I hate dragging soccer into this so early again. It's been the topic of most of my last few newsletters, and half of you are still nursing a World Cup hangover that preseason friendlies won't fix. 

But Sky Sports put a graphic up this week, and I haven't been able to leave it alone.

Most expensive British transfers ever. Morgan Rogers, £117m. Elliot Anderson, £116m. Then Bellingham, Rice, Grealish, Kane. Four of the six inside three years, two of them since June.

I have no idea what's running inside Chelsea's recruitment department, and neither does anybody outside that building, no matter what LinkedIn told you this week. What I can see is where the tools are pointed, because that part's the same everywhere.

Every one of them does the same two things. It finds you names, and it tells you roughly what they're worth. Both are useful. Both are also the cheapest work anybody does in a transfer.

That largely held everywhere I looked this month. Sport by sport, use case after use case.

The transfer portal opens, and a program has under 14 days to evaluate, price, and close a roster. It’s essentially NFL free agency at 4x the speed, except the contracts are secret and half the agents are strangers. That's more tape than any staff on earth gets through.

So the screening went to the machines. Every available body ranked on scheme fit and need, shortlist handed to humans, who at most programs now include somebody doing a job that didn't exist three years ago: a salary-cap analyst, in college football, mapping a roster against a finite pot.

Fine. Screening is also the cheapest link in the chain, and everyone's automating it, so your advantage lasts roughly as long as it takes a rival to sign the same contract.

Pricing is where it falls apart. The revenue-share cap sits at $21.3m for the year ending next June, and Opendorse found Power 4 schools adding an average of $13.5m in third-party NIL on top of it. That makes the cap a floor of around $30m rather than a ceiling, and nobody has to disclose a dollar of it. There's no contract database and nothing resembling a Spotrac for the SEC, so agents swear every starter is worth seven figures and the GMs lie right back at them.

Best screening in the country won't stop you overpaying a million for a guard.

A draft pick isn't a purchase. The rookie deal gets fixed by slot before anyone shakes hands, so what you're buying is cheap exposure to a second contract four or five years out. Scouting was always about that strike price. Nobody could put a number on it.

Caleb Downs skipped the 40 at the combine and again at his pro day, so teams ran computer vision over his college tape and built a speed figure they could set against Next Gen Stats. Karim Kassam at Teamworks told ESPN what came back: Downs just isn't that fast.

His draft slot held. What moved was the option. The data put him at box or slot safety rather than the deep-middle guy, and the market pays those roles very differently. Kassam's read on edge rusher Arvell Reese went the same way. Best at his position in the class, and still less likely than David Bailey or Rueben Bain to reach double-digit sacks. Double digits is the number that sets an edge rusher's price.

Which lands harder this year, with the cap over $300m for the first time at $301.2m and rookie deals the last cheap labor anybody has. Miss enough of these, and you're Miami, carrying a record $179m of dead money for men who don't work there.

Les Snead, the LA Rams' famed GM, recently told a room in Phoenix they'd turned the whole draft over to Claude. He was joking, of course. But every joke has an element of truth to it, and the truth here is what Snead says he wants: an assistant lieutenant, his phrase, that'll disagree with him because it has no promotion riding on the answer. A scout might think twice about telling his GM he's wrong on a player. Software won't. 

Every NBA club pulls from the same machine learning models built on the league's tracking data, used on front office calls as much as coaching ones. Ask an executive what they point them at, and you get nothing, which Yahoo turned into an entire article. 

The one answer anybody's given came from Daryl Morey, who back when he ran the Sixers said they treated one "almost like one scout."

Odd thing to guard. Thirty teams doing identical math on identical data land in roughly the same place, so whatever edge exists has to live somewhere the model can't see.

In this league that's the cost of the trade itself. Cross the second apron at $221.686m, and you lose the right to combine salaries in a deal, the taxpayer mid-level goes, and a first-rounder seven years out freezes solid. Those are real costs and not one of them appears on a P&L.

So somebody prices a frozen 2034 pick before the call gets made, which is a finance job sitting inside a basketball building. In Philadelphia it was Morey's, until a second-round sweep in May ended six years of it.

Same problem up here, opposite response. No GM in this league is carrying the cap math around in his head, because the NHL built the tool for them.

The ceiling climbed $8.5m to $104m this season, which sounds like room until you remember all 32 GMs got that raise on the same day. What changed was the difficulty of executing a deal, between long-term injured reserve, retained salary, and compliance that you check lineup by lineup rather than once a summer.

So the league and SAP put a cap projector inside the Front Office app. A GM drops in a projected lineup and learns on the spot whether it clears. Usage ran 16x normal around the March 6 deadline, which tells you exactly what they were modeling.

That's the whole product. It grades nobody and projects nothing. It tells you at 2 PM, with an hour left, whether the deal you shook on at 11 AM is one you're permitted to do.

Which is the single most useful thing I can tell anyone buying software this year. Adoption follows whatever constraint binds you. In hockey that's legality, so the legality tool is the one they open sixteen times over.

MLB front offices evaluate amateur pitchers on camera data. Velocity, spin rate, release point, how the elbow loads under stress. That's what decides whether a club spends a first-round pick on an eighteen-year-old, and what the same arm is worth in a trade three years later.

Clubs used to collect it themselves. They wired amateur venues and left rigs running, which is how a bank of cameras ended up at a California high school recording every pitch thrown by a kid scouts had marked as a first-round arm. The club with the best coverage had the best board.

The league ended that on January 1. Clubs can't contract with Trackman, Blast Motion or Edgertronic directly now. They buy a league-approved package, and MLB sends the same data to every club. Joe Doyle at Over Slot expects analytics staff cuts when the 2027 CBA lands.

The rules did the rest. Teams spent a decade paying premiums for catchers who framed pitches well, then the automated ball-strike challenge system made challenge discipline the skill that matters. A catcher signed through 2029 on his framing numbers is impaired, and the player didn't do a thing.

Two ways a valuation model stops working. Your rivals get your inputs, or the rulebook changes what the output was worth. Baseball managed both in one winter.

The recruitment tools sold into European football now work like this. Type "left back under €5m who can start in the Bundesliga," get a ranked list filtered for budget and work permits, then pull a failure-risk score on your top three before anyone bids. Gemini Sports does the squad-planning version of this: a Miami-built, mobile-first hub that pulls scouting reports, video, event data and the financial picture into one view a sporting director can actually use at 1am on deadline day, with clients from AS Monaco to QPR. The category is getting heavy Sky Sports coverage as clubs wake up to it.

Another sharp version belongs to Tony Bloom, a professional gambler who worked out that the data pricing a match also prices the players in it. His firm Jamestown handles recruitment for Brighton, Union Saint-Gilloise, Como and Hearts, where he paid £9.86m for 29% and exclusive Scottish access. Brighton cut most of their full-time scouts in 2024.

The buying works. Brighton's 2022/23 accounts showed a record £122.8m profit, £121.4m of it from selling players.

Selling, though, is what no model touches. Ask Bloom how he times it, and you get a standing rule: the best players always go. Fifteen years of that working doesn't make it a system.

As the Sky Sports graphic showed at the start of this piece, Villa took £117m for their best player and Forest £116m for theirs. No software told either of them when.

Fifteen years ago, a club that could reliably spot an undervalued player had something worth real money. Now, that's gone. Teamworks is worth $1.5bn because the product it sells works the same way for everyone who buys it, and by now everyone has. League by league.

What keeps striking me across all six of these leagues is that the decisions which settled the outcomes were the ones no software went near. Villa banking £117m came down to which summer they sold, nothing more. That catcher lost value because a committee changed a rule, not because anything about him changed. And a pick that freezes in 2034 gets weighed by whoever happens to be in the room, on instinct, in the twenty minutes before a deadline.

None of that lands anywhere in a set of accounts, which is my best guess at why nobody has built for it. Software gets sold against budget lines, and there's no budget line for a cost that never arrives.

So if you want to know where you're exposed, go looking for the decision nobody in your building has ever put a number on. I've yet to sit with a business that didn't have one.

It's what we do at pointcfo.co.