Sideline Quant
Loading…
2026/27 · live record · Sideline's Team

Track record

Before every deadline I publish the model’s predictions and enter its team in FPL. Once the GW settles, both get marked here, bad weeks included. Nothing is edited after the fact.

So far: ahead of the average manager in 8 of 9 gameweeks (+108 points in total). A baseline beat the model on 18 measures, in GW2, GW4, GW5, GW6, GW8 and GW9 (each report has the detail). Only 9 gameweeks in, so a pinch of salt.

Last settled GW9 · scored against the run of 31 Oct, 10:09 UTC · pre-registered · checkpoints at GW10, GW19, GW38

Points vs the average manager

+108

616 to the average manager’s 508 · above the average in 8 of 9 gameweeks · Triple Captain in GW1

GW1: 59 pts vs average 50 (+9) · TC+9GW1GW2: 91 pts vs average 81 (+10)+10GW2GW3: 56 pts vs average 51 (+5)+5GW3GW4: 83 pts vs average 69 (+14)+14GW4GW5: 30 pts vs average 48 (−18)−18GW5GW6: 79 pts vs average 58 (+21) · WC+21GW6GW7: 86 pts vs average 59 (+27)+27GW7GW8: 80 pts vs average 45 (+35)+35GW8GW9: 52 pts vs average 47 (+5)+5GW9

Against a sample of top-10k managers: −16.9 in GW2, −1.3 in GW3, −6.7 in GW4, −23.6 in GW5, +1.8 in GW6, +10.6 in GW7, +28.6 in GW8, −14.7 in GW9 (before hits · no sample for GW1). This early, the sample is chosen on the very weeks it is scored on, so a deficit is overstated.

Ordering the players

0.24vs0.12
Model vs last season's points per game · 1.00 = perfect order, 0 = none · mean of 9 GWs on the same 1,302 players

Spotting 6+ hauls · season

66%vs58%
The model put a player who went on to haul above one who didn't 66% of the time; last season's points per game managed 58% · 288 hauls on 1,207 shared players

Overall rank

256k
255,729 · after 9 gameweeks, and a single team's rank jumps around a lot this early

The model's own team, week by week

A real FPL entry, picked by the model every deadline. What it predicted against what it scored, with the average manager and a top-10k sample alongside.

A note on GW5. My daughter was born that week, so I didn’t update the team and the GW4 squad carried over. The model’s GW5 predictions were still published and are scored below as normal.

Model said · scored · the field

GW1TC
GW1: average manager 50GW1: model said 65.7 for the squad as submittedGW1: scored 59
+9 vs average · −6.7 vs model
GW2
GW2: average manager 81GW2: top-10k sample 107.9 (60 managers, before hits)GW2: model said 59.2 for the squad as submittedGW2: scored 91
+10 vs average · −16.9 vs top 10k · +31.8 vs model
GW3
GW3: average manager 51GW3: top-10k sample 57.3 (170 managers, before hits)GW3: model said 59.8 for the squad as submittedGW3: scored 56
+5 vs average · −1.3 vs top 10k · −3.8 vs model
GW4
GW4: average manager 69GW4: top-10k sample 89.7 (289 managers, before hits)GW4: model said 63.0 for the squad as submittedGW4: scored 83
+14 vs average · −6.7 vs top 10k · +20.0 vs model
GW5
GW5: average manager 48GW5: top-10k sample 53.6 (329 managers, before hits)GW5: model said 48.4 for the squad as submittedGW5: scored 30
−18 vs average · −23.6 vs top 10k · −18.4 vs model
GW6WC
GW6: average manager 58GW6: top-10k sample 77.2 (527 managers, before hits)GW6: model said 56.9 for the squad as submittedGW6: scored 79
+21 vs average · +1.8 vs top 10k · +22.1 vs model
GW7
GW7: average manager 59GW7: top-10k sample 75.4 (554 managers, before hits)GW7: model said 57.8 for the squad as submittedGW7: scored 86
+27 vs average · +10.6 vs top 10k · +28.2 vs model
GW8
GW8: average manager 45GW8: top-10k sample 51.4 (556 managers, before hits)GW8: model said 57.1 for the squad as submittedGW8: scored 80
+35 vs average · +28.6 vs top 10k · +22.9 vs model
GW9
GW9: average manager 47GW9: top-10k sample 66.7 (644 managers, before hits)GW9: model said 51.3 for the squad as submittedGW9: scored 52
+5 vs average · −14.7 vs top 10k · +0.7 vs model
ScoredModel’s prediction for the submitted squadAverage managerTop-10k sample, before hitsPoints per gameweek. Every mark is a row in the ledger.
Total points
616after hits
Team value
£99.7m+ £0.9m bank
Chips played
2TC GW1 · WC GW6
Every settled gameweek of the model’s real FPL team. Select a gameweek to see the team.
GWPtsvs avgvs 10ktotalrankpredictedchip
59+9—591.99m65.7TC
91+10−16.91501.98m59.2—
56+5−1.32061.68m59.8—
83+14−6.72891.33m63.0—
30−18−23.63192.75m48.4—
79+21+1.83981.69m56.9WC
86+27+10.6484857k57.8—
80+35+28.6564161k57.1—
52+5−14.7616256k51.3—

Select a gameweek to see the team below. A dash means the figure could not be computed for that week; it is never a zero.

The team in GW952 ptsShow the team
Tzolakis
1pts
MinutesMin90
BonusBns0
GKHUL
Gabriel
6pts
MinutesMin90
BonusBns0
DEFARS
Van Hecke
15pts
MinutesMin90
BonusBns3
DEFTOT
Virgil
1pts
MinutesMin90
BonusBns0
DEFLIV
Ajayi
2pts
MinutesMin90
BonusBns0
DEFHUL
Wirtz
2pts
MinutesMin80
BonusBns0
MIDLIV
V
Tavernier
3pts
MinutesMin90
BonusBns0
MIDBOU
E.Le Fée
8pts
MinutesMin90
BonusBns1
MIDSUN
Mbeumo
1pts
MinutesMin90
BonusBns0
MIDMUN
C
Haaland
2pts
MinutesMin90
BonusBns0
FWDMCI
Barry
9pts
MinutesMin79
BonusBns3
FWDEVE
benchΣ 8pts
1
Raya
5pts
MinutesMin90
BonusBns0
GKARS
2
Castagne
2pts
MinutesMin86
BonusBns0
DEFFUL
3
Hughes
1pts
MinutesMin2
BonusBns0
MIDCRY
4
João Pedro
0pts
MinutesMin0
BonusBns0
FWDCHE
Transfers
OUTCalafioriARSINVan HeckeTOT

How accurate were the predictions?

Every player the model expected to start, picked or not, against what they scored. Points per player first, then the order and the hauls, each week beside simple baselines.

Prediction accuracy per settled gameweek, over likely starters, with the season to date.
GWPoints per playerMAEBig missesRMSEOrderingrank ρPredicted → scoredΣ xP → Σ ptsPlayersnSpotting 6+AUCSpotting 10+AUC
GW1seed2.753.350.290.09–0.45332 → 30295——
GW22.373.170.360.23–0.48680 → 66620367%†57–7572%†57–85
GW32.302.990.160.03–0.28681 → 68519969%†60–7764%†46–80
GW42.643.440.310.18–0.43686 → 71520570%62–7858%†41–76
GW52.533.460.11-0.03–0.26636 → 72318858%50–6650%†33–67
GW62.243.010.260.12–0.38644 → 64019560%†50–7063%†42–83
GW72.332.990.300.16–0.43663 → 64619972%†64–8079%†68–89
GW82.513.150.14-0.01–0.28619 → 62718269%†61–7765%†52–78
GW92.142.700.270.13–0.40625 → 63019171%†60–8063%†41–84
SeasonGW1–GW92.403.14—5565 → 56341657——

† fewer than 50 hauls that week - too few to claim a result (GW2 6+: 43 · GW2 10+: 13 · GW3 6+: 42 · GW3 10+: 12 · GW4 10+: 14 · GW5 10+: 12 · GW6 6+: 38 · GW6 10+: 9 · GW7 6+: 43 · GW7 10+: 10 · GW8 6+: 44 · GW8 10+: 11 · GW9 6+: 43 · GW9 10+: 6).

Ordering and the haul figures are per gameweek and are not averaged; the season comparison on the same players is in the tiles above. Likely starters only: players the model expected to play 60+ minutes (predicted E[minutes] > 60 (n=1657 of 5788 player-fixtures)).

95% percentile intervals from a row-level bootstrap (1000 resamples, seeded). Player-fixtures in the same GW share fixtures and clean sheets, which makes these intervals understate the true width - read them as a floor on the uncertainty. Fewer than 50 hauls is too few to claim a result from a haul AUC.

What went wrong each week, and how the baselines did

GW96 misses · beaten by last season's points per game on spotting 10+ hauls; fpl's own projection (ep_next) on spotting 10+ haulsOpen
  • Last season's points per game was better than the model at spotting 10+ hauls this week: on the 150 players both scored, it ranked a hauler above a non-hauler 59% of the time, the model 56%.
  • FPL's own projection (ep_next) was better than the model at spotting 10+ hauls this week: on the 191 players both scored, it ranked a hauler above a non-hauler 72% of the time, the model 63%.
  • Spotting 10+ hauls was no better than a coin toss this week. It came out at 63% and the interval runs from 41% to 84%, and takes in 50%.
  • The team finished 14.7 behind a sample of top-10k managers (average 66.7, 644 sampled). This early, the sample is chosen on the very weeks it is scored on, so the gap is overstated.
  • 8 points sat on the bench and did not count (no Bench Boost).
  • Captain Haaland scored 2, doubled to 4.
  • Better than usual this week - the model was out by 2.14 points a player, against 2.40 across the season so far.
  • The model was closer than last season's points per game on points per player: 2.16 to 2.22 on the 150 players both scored.
  • The model was closer than last season's points per game on the big misses: 2.75 to 2.95 on the 150 players both scored.
  • The model ordered the players better than last season's points per game: 0.23 to 0.07 on the 150 players both scored.
  • The model was better than last season's points per game at spotting 6+ hauls: 70% to 57% on the 150 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.27 to 0.19 on the 191 players both scored.
  • The model was better than top-10k effective ownership at spotting 6+ hauls: 71% to 62% on the 191 players both scored.
  • The model was better than top-10k effective ownership at spotting 10+ hauls: 63% to 62% on the 191 players both scored.
  • The model ordered the players better than fpl's own projection (ep_next): 0.27 to 0.22 on the 191 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 6+ hauls: 71% to 66% on the 191 players both scored.
  • The team scored 52, 5 above the average manager's 47.
  • Only 43 players scored 6+ this week, too few to claim a result on the 6+ figure.
  • Only 6 players scored 10+ this week, too few to claim a result on the 10+ figure.

Scored against the run of 31 Oct, 10:09 UTC · rows fa6549fe…e54a · full report →

Against the baselines

Last season's points per gamesame 150 players
MeasureModelBaselineVerdict
points per player2.162.22model ahead
big misses2.752.95model ahead
ordering the players0.230.07model ahead
spotting 6+ hauls70%57%model ahead
spotting 10+ hauls56%59%baseline ahead
Top-10k effective ownershipordering only · same 191 players
MeasureModelBaselineVerdict
ordering the players0.270.19model ahead
spotting 6+ hauls71%62%model ahead
spotting 10+ hauls63%62%model ahead
FPL's own projection (ep_next)ordering only · same 191 players
MeasureModelBaselineVerdict
ordering the players0.270.22model ahead
spotting 6+ hauls71%66%model ahead
spotting 10+ hauls63%72%baseline ahead
GW84 misses · beaten by last season's points per game on spotting 10+ haulsOpen
  • Worse than usual this week - the model was out by 2.51 points a player, against 2.40 across the season so far.
  • Its ordering of the players barely matched the week's: rank correlation 0.14, where 1.00 would be a perfect order and 0 no relationship at all.
  • Last season's points per game was better than the model at spotting 10+ hauls this week: on the 140 players both scored, it ranked a hauler above a non-hauler 65% of the time, the model 61%.
  • The best team the model could legally have bought scored 55, 43% of the 128 a perfect-hindsight XI would have made.
  • The model was closer than last season's points per game on points per player: 2.58 to 2.65 on the 140 players both scored.
  • The model was closer than last season's points per game on the big misses: 3.21 to 3.34 on the 140 players both scored.
  • The model ordered the players better than last season's points per game: 0.12 to -0.03 on the 140 players both scored.
  • The model was better than last season's points per game at spotting 6+ hauls: 67% to 53% on the 140 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.14 to 0.11 on the 182 players both scored.
  • The model was better than top-10k effective ownership at spotting 6+ hauls: 69% to 62% on the 182 players both scored.
  • The model was better than top-10k effective ownership at spotting 10+ hauls: 65% to 54% on the 182 players both scored.
  • The model ordered the players better than fpl's own projection (ep_next): 0.14 to 0.04 on the 182 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 6+ hauls: 69% to 55% on the 182 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 10+ hauls: 65% to 62% on the 182 players both scored.
  • The team scored 80, 35 above the average manager's 45.
  • Only 44 players scored 6+ this week, too few to claim a result on the 6+ figure.
  • Only 11 players scored 10+ this week, too few to claim a result on the 10+ figure.

Scored against the run of 23 Oct, 16:44 UTC · rows c928dafd…ce6d · full report →

Against the baselines

Last season's points per gamesame 140 players
MeasureModelBaselineVerdict
points per player2.582.65model ahead
big misses3.213.34model ahead
ordering the players0.12-0.03model ahead
spotting 6+ hauls67%53%model ahead
spotting 10+ hauls61%65%baseline ahead
Top-10k effective ownershipordering only · same 182 players
MeasureModelBaselineVerdict
ordering the players0.140.11model ahead
spotting 6+ hauls69%62%model ahead
spotting 10+ hauls65%54%model ahead
FPL's own projection (ep_next)ordering only · same 182 players
MeasureModelBaselineVerdict
ordering the players0.140.04model ahead
spotting 6+ hauls69%55%model ahead
spotting 10+ hauls65%62%model ahead
GW71 miss · no baseline aheadOpen
  • 14 points sat on the bench and did not count (no Bench Boost).
  • Better than usual this week - the model was out by 2.33 points a player, against 2.40 across the season so far.
  • The model was closer than last season's points per game on points per player: 2.47 to 2.61 on the 148 players both scored.
  • The model was closer than last season's points per game on the big misses: 3.16 to 3.36 on the 148 players both scored.
  • The model ordered the players better than last season's points per game: 0.23 to 0.02 on the 148 players both scored.
  • The model was better than last season's points per game at spotting 6+ hauls: 70% to 52% on the 148 players both scored.
  • The model was better than last season's points per game at spotting 10+ hauls: 75% to 55% on the 148 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.30 to 0.11 on the 199 players both scored.
  • The model was better than top-10k effective ownership at spotting 6+ hauls: 72% to 52% on the 199 players both scored.
  • The model was better than top-10k effective ownership at spotting 10+ hauls: 79% to 53% on the 199 players both scored.
  • The model ordered the players better than fpl's own projection (ep_next): 0.30 to 0.00 on the 199 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 6+ hauls: 72% to 47% on the 199 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 10+ hauls: 79% to 70% on the 199 players both scored.
  • The team scored 86, 27 above the average manager's 59.
  • Only 43 players scored 6+ this week, too few to claim a result on the 6+ figure.
  • Only 10 players scored 10+ this week, too few to claim a result on the 10+ figure.

Scored against the run of 17 Oct, 09:10 UTC · rows ca838128…1cd9 · full report →

Against the baselines

Last season's points per gamesame 148 players
MeasureModelBaselineVerdict
points per player2.472.61model ahead
big misses3.163.36model ahead
ordering the players0.230.02model ahead
spotting 6+ hauls70%52%model ahead
spotting 10+ hauls75%55%model ahead
Top-10k effective ownershipordering only · same 199 players
MeasureModelBaselineVerdict
ordering the players0.300.11model ahead
spotting 6+ hauls72%52%model ahead
spotting 10+ hauls79%53%model ahead
FPL's own projection (ep_next)ordering only · same 199 players
MeasureModelBaselineVerdict
ordering the players0.300.00model ahead
spotting 6+ hauls72%47%model ahead
spotting 10+ hauls79%70%model ahead
GW66 misses · beaten by last season's points per game on spotting 6+ hauls; top-10k effective ownership on spotting 6+ hauls; top-10k effective ownership on spotting 10+ haulsOpen
  • Last season's points per game was better than the model at spotting 6+ hauls this week: on the 144 players both scored, it ranked a hauler above a non-hauler 62% of the time, the model 61%.
  • Top-10k effective ownership was better than the model at spotting 6+ hauls this week: on the 195 players both scored, it ranked a hauler above a non-hauler 64% of the time, the model 60%.
  • Top-10k effective ownership was better than the model at spotting 10+ hauls this week: on the 195 players both scored, it ranked a hauler above a non-hauler 68% of the time, the model 63%.
  • Spotting 6+ hauls was no better than a coin toss this week. It came out at 60% and the interval runs from 50% to 70%, and takes in 50%.
  • Spotting 10+ hauls was no better than a coin toss this week. It came out at 63% and the interval runs from 42% to 83%, and takes in 50%.
  • The best team the model could legally have bought scored 40, 30% of the 132 a perfect-hindsight XI would have made.
  • Better than usual this week - the model was out by 2.24 points a player, against 2.40 across the season so far.
  • The model was closer than last season's points per game on points per player: 2.32 to 2.41 on the 144 players both scored.
  • The model was closer than last season's points per game on the big misses: 3.15 to 3.24 on the 144 players both scored.
  • The model ordered the players better than last season's points per game: 0.25 to 0.18 on the 144 players both scored.
  • The model was better than last season's points per game at spotting 10+ hauls: 69% to 63% on the 144 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.26 to 0.07 on the 195 players both scored.
  • The model ordered the players better than fpl's own projection (ep_next): 0.26 to 0.19 on the 195 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 6+ hauls: 60% to 58% on the 195 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 10+ hauls: 63% to 53% on the 195 players both scored.
  • The team scored 79, 21 above the average manager's 58.
  • Only 38 players scored 6+ this week, too few to claim a result on the 6+ figure.
  • Only 9 players scored 10+ this week, too few to claim a result on the 10+ figure.
  • The transfers for this week could not be checked against FPL's records.

Scored against the run of 10 Oct, 09:21 UTC · rows dd306419…4f3b · full report →

Against the baselines

Last season's points per gamesame 144 players
MeasureModelBaselineVerdict
points per player2.322.41model ahead
big misses3.153.24model ahead
ordering the players0.250.18model ahead
spotting 6+ hauls61%62%baseline ahead
spotting 10+ hauls69%63%model ahead
Top-10k effective ownershipordering only · same 195 players
MeasureModelBaselineVerdict
ordering the players0.260.07model ahead
spotting 6+ hauls60%64%baseline ahead
spotting 10+ hauls63%68%baseline ahead
FPL's own projection (ep_next)ordering only · same 195 players
MeasureModelBaselineVerdict
ordering the players0.260.19model ahead
spotting 6+ hauls60%58%model ahead
spotting 10+ hauls63%53%model ahead
GW516 misses · beaten by last season's points per game on spotting 10+ hauls; top-10k effective ownership on spotting 6+ hauls; top-10k effective ownership on spotting 10+ hauls; fpl's own projection (ep_next) on ordering the players; fpl's own projection (ep_next) on spotting 6+ hauls; fpl's own projection (ep_next) on spotting 10+ haulsOpen
  • Worse than usual this week - the model was out by 2.53 points a player, against 2.40 across the season so far.
  • Too pessimistic overall: it predicted 636 points for 188 likely starters, who scored 723, so it was 12% too low.
  • Its ordering of the players barely matched the week's: rank correlation 0.11, where 1.00 would be a perfect order and 0 no relationship at all.
  • Last season's points per game was better than the model at spotting 10+ hauls this week: on the 144 players both scored, it ranked a hauler above a non-hauler 54% of the time, the model 52%.
  • Top-10k effective ownership was better than the model at spotting 6+ hauls this week: on the 188 players both scored, it ranked a hauler above a non-hauler 60% of the time, the model 58%.
  • Top-10k effective ownership was better than the model at spotting 10+ hauls this week: on the 188 players both scored, it ranked a hauler above a non-hauler 57% of the time, the model 50%.
  • FPL's own projection (ep_next) ordered the players better than the model this week: 0.12 to 0.11 on the 188 players both scored.
  • FPL's own projection (ep_next) was better than the model at spotting 6+ hauls this week: on the 188 players both scored, it ranked a hauler above a non-hauler 59% of the time, the model 58%.
  • FPL's own projection (ep_next) was better than the model at spotting 10+ hauls this week: on the 188 players both scored, it ranked a hauler above a non-hauler 50% of the time, the model 50%.
  • Spotting 6+ hauls was no better than a coin toss this week. It came out at 58% and the interval runs from 50% to 66%, and takes in 50%.
  • Spotting 10+ hauls was no better than a coin toss this week. It came out at 50% and the interval runs from 33% to 67%, and takes in 50%.
  • The team scored 30, below the average manager's 48.
  • The team finished 23.6 behind a sample of top-10k managers (average 53.6, 329 sampled). This early, the sample is chosen on the very weeks it is scored on, so the gap is overstated.
  • Captain João Pedro scored 0, doubled to 0.
  • The team scored 18.4 fewer than the model predicted for it at the deadline: 48.4 predicted, 30 scored.
  • The best team the model could legally have bought scored 49, 33% of the 150 a perfect-hindsight XI would have made.
  • The model was closer than last season's points per game on points per player: 2.51 to 2.51 on the 144 players both scored.
  • The model was closer than last season's points per game on the big misses: 3.39 to 3.44 on the 144 players both scored.
  • The model ordered the players better than last season's points per game: 0.10 to 0.08 on the 144 players both scored.
  • The model was better than last season's points per game at spotting 6+ hauls: 55% to 54% on the 144 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.11 to 0.08 on the 188 players both scored.
  • Only 12 players scored 10+ this week, too few to claim a result on the 10+ figure.

Scored against the run of 18 Sept, 16:54 UTC · rows a4155c76…b085 · full report →

Against the baselines

Last season's points per gamesame 144 players
MeasureModelBaselineVerdict
points per player2.512.51model ahead
big misses3.393.44model ahead
ordering the players0.100.08model ahead
spotting 6+ hauls55%54%model ahead
spotting 10+ hauls52%54%baseline ahead
Top-10k effective ownershipordering only · same 188 players
MeasureModelBaselineVerdict
ordering the players0.110.08model ahead
spotting 6+ hauls58%60%baseline ahead
spotting 10+ hauls50%57%baseline ahead
FPL's own projection (ep_next)ordering only · same 188 players
MeasureModelBaselineVerdict
ordering the players0.110.12baseline ahead
spotting 6+ hauls58%59%baseline ahead
spotting 10+ hauls50%50%baseline ahead
GW46 misses · beaten by last season's points per game on points per player; top-10k effective ownership on spotting 10+ hauls; fpl's own projection (ep_next) on spotting 10+ haulsOpen
  • Worse than usual this week - the model was out by 2.64 points a player, against 2.40 across the season so far.
  • Last season's points per game was closer than the model on points per player: 2.71 to 2.71 on the 162 players both scored.
  • Top-10k effective ownership was better than the model at spotting 10+ hauls this week: on the 205 players both scored, it ranked a hauler above a non-hauler 66% of the time, the model 58%.
  • FPL's own projection (ep_next) was better than the model at spotting 10+ hauls this week: on the 205 players both scored, it ranked a hauler above a non-hauler 74% of the time, the model 58%.
  • Spotting 10+ hauls was no better than a coin toss this week. It came out at 58% and the interval runs from 41% to 76%, and takes in 50%.
  • The team finished 6.7 behind a sample of top-10k managers (average 89.7, 289 sampled). This early, the sample is chosen on the very weeks it is scored on, so the gap is overstated.
  • The model was closer than last season's points per game on the big misses: 3.45 to 3.51 on the 162 players both scored.
  • The model ordered the players better than last season's points per game: 0.30 to 0.24 on the 162 players both scored.
  • The model was better than last season's points per game at spotting 6+ hauls: 67% to 62% on the 162 players both scored.
  • The model was better than last season's points per game at spotting 10+ hauls: 62% to 56% on the 162 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.31 to 0.18 on the 205 players both scored.
  • The model was better than top-10k effective ownership at spotting 6+ hauls: 70% to 61% on the 205 players both scored.
  • The model ordered the players better than fpl's own projection (ep_next): 0.31 to 0.26 on the 205 players both scored.
  • The model was better than fpl's own projection (ep_next) at spotting 6+ hauls: 70% to 60% on the 205 players both scored.
  • The team scored 83, 14 above the average manager's 69.
  • Only 14 players scored 10+ this week, too few to claim a result on the 10+ figure.

Scored against the run of 12 Sept, 11:50 UTC · rows a0497656…c5bd · full report →

Against the baselines

Last season's points per gamesame 162 players
MeasureModelBaselineVerdict
points per player2.712.71baseline ahead
big misses3.453.51model ahead
ordering the players0.300.24model ahead
spotting 6+ hauls67%62%model ahead
spotting 10+ hauls62%56%model ahead
Top-10k effective ownershipordering only · same 205 players
MeasureModelBaselineVerdict
ordering the players0.310.18model ahead
spotting 6+ hauls70%61%model ahead
spotting 10+ hauls58%66%baseline ahead
FPL's own projection (ep_next)ordering only · same 205 players
MeasureModelBaselineVerdict
ordering the players0.310.26model ahead
spotting 6+ hauls70%60%model ahead
spotting 10+ hauls58%74%baseline ahead
GW34 misses · no baseline aheadOpen
  • Its ordering of the players barely matched the week's: rank correlation 0.16, where 1.00 would be a perfect order and 0 no relationship at all.
  • Spotting 10+ hauls was no better than a coin toss this week. It came out at 64% and the interval runs from 46% to 80%, and takes in 50%.
  • The team finished 1.3 behind a sample of top-10k managers (average 57.3, 170 sampled). This early, the sample is chosen on the very weeks it is scored on, so the gap is overstated.
  • The best team the model could legally have bought scored 51, 40% of the 127 a perfect-hindsight XI would have made.
  • Better than usual this week - the model was out by 2.30 points a player, against 2.40 across the season so far.
  • The model was closer than last season's points per game on points per player: 2.40 to 2.48 on the 156 players both scored.
  • The model was closer than last season's points per game on the big misses: 3.08 to 3.25 on the 156 players both scored.
  • The model ordered the players better than last season's points per game: 0.20 to 0.07 on the 156 players both scored.
  • The model was better than last season's points per game at spotting 6+ hauls: 68% to 56% on the 156 players both scored.
  • The model was better than last season's points per game at spotting 10+ hauls: 64% to 43% on the 156 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.16 to 0.09 on the 199 players both scored.
  • The model was better than top-10k effective ownership at spotting 6+ hauls: 69% to 55% on the 199 players both scored.
  • The model was better than top-10k effective ownership at spotting 10+ hauls: 64% to 49% on the 199 players both scored.
  • The team scored 56, 5 above the average manager's 51.
  • Only 42 players scored 6+ this week, too few to claim a result on the 6+ figure.
  • Only 12 players scored 10+ this week, too few to claim a result on the 10+ figure.

Scored against the run of 4 Sept, 16:27 UTC · rows 6f8b387e…5a77 · full report →

Against the baselines

Last season's points per gamesame 156 players
MeasureModelBaselineVerdict
points per player2.402.48model ahead
big misses3.083.25model ahead
ordering the players0.200.07model ahead
spotting 6+ hauls68%56%model ahead
spotting 10+ hauls64%43%model ahead
Top-10k effective ownershipordering only · same 199 players
MeasureModelBaselineVerdict
ordering the players0.160.09model ahead
spotting 6+ hauls69%55%model ahead
spotting 10+ hauls64%49%model ahead

FPL's own projection (ep_next) isn’t available this GW: no FPL ep_next capture in this gameweek's pre-deadline window.

GW25 misses · beaten by last season's points per game on spotting 10+ hauls; top-10k effective ownership on spotting 6+ hauls; top-10k effective ownership on spotting 10+ haulsOpen
  • Last season's points per game was better than the model at spotting 10+ hauls this week: on the 163 players both scored, it ranked a hauler above a non-hauler 82% of the time, the model 76%.
  • Top-10k effective ownership was better than the model at spotting 6+ hauls this week: on the 203 players both scored, it ranked a hauler above a non-hauler 67% of the time, the model 67%.
  • Top-10k effective ownership was better than the model at spotting 10+ hauls this week: on the 203 players both scored, it ranked a hauler above a non-hauler 74% of the time, the model 72%.
  • The team finished 16.9 behind a sample of top-10k managers (average 107.9, 60 sampled). This early, the sample is chosen on the very weeks it is scored on, so the gap is overstated.
  • 12 points sat on the bench and did not count (no Bench Boost).
  • Better than usual this week - the model was out by 2.37 points a player, against 2.40 across the season so far.
  • The model was closer than last season's points per game on points per player: 2.34 to 2.39 on the 163 players both scored.
  • The model was closer than last season's points per game on the big misses: 3.21 to 3.25 on the 163 players both scored.
  • The model ordered the players better than last season's points per game: 0.43 to 0.23 on the 163 players both scored.
  • The model was better than last season's points per game at spotting 6+ hauls: 71% to 66% on the 163 players both scored.
  • The model ordered the players better than top-10k effective ownership: 0.36 to 0.27 on the 203 players both scored.
  • The team scored 91, 10 above the average manager's 81.
  • Only 43 players scored 6+ this week, too few to claim a result on the 6+ figure.
  • Only 13 players scored 10+ this week, too few to claim a result on the 10+ figure.

Scored against the run of 28 Aug, 17:22 UTC · rows dc666d46…9013 · full report →

Against the baselines

Last season's points per gamesame 163 players
MeasureModelBaselineVerdict
points per player2.342.39model ahead
big misses3.213.25model ahead
ordering the players0.430.23model ahead
spotting 6+ hauls71%66%model ahead
spotting 10+ hauls76%82%baseline ahead
Top-10k effective ownershipordering only · same 203 players
MeasureModelBaselineVerdict
ordering the players0.360.27model ahead
spotting 6+ hauls67%67%baseline ahead
spotting 10+ hauls72%74%baseline ahead

FPL's own projection (ep_next) isn’t available this GW: no FPL ep_next capture in this gameweek's pre-deadline window.

GW1seed5 misses · no baseline aheadOpen
  • Worse than usual this week - the model was out by 2.75 points a player, against 2.40 across the season so far.
  • Too optimistic overall: it predicted 332 points for 95 likely starters, who scored 302, so it was 10% too high.
  • 11 points sat on the bench and did not count (no Bench Boost).
  • Captain Mbeumo scored 2, tripled to 6 by the Triple Captain.
  • The best team the model could legally have bought scored 38, 26% of the 148 a perfect-hindsight XI would have made.
  • The model was closer than last season's points per game on points per player: 2.75 to 2.93 on the 95 players both scored.
  • The model was closer than last season's points per game on the big misses: 3.35 to 3.38 on the 95 players both scored.
  • The model ordered the players better than last season's points per game: 0.29 to 0.27 on the 95 players both scored.
  • The team scored 59, 9 above the average manager's 50.
  • No haul figures for this week. The predictions it's marked against didn't keep haul probabilities, so I've nothing to score this on.
  • GW1 is marked against the predictions committed before the season started (the "seed"), because I wasn't keeping the live runs yet.
  • There's no elite comparison this week because no top-10k sample was taken.

Scored against the seed of 18 Aug, 21:51 UTC · rows 4d566e22…dc02 · full report →

Against the baselines

Last season's points per gamesame 95 players
MeasureModelBaselineVerdict
points per player2.752.93model ahead
big misses3.353.38model ahead
ordering the players0.290.27model ahead

Top-10k effective ownership isn’t available this GW: no top-10k capture before this gameweek's deadline.

FPL's own projection (ep_next) isn’t available this GW: no FPL ep_next capture in this gameweek's pre-deadline window.

Every group of players, and the reliability tablesThe same error on four groups of players, and a check on whether the model’s probabilities hold up.Open

Points per player, on every group I could score it on

GroupGW1GW2GW3GW4GW5GW6GW7GW8GW9Season
Every player the model projected1.49n=5841.18n=6021.08n=6181.18n=6541.11n=6591.06n=6671.09n=6671.09n=6681.00n=6691.14n=5788
Players the model gave any minutes1.67n=5191.27n=5561.29n=5181.41n=5491.33n=5501.26n=5581.31n=5591.31n=5571.20n=5591.34n=4925
Players who actually playedhindsight2.21n=3011.89n=2991.91n=2902.13n=3072.04n=3011.88n=3151.93n=3041.99n=3031.81n=3101.98n=2730
Likely startersthe headline2.75n=952.37n=2032.30n=1992.64n=2052.53n=1882.24n=1952.33n=1992.51n=1822.14n=1912.40n=1657

About half of all projected rows score exactly 0. That drags the every-player figure down and I never show it on its own. The players-who-played row is picked after the fact (a player who didn’t play isn’t in it), so it’s marked hindsight. If you’re comparing this with someone else’s error figure, check it’s on the same group of players first or it won’t tell you much.

When the model says 70%, does it happen 70% of the time?

Plays 60+ minutes
played 60+ minutes · every projected player-fixture whose run kept this probability (n=5788 of 5788)
PredictedObserved95% rangen
0%0%0–01158
1%0%0–1579
2%2%1–4578
6%7%5–9579
15%21%17–24579
35%40%36–44578
66%68%64–72579
85%89%86–91579
93%93%91–95579
Scores 6+
scored 6+ points · every projected player-fixture whose run kept this probability (n=5204 of 5788)
PredictedObserved95% rangen
0%0%0–01076
0%0%0–1499
1%1%0–2507
2%1%1–3520
5%3%2–5521
9%9%7–11520
15%15%12–18522
21%23%19–26518
32%34%30–39521
Scores 10+
scored 10+ points · every projected player-fixture whose run kept this probability (n=5204 of 5788)
PredictedObserved95% rangen
0%0%0–01119
0%0%0–1452
0%0%0–1530
0%0%0–1510
1%1%0–2512
2%2%1–4520
3%2%1–4521
5%6%4–8519
10%8%6–11521

Bins of the model’s stated probability against the rate at which it happened, with Wilson 95% intervals. A calibrated forecaster’s observed column tracks its predicted column; the range says how much a small bin can be trusted.

How well did it predict the lineups?

Before every deadline the model names each club's likeliest XI. Here's how many of the real starters it got.

Starters named

86.0%9.46 of 11 a side
1514 of 1760 starters across 160 sides, GW2–9.

Against last match's XI

86.2%vs83.7%
+0.27 starters a side (95% CI +0.13 to +0.41), on the 159 sides where both exist.

9+ of 11 right

84%of sides
134 of 160 sides. 88 had 10+ right and 25 were perfect.
Predicted XIs against actual starters per settled gameweek, beside last match’s XI.
GameweekStarters namedOf 11Perfect10+ rightLast match’s XI
GW1seed69.5%7.650/202/20—
GW285.5%9.404/2011/2084.7%
GW388.6%9.753/2014/2088.2%
GW486.4%9.504/209/2086.4%
GW588.2%9.703/2014/2083.6%
GW687.3%9.603/2012/2083.2%
GW782.3%9.052/209/2081.4%
GW883.2%9.152/208/2077.7%
GW986.8%9.554/2011/2084.5%
SeasonGW2–986.0%9.4625/16088/16083.7%on 159 sides

Each side's XI is its likeliest keeper and ten likeliest outfielders by expected minutes, from the pre-deadline file this page hashes for that gameweek, so anyone can rebuild it. The Fixtures page ranks by chance of starting; across three past seasons the two rankings named the same share of starters. Scored on every side with 11 recorded starters.

GW1 is marked against the pre-season seed, not a deadline run, so it's shown but left out of the season figure.

This week’s predicted lineups →

Who beat the model's number, and who fell short

The players furthest above and below their predictions this season.

Points scored minus the model’s prediction, added up over the settled gameweeks. Only weeks the model expected the player to start (60+ minutes) count. Most of it is variance, so I wouldn’t make a transfer off the back of it.

Scored above the model’s number

GroßBHAMIDxA +4.4
9 GWs · predicted 35.6 → scored 60
+24.4
VuskovicBHADEF
8 GWs · predicted 27.7 → scored 52
+24.3
VerbruggenBHAGK
9 GWs · predicted 28.5 → scored 51
+22.5
De CuyperBHADEFxA +2.9
8 GWs · predicted 29.7 → scored 52
+22.3
KonsaARSDEF
5 GWs · predicted 18.4 → scored 40
+21.6

Scored below the model’s number

WissaNEWFWDxG −2.4
7 GWs · predicted 33.5 → scored 15
−18.5
GonzaloFULFWDxG −3.0
8 GWs · predicted 34.3 → scored 16
−18.3
O'ReillyMCIDEFxG −1.2
5 GWs · predicted 22.9 → scored 7
−15.9
CanvotCRYDEF
7 GWs · predicted 21.6 → scored 6
−15.6
ThiawNEWDEF
8 GWs · predicted 29.5 → scored 16
−13.5