in build-up possession chains, because if we want to reduce Locatelli's workload in that area, we need a CDM who can carry some of that function. For CMs, I searched for volume of wall passes in third man combinations — because Locatelli leads my entire dataset in that metric for his position group, and finding someone who can take that role relieves pressure on him to do a play after the progression.
Locatelli's progressive actions from deep in 25/26 — 158 in total, 139 passes and 19 carries. The spread across the pitch is genuinely wide, which tells you this is a complete build-up midfielder rather than a specialist on one side. He's doing an enormous volume of this work for Juventus, and the point of recruiting alongside him is partly to distribute that load.
These two filters produced 22 CMs and 22 CDMs who sat in the top 80 of both their weighted ranking and the relevant skill volume. Forty-four players is still a lot, but they're all reasonably pre-qualified.
Filtering by transfer corridor
This is where the transfer success pipeline I've been building becomes directly useful. The question isn't just whether a player is good — it's whether players from that league have historically succeeded when they've moved to Serie A. For a club like Juventus, you don't want to be making probability bets on corridors with thin or negative track records.
The transfer success matrix — success rate and sample size for every origin league × destination league pair in my dataset. Success here means a player maintained both their playing time share and their VAEP performance within 10% of their origin baseline, after adjusting for the difference in league difficulty between the two competitions. The Serie A column is what matters for this exercise. Higher is better — darker blue means more transfers came through successfully.
The Scottish Premiership has a 14% success rate to Serie A across all positions, and 0 for 2 on CDMs and CMs specifically. That's enough to cut. MLS, lower English leagues, and the Belgian Pro League (33% overall) were also removed — not because those leagues don't produce good players, but because the data doesn't support confidence in a Juventus-level move from those corridors. This cut removed a handful of interesting names.
After the corridor filter: 25 players.
Age filter
Players born in 1997 or earlier were removed. At 28 or 29, you're paying for what a player has already done, and their value as a market asset is declining rather than growing. Schouten, Maxime Lopez, Pablo Rosario, Thiago Maia, Aleix Garcia, and Bruno Guimarães all came out here. Bruno deserves a specific mention: if he were available at a reasonable price, he'd probably be close to the ideal profile for what Juventus need. He isn't, and 60 to 70 or more million for a player heading toward 30 isn't a conversation Juventus should be having regardless of how good the fit looks on paper.
After the age filter: 17 players. Two more cuts for injury record concerns, two for a position clustering assignment error in my model. Thirteen remaining.
Phase two: where do they actually play?
Thirteen players is still too many to evaluate in detail, and not all of them solve the positional problem. The next stage was heatmaps — working out which of the 13 genuinely play centrally or with a right-side tendency, versus which would replicate the left-heavy imbalance Juventus already have.
This is the stage where Kökçü, Veerman, Janelt, and Elliott Anderson all came out. Good players, wrong spatial answer for this specific question. Wharton is worth showing because he actually passes the spatial test convincingly:
Wharton's activity spread is clearly right leaning. The heatmap passes the test. The problem is he’s probably with Anderson the most sought after English midfielder right now and that’s not the table Juventus can sit at.
Stiller's heatmap has the slight left lean that gave me pause initially. But look at the right side — multiple hot spots, genuine presence across the width. Compare this to Kökçü's heatmap and the difference is stark. This isn't a player who avoids the right; it's a player whose most active channel happens to lean slightly left. There's more to check — and I will check it — but this doesn't disqualify him.
After removing the spatially misaligned players and Zanocelo for insufficient minutes at Ceará (around 1,200 in 2025), eight players remain.
Passing skill
Passing Skill Above Average (PSAA) — estimated from a Bayesian hierarchical logistic regression that controls for pass difficulty and competition level, an upgrade of this. The point is to separate players who complete passes because they're good passers from players who complete passes because they play simple ones. Blue dots are confidently above average; grey straddles zero; the red dot is Wharton, confidently below average. Dot size reflects pass count. The error bars are 94% credible intervals — wider bars mean less certainty, usually because of smaller samples.
Camavinga and Stiller sit furthest right at 100% probability of being above average, both with large sample sizes. Avdullahu is at 99.9%, Fernandes at 98.5%. Holsgrove sits in the uncertain zone at 94.1%. This chart matters because it's telling you something independent of volume or role: these players are actually good at passing, not just the recipients of good passing situations.
Reception into progression
The last specific skill: receiving a pass and immediately turning it into something progressive. This matters for Juventus' midfield specifically because Locatelli's value is partly built on exactly this — he receives in tight spaces and keeps the team moving forward. We want his partner to be able to do some of that too.
Volume ranking for number of this receptions every 30 minutes in possession. Holsgrove leads at 26.57, Stiller second at 23.36, Avdullahu third at 20.23. Camavinga and Fernandes are further down the list.
Value ranking — aVAEP generated from receptions followed by a progressive action, normalized per 30 minutes in possession. Stiller leads at 0.1263, Holsgrove second at 0.1239. Avdullahu drops to 0.0953. Camavinga and Fernandes come in at 0.1046 and 0.0927 respectively.
The pattern is consistent: Stiller and Holsgrove are the best at this specific skill, by both volume and value. Avdullahu is solid on volume but the value drops off. Camavinga and Fernandes aren't strong here — which shapes their position in the final recommendations despite being interesting for other reasons.
The final check: directional passing after receiving sideways
At this point I was confident in Stiller but still wanted to address the left-lean question properly rather than just wave it away. The question: when Stiller receives a sideways pass in midfield — the most common midfield receipt situation — does he show a strong directional preference in what he does with the ball next?
I don't have foot-specific passing data in this dataset. As a proxy, I split receptions into "left to right" (the ball arriving from his left, which typically means he's receiving on his weaker side relative to the direction of play) and "right to left," then looked at the distribution of subsequent passes in each case. If he avoids passing into certain zones when facing one direction, that would show up as a gap in the sonar.
149 left-to-right receptions, 138 right-to-left — almost exactly balanced in volume. The patterns differ slightly: receiving left to right, his passes are slightly more concentrated and forward-facing with higher VAEP; receiving right to left, the distribution spreads wider. But there's no obvious avoidance in either case — no dead zones, no systematic refusal to play in any direction. The slight left lean in his heatmap looks like it reflects where Stuttgart tend to initiate play, not a personal reluctance to operate on the right.
That's the reassurance I needed. And the corridor number does the rest: Serie A recruits at a 53% success rate from the Bundesliga in my data — tied third highest, with 30 transfers to draw from. That's a real sample with a real result.
So.
Stiller is the name. He does most of what Locatelli does, brings more forward creativity, distributes without obvious directional limits, and comes from a corridor that works for Serie A. At 25 he's entering peak years, not leaving them.
Camavinga is the opportunity you can't ignore even if the data doesn't rank him first — 45 to 50 million for a Real Madrid starter is a number Juventus might not see again for a player at that age.
Fernandes is next tier with Avdullahu a step lower: both young, both from corridors that work. Fernandes is the more complete profile; Avdullahu is more of a bet.
Holsgrove is the biggest bet, probably an unreal scenario.
There are real limitations here. No video, data that's mid-season rather than end-of-season, which means sample sizes for some players are still building. A transfer success model that doesn’t work on huge numbers. And Juventus have a scouting operation with access to everything I don't have.
I'll probably keep covering Juventus positions over the coming weeks, partly because it's interesting and partly because this newsletter is meant to function as a public portfolio for the kind of work I want to do professionally. If you have thoughts on the methodology — things I've missed, things I've overcomplicated, things where the logic breaks — I'd genuinely like to hear them.
See you soon.
Data: Data from the corresponding leagues + UCL and UEL until 10th March 2026. Claude was used to correct the draft of this post.
