Open data

The projections, in full, for anyone

Name: Sideline Quant FPL projections
Creator: Sideline Quant
License: https://creativecommons.org/licenses/by/4.0/

Every player’s expected points for the next gameweek, free for anyone to use with a credit. I publish the model’s accuracy on the settled gameweeks with them, which is how you’d check whether they’re any good.

Currently published

Gameweek 9 (669 players, generated 2026-10-27T03:34:58Z)

https://sidelinequant.com/data/gw/9/projections.json (every player and column)
https://sidelinequant.com/data/gw/9/projections.md (ranked extracts, easy to quote)
https://sidelinequant.com/data/latest.json (redirects to the current gameweek’s own address)

What makes this different

Every player. All of them, with the FPL id on every row so you can join it straight onto the official game’s data or your own.
Floor, ceiling and haul chance. Each comes from that player’s own simulations, which is why a nailed defender and a rotation-risk striker on the same xP can get very different ranges.
Start probability. Every row has one. Plenty of low-owned players are low-owned because they aren’t starting.
How it’s gone so far. The payload carries the model’s error on the settled gameweeks and links to the full record.
One address per gameweek. Each gameweek has its own URL and latest redirects to the current one. I only keep the current gameweek. Once the feed moves on, the old address returns a 404, so you can’t end up pulling this week’s numbers thinking they’re last week’s.

Columns

Every field the feed publishes, and what it means. Rows are at fixture grain: a player with two fixtures in the gameweek has two rows.

Field	Meaning
player_id	FPL element id. Stable across the season; join key to the official game.
fixture_id	FPL fixture id. Rows are at fixture grain, so a double gameweek is two rows.
gameweek	FPL gameweek number.
name	Full player name.
web_name	Short display name, as FPL renders it.
club	Player's club.
team	Player's club (same value as `club`).
position	GK, DEF, MID or FWD.
price_tenths	Current FPL price in tenths of a million (96 = £9.6m).
kickoff_time	Fixture kickoff, ISO 8601 UTC.
opponent	Opposition club in this fixture.
was_home	True when the player's club is at home.
xP	Expected FPL points for this fixture. An expectation over simulated minutes states, not a per-90 rate scaled by expected minutes.
sigma	Standard deviation of this fixture's simulated points.
gw_sigma	Standard deviation of the player's gameweek points total, summed across their fixtures in the gameweek.
gw_p10	10th percentile of the simulated gameweek points total (a realistic floor).
gw_p90	90th percentile of the simulated gameweek points total (a realistic ceiling).
horizon_sigma	Standard deviation of the player's points total summed over every gameweek the model currently covers (not just this one).
horizon_p10	10th percentile of the simulated points total over the whole horizon.
horizon_p90	90th percentile of the simulated points total over the whole horizon.
xp_cond_start60	Expected points if the player starts and plays 60+ minutes.
xp_cond_start_lt60	Expected points if the player starts but goes off before 60 minutes.
xp_cond_cameo	Expected points if the player comes off the bench.
e_mins	Expected minutes.
p_start	Probability the player starts.
p_start60	Probability the player starts and plays at least 60 minutes.
p_start_sub	Probability the player starts but is substituted before 60 minutes.
p_cameo	Probability the player appears as a substitute.
p_unused	Probability the player does not appear at all.
startsub_mean	Model's fitted average minutes for a start that ends before 60 minutes (this export's own estimate).
cameo_mean	Model's fitted average minutes for an appearance off the bench.
status	FPL's own availability flag (a available, d doubtful, i injured, s suspended, …).
chance_of_playing_next_round	FPL's own percentage chance of playing the next round, when FPL has set one.
news	FPL's own news text for the player, when there is any.
p_goal	Probability of scoring at least one goal.
p_goals_2plus	Probability of scoring two or more goals.
p_assist	Probability of at least one assist.
p_cs	Probability of a clean sheet for this player (conditional on playing 60+ minutes).
p_return	Probability of a goal or an assist.
p_defcon	Probability of hitting the defensive-contribution points threshold.
p_ge_6	Probability the gameweek points total is 6 or more.
p_ge_10	Probability the gameweek points total is 10 or more (a haul).
bonus_mean	Expected bonus points.
xbps_mean	Expected BPS (the score bonus points are awarded from).
xg_mean	Expected goals for the player in this fixture (mean over the simulation).
xa_mean	Expected assists for the player in this fixture (mean over the simulation).
total_points	FPL points actually scored so far this season.
xp_appearance	Expected points from minutes played (the appearance points).
xp_goals	Expected points from goals scored.
xp_assists	Expected points from assists.
xp_clean_sheet	Expected points from clean sheets.
xp_conceded	Expected points lost to goals conceded (zero or negative).
xp_saves	Expected points from saves.
xp_penalty_save	Expected points from penalty saves.
xp_penalty_miss	Expected points lost to missed penalties (zero or negative).
xp_defcon	Expected points from defensive contributions.
xp_cards	Expected points lost to yellow and red cards (zero or negative).
xp_own_goal	Expected points lost to own goals (zero or negative).
p_card	Probability of at least one yellow or red card.
form	Mean points over the player's last four gameweeks. Not FPL's own 30-day form.
l4_gws	How many gameweeks form and the other last-four fields cover (four or fewer).
starts_l4	Starts over those gameweeks (a double gameweek can count two).
mins_l4	Minutes played over those gameweeks.
xp_l4_gws	How many of those gameweeks have the model's last pre-deadline prediction.
xp_l4	The model's pre-deadline expected points summed over those covered gameweeks.
pts_vs_xp_l4	FPL points actually scored minus xp_l4, over the same covered gameweeks.
owned_pct	Percentage of all FPL managers owning the player.
elite_10k_owned_pct	Percentage of a stratified sample of overall top-10k managers owning the player.
elite_10k_eo_pct	Effective ownership in that top-10k sample: owned plus captained, as a percentage. A triple captain counts as a captain: the chip was played in the sampled gameweek, not this row's.
elite_10k_captain_pct	Percentage of that top-10k sample captaining the player.
elite_1k_owned_pct	Percentage of the sampled overall top-1k managers owning the player.
elite_1k_captain_pct	Percentage of the sampled overall top-1k managers captaining the player.
elite_delta_pct	elite_10k_owned_pct minus owned_pct, in percentage points.
price_change_percent	FPL's own progress towards the player's next price change, as a signed percent.
price_change_as_of	When that price state was captured, ISO 8601 UTC.
price_change_direction	Which way the counter is heading: rise or fall.
price_change_expected	True when the counter is past the change threshold tonight.
price_change_locked	True when FPL has price-locked the player (new to the game).
price_change_season_tenths	Price change since the season started, in tenths of a million.
price_change_per_hour_pct	Captured net-transfer rate, in the same percent units as price_change_percent, per hour.
price_change_projected_pct	Where the counter is projected to stand at tonight's change point, assuming the current rate holds (FPL's own assumption).
price_change_projected_tenths	The model's expected price change over the next three change points, in tenths of a million. An expectation, so fractional.
lambda_home	Model expected goals for the home side in this fixture.
lambda_away	Model expected goals for the away side in this fixture.
p_cs_home	Model clean-sheet probability for the home side.
p_cs_away	Model clean-sheet probability for the away side.
d_lambda_home	Change in lambda_home over the odds-drift window.
d_lambda_away	Change in lambda_away over the odds-drift window.
d_p_cs_home	Change in p_cs_home over the odds-drift window.
d_p_cs_away	Change in p_cs_away over the odds-drift window.
drift_from_ts	Start of the odds-drift window, ISO 8601 UTC.
drift_to_ts	End of the odds-drift window, ISO 8601 UTC.
pen_taker	True when the player is first-choice penalty taker.
corner_taker	True when the player is first-choice corner / indirect free-kick taker.
fk_taker	True when the player is first-choice direct free-kick taker.
rates_source	Where this fixture's team rates came from: market odds, or the team model.
has_unseen_team	True when a club in this fixture has no played matches to fit on yet.
shrinkage_weight	How much the player's scoring rates lean on the prior rather than his own minutes: near 1 is mostly prior (little data), near 0 his own record.

Licence and use

Licensed CC BY 4.0. Free to use, quote and redistribute, including commercially, with attribution to Sideline Quant and a link to https://sidelinequant.com. Please cache rather than poll: the payload carries an ETag, which only changes when a model refresh lands.

Everything the model predicts is free on the site, for every player and gameweek it covers, along with the whole-season view. This feed is the next gameweek of that, with all the columns, in a form a script can read. The optimiser’s free with an account, and supporters help pay for the site (see support).