The Elite Edge: What Separates the Top 0.1% of FPL Managers

Why this changes decisions

Take Liverpool at home to a low-block, defensively stubborn mid-table side.

For an attacking asset — Isak, or whoever's on penalties that week — this can be a Category 1 mismatch. Liverpool generate serious non-penalty xG at Anfield, and a side that parks ten men behind the ball will eventually concede volume: box entries, cut-backs, second balls, corners. Sit deep for ninety minutes against that and something usually gives.

For a defensive asset — Kerkez, or Alisson — the same fixture is noticeably less appealing than it looks on paper. If that opponent scores 80% of its goals from direct counter-attacks and set pieces, it's precisely the kind of side that has one shot all afternoon and scores with it. Low block, high clean-sheet wipeout risk. Same fixture, same FDR, two completely different answers.

So top managers keep two schedules, not one: an attacking board and a defensive board. And when captaincy comes round, they don't default to whoever scored a brace last weekend. They scan the board for the single widest positive gap between one club's attacking output and a specific opponent's defensive weakness. The armband goes to the matchup, not the name.

What it looks like in our own numbers

Our fixture matrix already does this split. Every fixture gets a separate, position-specific difficulty score from the Dixon-Coles engine — 0 is the easiest fixture on the board for that position, 100 the hardest — built from bookmaker-calibrated expected goals for and against. Put those next to the official FDR for a handful of upcoming fixtures and the single number starts to look very thin indeed:

Read across those rows and FDR starts to look like a weather forecast that only reports the average temperature. Chelsea's two 3s are the same number on the FPL site, but for Chelsea's attackers the trip to Sunderland is about as good as fixtures get (3.2 out of 100) while their defenders are looking at a thoroughly mid-table 35. Everton's FDR 2 at Hull is the reverse: the model has Everton scoring barely a goal (1.04) but conceding only 0.62 — a fixture for Everton's back line, not their forwards.

Then there's Liverpool v Arsenal in GW9: an FDR 4 both ways and a genuinely grim fixture for either set of forwards (83 and 94 out of 100), with 49% and 42% clean-sheet odds for the two defences. If you own premium attackers on both sides, that isn't a "tough week" — it's a week where two of your biggest slots are likely to be defined by a low-scoring draw. If you own their defenders, it's quietly fine.

And every so often you get Man City at home to Ipswich: zero out of 100 at both ends, 3.27 expected goals, a 69% clean sheet. Asymmetry analysis doesn't find those — anyone can find those. Its job is to find the Chelsea-at-Sunderland-shaped edges hiding in fixtures the FDR has painted a neutral grey.

The information timeline

Every blank and double has a lifecycle. The question isn't whether you respond to it — everyone does eventually — but where on this line you respond:

 CUP DRAW          PROBABILISTIC        PRE-EMPTIVE          OFFICIAL
 (months out)      MODELLING            SQUAD STRUCTURE      CONFIRMATION
     │                  │                     │                   │
     ●──────────────────●─────────────────────●───────────────────●────▶ time
     │                                        │                   │
     │◀──────────── THE ELITE ACT HERE ──────▶│                   │
     │                                                            │
     │                                      CASUALS BUY HERE ────▶│
     │                                      (full price, plus -4s)

Everything to the left of the official announcement is probabilistic. Everything to the right of it is priced in. The elite are perfectly happy to act on a 60% probability if it means acting four weeks early; the casual manager waits for 100% certainty and pays for it in points and transfers.

Modelling the cup trees

Blanks and doubles aren't acts of God. They're the mechanical consequence of domestic cup fixtures (and occasionally European ones) colliding with league fixtures, which means they can be forecast — not with certainty, but with a probability tree, the same way you'd price any conditional bet.

The elite map the EFL Cup and FA Cup knockout rounds months ahead, as conditional paths. If Team A beats Team B in the fifth round, A's league fixture in the corresponding gameweek has to move — so, for example, Gameweek 29 becomes a blank for both clubs in that league fixture, and it gets re-homed in a later midweek, most likely feeding a double in Gameweek 34 or 37. Every tie that could produce a clash gets a branch. Every branch gets a probability. Multiply through and you have a live estimate of who's going to blank and double, updated every time a ball is kicked in the cup.

 FA CUP R5: TEAM A v TEAM B
 │
 ├── Team A win (p = 0.62)
 │     └── A's GW29 league game clashes with the QF ──▶ GW29 BLANK
 │           └── Rearranged into a midweek ──▶ DOUBLE in GW34 or GW37
 │
 └── Team B win (p = 0.38)
       └── B's league game clashes instead ──▶ the blank moves to B

(The probabilities above are illustrative. In practice you'd lift them straight from the cup-tie match odds, which is precisely why the betting markets are the first place the elite look.)

At the time of writing, our own fixture matrix has zero blank or double cells flagged for 2026/27. That's exactly how it should look in September — and exactly the point. The information doesn't exist yet in any official form. The managers who'll profit from it most are the ones who'll have started positioning before it does.

Pre-emptive accumulation

The elite don't scramble for -8s the week a double is announced. They quietly spend their rolling free transfers three to five weeks in advance, targeting a very particular kind of player: someone worth owning right now on the immediate fixtures, who also carries a high probability of an eventual double.

That dual requirement is the whole trick. Buying a player purely for a double that's six weeks away means carrying dead weight in the meantime (see section 3 on why the future is worth less than you think). Buying a player purely for this week means you'll have to sell him again later. The player who satisfies both is the one who lets you arrive at the double with your squad already built, your transfers already banked, and nothing to do on announcement day except watch everyone else's price-rise notifications.

Chip architecture: the Wildcard into Bench Boost

The crown jewel of FPL logistics is the structural stack. Rather than firing the Wildcard in a fit of pique after a bad gameweek, the elite hold it to build a bespoke fifteen-man squad of nailed-on double-gameweek starters — then trigger the Bench Boost the following week, so all fifteen get to exploit a double that might contain 30 fixtures' worth of returns, without leaving a single structural problem behind for the weeks afterwards.

Compare that with the ordinary Bench Boost, played on a squad whose bench is a £4.0m goalkeeper who's never started, a defender who's lost his place and a striker who's injured. Same chip. A fraction of the value.

What the elite aren't is sentimental about chips. Our capture of the current top 1,000 managers' squads makes that plain:

71.6% of them fired a Bench Boost in Gameweek 1. That isn't recklessness. A fresh Gameweek 1 squad — fifteen players picked on day one, every one fit, every one expected to start — is its own kind of structure, and the elite recognised it and cashed it. The principle is identical to the Wildcard-into-Bench-Boost stack: a chip is worth exactly what the structure around it makes it worth, and when the structure appears, you play it. What you don't do is hoard a chip because it feels important, or burn one because the week's gone badly and you'd like to feel something.

"Doesn't decay apply to everyone equally?"

It's the most common objection, and it sounds right: if every player's future is discounted by the same γ, surely the rankings don't change? For a one-off comparison of two players with identically shaped fixture runs, they don't. But you aren't making one-off comparisons. You're running a portfolio with a limited, irreversible supply of transfers, and in that setting decay changes the optimal sequence of moves completely — because it punishes players whose points arrive late.

Here are two players with identical five-week totals and mirror-image fixture runs:

Without decay, your spreadsheet (or a lazy solver) thinks these two players are interchangeable: 23.0 apiece. With decay, Player A is ahead by nearly three projected points, 18.53 to 15.67.

Why should that be? Because points delivered this weekend come with almost total information certainty: you've seen the team news, the press conference, the likely XI. Points projected for Gameweek 5 carry enormous real-world entropy — a hamstring in training, a tactical reshuffle, a fifth yellow card, a Europa League trip somewhere with a four-hour flight. Nearly every one of those can only take points away from a projection; almost nothing adds them back. The discount factor is simply the honest price of that uncertainty.

How much γ matters

The gap isn't a quirk of 0.85. For these two players it's zero only when γ = 1 — that is, only if you believe a projection for five weeks' time is as reliable as one for this Saturday, which nobody who's ever owned a Premier League full-back believes.

Top managers rarely buy a player today purely for a fixture he has in a month. They weight immediate fixtures heavily and keep enough flexibility to pivot before the back end of a run arrives — because by the time it does, the thing they were planning for has usually changed anyway.

What the solver is actually doing

"Mixed-integer linear programming" sounds like something you'd need a lanyard to discuss, but the structure is simple. Every player in every gameweek becomes a yes/no decision — own him or don't, start him or don't, captain him or don't — and the solver searches every legal combination for the one that maximises the discounted objective above, subject to the rules of the game:

• →Budget: the squad has to fit inside your bank plus your selling prices — not current prices, which matters more than people think once team value starts moving.
• →Squad rules: two goalkeepers, five defenders, five midfielders, three forwards, and no more than three players from any one club.
• →Formation: every starting XI has to be legal — one goalkeeper, at least three defenders, at least two midfielders and at least one forward.
• →Transfer economy: each gameweek's free transfers, how many roll forward, and a −4 for every move beyond them.
• →The objective: the discounted sum of projected points across the horizon — so a move that pays off this week beats an identically sized one that pays off in five.

What the option actually buys you

"Flexibility" is an easy word to wave around, so here's what 1.5–2.2 points of it actually consists of:

• →Structural power. Going from one free transfer to two lets you do a mini-overhaul — downgrade a defender to upgrade a striker, or shuffle a double-up out of a team whose fixtures have turned — without paying a single point in hits. One transfer can only swap like for like within your budget; two can move money around the squad.
• →Emergency cover. It keeps you able to react instantly to the catastrophic Friday press conference — the "he'll be out for a few weeks" that lands eighteen hours before the deadline — without a hit, and without fielding ten men.
• →Information timing. A rolled transfer can be spent next week, with a week's more team news, another fixture's worth of form data and a fresh set of odds. A spent transfer is locked in at today's information level for ever.

The +0.8 trap, in real numbers

Here's where most transfers go to die. Your proposed single move nets an expected gain of +0.8 xPts over the next three weeks compared with the player you'd be selling. It feels like progress — the number's positive, after all. But you've traded an option worth around 1.8 points for an upgrade worth 0.8. That's a net loss of about a point of expected value, and you've also lost the ability to do anything about next Friday's injury news.

It isn't a hypothetical. Ahead of GW6, our model projects Mbeumo (£7.9m) for 3.68, 2.84 and 3.19 over the next three gameweeks, against Morgan Rogers' (£7.7m) 3.25, 2.69 and 3.16. Rogers was in 67.1% of top-1,000 squads at GW4; Mbeumo is exactly the kind of "he's flying" punt that fills the Tuesday-night transfer threads. The swap is worth +0.61 over three weeks — and costs £0.2m and a free transfer to make.

The decision tree

 SHOULD I MAKE THIS TRANSFER?
 │
 ├── Is a starter injured, suspended or dropped?
 │     ├── YES ──▶ Replace him. The option exists for exactly this.
 │     └── NO
 │          │
 │          ├── Is ΔxP over the next 3 GWs comfortably above ~2?
 │          │     ├── YES ──▶ Make it (but at the deadline — see section 6)
 │          │     └── NO
 │          │          │
 │          │          ├── Am I already at the maximum banked?
 │          │          │     ├── YES ──▶ Use it on the best move going:
 │          │          │     │           an FT you can't bank is worth zero
 │          │          │     └── NO  ──▶ ROLL IT

The asymmetric downside of the fun punt

If an elite asset like Erling Haaland is started by 90% of managers in your rank tier and captained by 70%, his EO is 160%. Plug that into the rank-swing formula and every point he scores costs non-owners 1.6 points of ground, costs uncaptained owners 0.6, and gains the captainers a modest 0.4.

Now listen to the casual manager in the Tuesday group chat: "Everyone's captaining Haaland at home to a promoted side, so I'm going to captain a 5%-owned winger to be different and fly up the ranks!" This is mathematically reckless. The winger might well score. But the fixture that made Haaland 70% captained hasn't gone anywhere, and the downside is asymmetric: if Haaland hauls, you bleed ground at 1.6 times his points; if the winger hauls, you gain at roughly twice his points — from a much lower probability of it happening. Here's what a Haaland hat-trick (call it 17 points) does to each camp:

 HAALAND SCORES A HAT-TRICK (17 pts) AT 160% EO
 │
 ├── Not owned ............. 17 × (0 − 1.6) = −27.2   ▼▼▼ rank destruction
 ├── Owned, not captained .. 17 × (1 − 1.6) = −10.2   ▼   still bleeding
 ├── Captained ............. 17 × (2 − 1.6) =  +6.8   ═   shield held
 │
 └── The real rank gains? Not here. They're won in your other
     7–8 slots, where ownership is spread thin.

Look at the captain line. Even the manager who did the "right" thing only gains 6.8 on the field — near enough neutral. That's the point. Elite managers treat captaining a 150%+ EO asset as an insurance policy: a risk-management tool that flattens their variance against the field. They know rank isn't gained by gambling against overwhelming baseline projections. It's accumulated systematically through the seven or eight squad slots where ownership is diffuse — where a 12-point return moves you forward almost one-for-one, because barely anyone in your tier has him.

What the top 1,000 actually owned

Our elite capture pulls every squad in the top 1,000 of the overall league after each deadline. The GW4 numbers show both halves of the argument at once:

Three things jump out. First, the shield. João Pedro was started by 99.7% of the top 1,000 and carried a 145.2% EO, so every point he scored cost anyone in that tier who didn't own him 1.45 points of ground — and even owners who didn't captain him lost 0.45 a point. At that level, not owning him wasn't a differential. It was a donation.

Second, the elite's captaincy consensus formed around the matchup, not the name. Haaland was in 75.5% of top-1,000 squads but wore the armband in just 8.7% of them, for an EO of 84.3%. Meanwhile 85.3% of the top 1,000 put the armband on one of two Chelsea players at home to Hull — Palmer (43.8%, including 10.5% on a Triple Captain) or João Pedro (41.5%). That's section 1's asymmetry thinking and section 5's EO thinking arriving at the same place from opposite directions.

Third, "differential" is a relative word. Pascal Groß was in 19.9% of squads across the whole game — a differential by any normal definition — and in 73.5% of top-1,000 squads. If you're chasing rank at the very top, Groß wasn't a differential at all; he was the template. The only ownership that matters is the ownership of the people you're actually trying to beat.

For reference, here's how the model rates those three captaincy names now, heading into GW6:

When to harvest variance, and when to dampen it

The elite don't have a single risk setting. They tune it to their competitive position, the way a fund manager behaves differently protecting a lead than chasing a benchmark — and the trigger for switching is arithmetic, not mood:

 WHAT'S MY SITUATION?
 │
 ├── Leading, or protecting a high overall rank
 │     └── DAMPEN VARIANCE
 │           • Match the template's core structure
 │           • Captain the dominant-EO asset
 │           • Nailed 90-minute starters, steady baseline BPS
 │
 ├── Behind, but with plenty of gameweeks left
 │     └── STAY OPTIMAL: good process closes gaps on its own
 │
 └── Behind, and running out of gameweeks
       └── HARVEST VARIANCE
             • Fat-tailed assets: explosive per-90s, 18-point ceilings
             • Accept shaky minutes as the price of the ceiling
             • Diverge from the leader's picks, not from good picks

Variance dampening is for when you're ahead, or high in the overall rankings: maximise the squad's floor. Match the template's structure, captain the dominant-EO asset, and lean on secure 90-minute starters who deliver consistent baseline Bonus Point System output. You're not trying to win the week. You're trying to make sure nobody overtakes you in it.

Variance harvesting is the opposite, and it has a trigger. Once you're 60 points behind with six gameweeks left, playing the "optimal" template isn't safe — it's a guarantee of second place, because the leader already owns those players and every point they score is a point you both get. Top managers in that spot deliberately pivot to high-variance, fat-tailed assets: wingers with explosive per-90 numbers and dubious minutes security, but genuine 18-point ceilings. It lowers their expected score slightly. It raises the probability of the one outcome that matters — finishing first — considerably.

The two strategies, side by side

Pricing the wait

You can put rough numbers on this. Suppose there's a 10% chance that between Tuesday night and Friday's press conferences, the player you're buying picks up a knock, gets rotated after a cup tie, or is quietly dropped. If that happens you're either fielding a blank or spending a −4 to fix it — call the damage eight points either way.

Is £0.1m worth 0.8 points? Occasionally — when it's the exact £0.1m that would block a planned move three weeks later. Most weeks it's worth nothing at all, because you never actually spend it. Team value is a paper currency. It only becomes real when you cash it in, and the best thing to cash it in for is certainty: confirmed press-conference news, European minutes loads and tactical leaks. Spending team value to buy that certainty is one of the most profitable trades an FPL manager can make.

If you do want to play the price market, play it deliberately. Our Price Change Predictions page tracks each player's progress towards a rise or fall, so you can see whether a move genuinely has to happen tonight — or whether that's just the transfer threads talking.

When to pull the trigger

 WHEN SHOULD I MAKE THIS TRANSFER?
 │
 ├── Is the player I'm selling already ruled out?
 │     └── Usually still wait: a £0.1m drop is cheaper than
 │         picking his replacement before the team news is in
 │
 ├── Is tonight's price rise near-certain, AND would that £0.1m
 │   block a planned move within the next 2–3 gameweeks?
 │     ├── YES ──▶ Move early. A deliberate trade, not a reflex.
 │     └── NO  ──▶ Wait.
 │
 └── Default ──▶ The final 24 hours, after the press conferences

The binary nature of clean sheets

Attacking returns are individual and close to continuous. A goal, an assist, a bonus point: each arrives through a separate chain of events, and even two attackers from the same side only loosely move together — the winger can blank while the striker scores twice.

Clean sheets are neither individual nor continuous. They're a single, binary event shared by everyone in the same backline: either the team concedes or it doesn't. When it keeps one, every goalkeeper and defender you own from that side collects four points. When a 93rd-minute deflection goes in, every one of them loses the same four points in the same instant. There's no partial credit, and no diversification inside a single defence. In portfolio terms, the correlation between two same-team clean sheets is exactly 1.

The negative case earns a sentence of its own. Owning the defenders of both teams in the same fixture is the one structure where clean sheets hedge each other: only a 0-0 pays both, and any goal pays at most one side. It's rarely what you want, but it's a useful reminder that correlation, not quality, is what shapes a defensive portfolio.

The wipeout, in real numbers

Take GW6. The bookmakers give Arsenal a 67% clean-sheet chance at home to Leeds, and Everton 54% away at Hull — the same odds-implied numbers as section 1's table. Here are two ways of filling two premium defensive slots, counting clean-sheet points only:

The stack wins on raw expectation by about half a point, and in a vacuum that settles it. But look at the right-hand columns. One week in three, the stacked structure returns nothing from both slots on the same afternoon. The split cuts that to roughly one week in seven, and trims the week-to-week standard deviation by more than a quarter. You're paying 0.52 expected points for a much flatter distribution.

The stacked downside is also worse than zero. Goalkeepers and defenders lose a point for every two goals conceded, so a 3-1 defeat doesn't just wipe out both clean sheets — it docks both players at once. Correlation works on the negative points as well as the positive ones.

And that half-point gap is before counting the part of defending that isn't binary. Since 2025/26, defenders have earned two points for reaching 10 clearances, blocks, interceptions and tackles in a match: a return that's individual, close to continuous, and independent of whether the keeper gets beaten. Tarkowski is averaging 10.33 of those actions per 90 this season, which makes his defensive-contribution bonus something close to a weekly coin flip that doesn't care about Hull's finishing. Gabriel averages 7.0; Raya, as a goalkeeper, can't earn it at all. It doesn't erase the gap — across GW6–8 our model, which counts every scoring route, still has Raya ahead of Tarkowski by 7.67 to 5.87, about 0.6 a week — but it changes what you're buying. A real slice of Tarkowski's points arrives whether or not Hull score. The best defensive portfolios are built from assets whose points come from different places, and here the price of that decoupling is roughly 0.6 points a week: worth paying when you're protecting a rank, not when you're chasing one.

Floor vs ceiling: when to stack and when to split

None of this makes the double-up wrong. It makes it a variance decision, and section 5 has already told you when to buy variance. A stacked backline is a bimodal bet: nothing or 12-plus points from two slots, with very little in between. That's exactly the shape you want when you're chasing and need a fat tail. It's exactly the shape you don't want when you're protecting a lead or a high overall rank, where one deflection can erase eight to twelve points of your defensive foundation at once.

 SHOULD I DOUBLE UP ON ONE DEFENCE?
 │
 ├── Protecting a lead or a high overall rank?
 │     └── SPLIT: premium defenders across two elite backlines
 │           (one Arsenal, one Liverpool or Everton — not two Arsenal)
 │
 ├── Behind, and running out of gameweeks?
 │     └── STACK: the bimodal tail is the point
 │
 └── Neither — just optimising?
       │
       ├── Does the second asset have a non-clean-sheet route to points?
       │   (set-piece goals, attacking full-back, DefCon volume)
       │     ├── YES ──▶ Partly decoupled already: the double-up is fine
       │     └── NO  ──▶ Split, unless the clean-sheet odds gap is large

That last branch is why Gabriel is the Arsenal defender elite managers double up on most comfortably. A meaningful slice of his projection comes from set-piece goals, which arrive independently of the clean sheet — he's partly an attacking asset wearing a defender's price tag. A second defender whose entire case is the clean sheet adds almost pure correlation.

The same logic applies to the goalkeeper-plus-defender stack, with one small natural hedge built in: save points rise when a keeper is busy, and busy keepers concede more often. It isn't enough to decouple the pair, but it's why the elite, when they do stack a defence, prefer keeper-plus-defender to two outfield defenders with nothing but the clean sheet between them.

The dead-wood boundary

The floor isn't a constant. It moves every season with pricing and with the squads of the promoted clubs, so measure it rather than assume it. Here's what the cheapest end of the market looks like in our model right now:

Look at the midfield row. The line isn't gradual. Between £5.0m and £5.5m the three-gameweek projection roughly doubles, from under 3.0 to around 6.0, because below it you're buying rotation pieces and holding midfielders who don't shoot or take set pieces. That cliff — typically somewhere around £5.0–5.3m for midfielders and forwards, lower for defenders — is the dead-wood boundary. Once the APR on your last five or six slots falls beneath it, you're no longer choosing between players. You're choosing which slots to abandon.

Defence is the exception worth exploiting. This season's promoted sides have produced genuine £4.0–4.1m starters, which makes them some of the most valuable assets in the game per pound. Find them early and the APR problem gets dramatically easier everywhere else — although the three-per-club rule and the handful of them that exist mean you can't build a whole back line out of them.

Two builds, one budget

Here's the audit applied to two versions of the same core, at GW6 prices. Both start with Haaland, Bruno Fernandes, Palmer, Rogers, Raya and a £4.0m backup keeper. Build A then adds Mbeumo too. Build B stops there and spreads the difference across the squad.

(The defensive floor is set at £4.5m, not £4.0m, because you can't fill five defensive slots with the few £4.0m starters available.)

Build A's APR is only £0.36m lower than B's, which looks harmless. The slack tells the truth: A is £2.0m short of fielding a playing footballer in every remaining slot, so at least two of those eight shirts will be worn by people who don't play. In practice that tends to mean a non-starting £4.0m defender, a £4.5m midfielder who's one injury away from being your eleventh man, and a second forward slot with no second forward in it. In return, A gains about 3.7 projected points over three weeks in one slot: Mbeumo's 9.72 against the 6.05 of a £5.5m Wharton, which is how B fills it. What A pays is structural — dead shirts that can't cover an injury or a blank, and can't be sold for anything useful, which is where the next subsection comes in. Sometimes that trade is right. The point of the audit is to make it deliberately, not to discover it in October.

Fluidity: why a balanced squad moves faster

APR matters long after the draft, because it decides how quickly your squad can move money around once the season starts. A transfer can only reach what the outgoing player's selling price plus your bank can buy:

• →A £5.5m starting midfielder can pivot to almost any breakout budget option — the £5.5m set-piece taker just moved into the front three, the £5.8m defender whose side has found a new system — with a single free transfer.
• →A £4.0m non-starter can only reach another £4.0–4.3m player. To move that money anywhere useful you have to pair it with a sale elsewhere: two transfers, and very often a −4, to do what a balanced squad does with one.
• →Carry three or four of those slots and the squad is structurally frozen. Every opportunity needs multi-transfer surgery — exactly the −4 and −8 culture that section 4's option-value argument warns against.
• →The elite rule of thumb: one, at most two, deliberate fodder slots — the backup keeper and the fifth defender. Every other shirt should both start and be sellable for something useful.

Building the matrix

Instead of fixing one target for gameweek N+2, map a price tier of equivalents: players at neighbouring prices whose bilateral fixture deltas (section 1) and roles point the same way. Here's a real one — the £7.4–7.9m midfield tier heading into GW6, ranked by our projections over the next three gameweeks:

If Mbeumo rises to £8.0m overnight and your route was balanced to the last £0.1m, you don't take a hit. You default to Rogers, keep about 94% of the projected value (9.10 against 9.72), and leave the transfer economy intact. Forcing the original move with a −4 to free money elsewhere would cost more than six times the gap it was trying to close.

The matrix runs upwards too. If your outgoing player rises or a target above the tier drops, the same map tells you instantly whether a Tier A+ option has just become affordable — the same speed of decision, in the other direction.

The pivot test

 MY ROUTE JUST BROKE BY £0.1m
 │
 ├── Is there a Tier B with the same fixture delta and role?
 │     ├── YES ──▶ Pivot. A gap under 1 xP beats any −4.
 │     └── NO
 │          │
 │          ├── Can a bench-slot downgrade find £0.1m for under 1 xP?
 │          │     ├── YES ──▶ Make the downgrade, keep Plan A
 │          │     └── NO  ──▶ Delay a week and roll the transfer (section 4)
 │
 └── Never ──▶ Take a −4 purely to repair a £0.1m break

• →Keep £0.1–0.3m in the bank as deliberate price-rise insurance. It's the cheapest contingency in the game.
• →Price any target who's close to a rise at his next price, not his current one. Our Price Change Predictions page shows who's genuinely about to move.
• →Build tiers across different clubs where you can, so a single injury or rotation scare can't wipe out Plan A and Plan B at once.
• →Match role as well as price: a penalty taker substitutes for a penalty taker, a set-piece defender for a set-piece defender.
• →Refresh the matrix after the press conferences. Deadline arbitrage (section 6) and contingency planning are the same habit seen from opposite ends: the matrix is what makes waiting safe.

How the auto-sub actually works

The rules are simple, and one detail changes everything:

• →Auto-subs happen after the gameweek finishes. Any starter who played zero minutes is replaced by the first player in your bench order who did play.
• →A substitute who didn't play is simply skipped, at no cost. The game moves on to the next name.
• →The formation must stay legal. If a blanking starter would take you below three defenders or one forward, the first bench player who keeps it legal comes on instead, whatever the order says.
• →Your goalkeeper can only be replaced by your backup goalkeeper, who sits outside the outfield order entirely.
• →A starter who plays a single minute is not substituted. A one-minute cameo scores one point and blocks the auto-sub completely.

Rank by points if he plays, not points on average

The second rule is the one most managers miss. Because a substitute who doesn't play costs nothing — he's skipped — the only thing that matters about your first sub is how many points he scores if he does get on the pitch. His own minutes risk is already handled by the rules. So the right number to rank by isn't his raw projection. It's his projection conditional on playing:

At one zero-minute starter a week, Sub 1 comes on in roughly five gameweeks out of eight — around two dozen appearances over a season. That isn't a bench slot. It's a part-time twelfth starter, and picking it by shirt number is exactly as sensible as it sounds.

Ceiling first: using the matchup delta

Once you're ranking on points-if-he-plays, section 1's bilateral deltas do the rest. A bench attacker facing a leaky defence with a big attacking delta carries a real ceiling: one goal involvement is worth five to eight points to him. A secure defender facing a top-six attack is a near-certain two points with an unlikely clean sheet on top. On raw projections they can look similar. On a conditional, ceiling-weighted basis they're not close:

The defender never becomes a bad pick — he'll still get his two points when he's needed. He just belongs behind a forward who, if he gets on the pitch, is worth more than a third as much again. Over two dozen auto-subs a season, ordering for ceiling instead of nominal baseline is worth a handful of points you'll never notice winning, which is exactly how most elite edges feel from the inside.

Three refinements. First, the formation rule can overrule you: if you start three at the back, a blanking defender pulls in your first bench defender regardless of order, so the best defender on your bench should be the first defender listed. Second, the cameo trap cuts the other way: a starter at real risk of a ten-minute substitute appearance is worse than one who's ruled out, because his single point blocks the auto-sub. When a nailed bench player projects better than a rotation-risk starter, start the nailed player — the bench only protects you from zero minutes, not from one. Third, section 5 still applies: when two bench options are close, the tie goes to the higher floor when you're protecting a rank and to the higher ceiling when you're chasing one.

The ten rules

• →Stop rating fixtures as a whole. Rate attacks against defences, and defences against attacks.
• →Stop reacting to the calendar. Model the cup draws and position your squad for blanks and doubles before the rest of the game catches on.
• →Remember that time decays certainty. Value points this weekend over theoretical promises next month.
• →Respect the option value of an unspent transfer. Don't spend a precious free transfer on a marginal +0.5 upgrade.
• →Use the captaincy to manage tail risk. Let your low-EO differential slots do the heavy lifting.
• →Value deadline information over Monday-night pennies. Team value exists to be spent on information security.
• →Treat a defensive double-up as a variance decision, not a points decision. Decouple your backline when protecting, stack it when chasing.
• →Audit your Average Price Remaining before you confirm a premium. Negative slack means dead wood, and dead wood freezes the squad.
• →Never build a single-track route. Carry price-tier equivalents so a £0.1m break costs you a fraction of a point, not a −4.
• →Order your bench by points-if-he-plays and matchup delta. Sub 1 is a part-time starter, so pick it like one.

The habit ledger

FPL King

Prediction Engine Team

Written by the team behind FPL King's prediction engine, turning bookmaker odds, xG modelling and squad optimisation into plain-English advice.