Nobody is buying the best players in the game
Three gameweeks of underlying data reveal the transfer targets you should actually buy ahead of gameweek 4.
Three gameweeks into the season, three quarters of managers have missed the point entirely. 
João Pedro sits in 72.4 per cent of squads despite producing an xGI per 90 of 0.68. Meanwhile, 
Saka sits in just 11.4 per cent of teams, and 
Tavernier is chilling in 3.2 per cent of squads while putting up 0.62 xGI per 90 and 28 defensive actions.
Points are what happened. Underlying numbers are what happens next. When you look across the next five weeks, the gaps between ownership and actual threat are wide enough to drive a team bus through.
Master Score combines underlying production, form, and fixture difficulty on a 0-100 scale over gameweeks 4 to 8.
The model rates B.Fernandes top of the entire game with a 99.0 Master Score. That is not nostalgia. Across three starts and 270 minutes, he has put up 3.05 xGI, averaging 1.02 xGI per 90. He has scored three goals from 2.25 xG, added an assist from 0.8 xA, and chipped in with 16 defensive actions. Yet 53.8 per cent of the game is managing without him.
If 12.0m is too rich, his new clubmate Mbeumo is generating almost identical numbers at 7.9m.
B.FernandesMbeumoSeason to date across 270 minutes each in gameweeks 1 to 3.
Mbeumo has accumulated 2.31 xG and scored twice. His goalsMinusXg sits at -0.31, which means he is actually underperforming his chances slightly rather than riding a lucky streak. He delivers 0.93 xGI per 90 and sits on a Master Score of 94.2. Only 25.5 per cent of managers own him.
Then there is Manchester City. Haaland is Haaland, owned by 71.2 per cent of managers, boasting 2.46 xG and 0.99 xGI per 90. That is fine. But beside him, Cherki is quietly breaking the system at 7.8m.
0.99Cherki xGI / 90Matches Haaland30.3%Cherki ownership7.8m midfielder70.4MCI fixture ratingGameweeks 4 to 8Data covers gameweeks 1 to 3. City hold the easiest attacking run over the next five gameweeks.
In just 173 minutes across two starts, Cherki has racked up 25 points, two goals, two assists, and an xGI per 90 of 0.99. His goalsMinusXg of 1.44 tells us he will not keep finishing at this ridiculous rate forever, but his 1.35 xA shows he is feeding the machine. With Manchester City holding a fixture rating of 70.4 over the next five gameweeks, he is the clearest route into their attack.
Defensively, everyone is looking at the wrong end of the price list. 
Gabriel costs 8.0m and takes up a chunk of budget. 
Mitchell at 
Crystal Palace costs 4.5m, is owned by 5.7 per cent of managers, and gets a fixture rating of 60.7 over the next five weeks, starting with 
Ipswich at home in gameweek 4.
Man CityCrystal PalaceArsenaleasierharderInternal model rating, 0 to 100. Not FPL's fixture rating.
Fixtures scored 0 to 100 on internal model ease rating. Higher score means easier match.
Sample sizes after three weeks are thin. One deflected shot or a soft penalty skews three weeks of points completely. But xGI per 90 and fixture runs cut through the noise quickly.
If you have a transfer to make before gameweek 4, lock in Mbeumo or Cherki before the rest of your league notices what the numbers are saying.