FPL Differentials — Low-Ownership Picks by Model xPts

#	Player	Club	Pos	Price	Owned	xPts	Big haul	Bad / good week	Attack	Defence	Involve
1	O.Dango	BRE	MID	£6.5m	1.4%	3.63	9%	1 – 9	1.45	0.35	1.83
2	Thiawnew mgr	NEW	DEF	£5.0m	2.0%	3.62	7%	1 – 8	0.72	1.04	1.85
3	Tarkowski	EVE	DEF	£6.0m	9.7%	3.60	6%	1 – 8	0.51	1.27	1.82
4	Collins	BRE	DEF	£5.5m	2.1%	3.58	5%	1 – 8	0.49	1.23	1.87
5	Enzonew mgr	CHE	MID	£7.0m	5.5%	3.56	10%	1 – 9	1.41	0.38	1.78
6	Richardsnew mgr	CRY	DEF	£5.0m	0.8%	3.56	6%	1 – 8	0.64	1.20	1.72
7	Sarrnew mgr	CRY	MID	£6.5m	7.9%	3.51	9%	0 – 9	1.64	0.24	1.64
8	Vuskovicnew club	BHA	DEF	£5.0m	2.7%	3.41	6%	0 – 8	0.70	1.00	1.72
9	Canvotnew mgr	CRY	DEF	£5.0m	0.6%	3.41	6%	1 – 8	0.40	1.14	1.88
10	Cherki50%France 3rd-place PO 18 Jul; back ~10 Aug, City tip Community Shield/opener return	MCI	MID	£7.5m	8.9%	3.33	8%	1 – 8	1.32	0.27	1.74
11	Dewsbury-Hall	EVE	MID	£6.5m	3.7%	3.27	8%	1 – 8	0.94	0.46	1.87
12	Kelleher	BRE	GKP	£5.0m	5.7%	3.26	4%	1 – 8	0.01	1.14	2.04
13	Whartonnew mgr	CRY	MID	£5.5m	0.4%	3.25	5%	1 – 7	0.81	0.68	1.77
14	Stach	LEE	MID	£6.0m	1.1%	3.20	6%	1 – 7	0.74	0.61	1.85
15	Ballard	SUN	DEF	£5.0m	4.4%	3.20	5%	1 – 8	0.42	1.02	1.76

Attack = goals + assists points · Defence = clean sheets, goals conceded, saves and defensive-contribution points · Involve = appearance, bonus and card points. The three columns sum to xPts — check the arithmetic any time.
Big haul is the chance he scores 10 or more — the column that matters most here, because a differential only earns its place by having a week nobody else has. Bad / good week is the score he lands on when it goes badly and when it goes well.

Be honest with yourself about what this table is: differentials are the volatile end of the game. A low-owned player is usually low-owned for a reason — rotation risk, a new signing, a small club — and even the model's best differential call blanks more often than a template pick. The table exists so you pick your risks on projection instead of on last week's highlights, not so you can avoid risk. The full every-player table lives on predicted points.

#	Player	Team · pos	Owned	EO*	xPts	Rank upside
1	O.Dango	BRE · MID	1.4%	1.4%	3.63	3.58
2	Thiaw	NEW · DEF	2.0%	2.0%	3.62	3.55
3	Richards	CRY · DEF	0.8%	0.8%	3.56	3.53
4	Collins	BRE · DEF	2.1%	2.1%	3.58	3.50
5	Canvot	CRY · DEF	0.6%	0.6%	3.41	3.39
6	Enzo	CHE · MID	5.5%	5.5%	3.56	3.36
7	Vuskovic	BHA · DEF	2.7%	2.7%	3.41	3.32
8	Tarkowski	EVE · DEF	9.7%	9.7%	3.60	3.25
9	Wharton	CRY · MID	0.4%	0.4%	3.25	3.24
10	Sarr	CRY · MID	7.9%	7.9%	3.51	3.23
11	Stach	LEE · MID	1.1%	1.1%	3.20	3.16
12	Dewsbury-Hall	EVE · MID	3.7%	3.7%	3.27	3.15
13	Kelleher	BRE · GKP	5.7%	5.7%	3.26	3.07
14	Ballard	SUN · DEF	4.4%	4.4%	3.20	3.06
15	Cherki	MCI · MID	8.9%	8.9%	3.33	3.03

*EO is overall ownership until GW1 locks; the elite version (owned + captaincy among the top managers) activates automatically once the season is live, and re-sorts this table by the field you're actually racing.

The two multiply into a differential value read: a player is only interesting when the ownership is low and the projected return is high. A cheap, unowned player the model rates poorly is not a differential — he is a trap; a low-owned player with a soft run and strong xPts is the one worth catching. That forward xPts column comes from the same table on predicted points, and pairs with the model's fixture difficulty so you can see which low-owned picks have the easiest run. For the captaincy version of the same call — when a differential is worth the armband — Bawler's weekly captain shortlist publishes on captain picks.

Across the 2025/26 walk-forward backtest, every time the model projected 4.5+ points for a player under 10% owned — 121 calls in total — those picks averaged 5.53 points. That is more than the 4.4 averaged the same weeks by the template players owned by 30%+ of the field, and 28.1% of them hauled (8+ points). For scale: pick the same number of under-10%-owned players by recent form instead and the haul rate is about 10% — low-owned players as a group haul roughly one week in ten. The model's 28.1% is the lift over that, not over zero. In plain terms: when the projection said a low-owned player was worth backing, backing him beat holding the crowd's picks — on average, across a full season, misses included.

The best calls the crowd missed

A sample of the backtest's biggest hits — weeks where almost nobody owned the player, the model projected him anyway, and he delivered:

Read it honestly: this table is the wins, cherry-picked by construction. The averages and haul-rate above include every call — blanks and all — and that is the number to judge the model on. The full backtest (accuracy vs form and points-per-game, calibration, captaincy) is published on predicted points.

Read the extremes: Gibbs-White was the fifth-highest scorer in the entire game (188 points) while sitting in just 9.2% of squads — a season-long differential hiding in plain sight. Casemiro returned 165 points from only 3.8% ownership, the best haul of any sub-5% player. And E.Le Fée — owned by 1.6% at £4.8m — banked 147 points at 30.6 points per million, one of the best value returns in the game from a player almost nobody had. By contrast, the ten highest scorers overall averaged 32.4% ownership.

Caveat: the ownership figures above are season-final. A player who was a differential in August can be widely owned by May (or vice versa), so this table is a retrospective study of who finished low-owned and high-scoring — not a claim about when they were catchable. The ranked table at the top of this page solves exactly that: it projects forward from current ownership, week by week.