Sportpools selection optimiser

This package optimises your Sportpools selection for a Grand Slam. It supports a web UI and a CLI, both running on the same prediction pipeline:

• Elo ratings from Tennis Abstract, an exact dynamic-programming bracket simulation, and an integer program that optimises the full team (13 players + joker + kluns) under the real Sportpools rules.

Web UI

uv run sportpools-ui

Then open http://127.0.0.1:8000. One command runs the FastAPI backend and serves the built React frontend. The UI offers:

• a prediction runner with live stage-by-stage progress;
• the optimal team with per-player round probabilities (R4/QF/SF/F/title);
• joker and kluns alternatives with the underlying statistics (bonus formula, exit-round chances, and the total team value when forcing each kluns);
• the most probable route to the final for every player (most likely opponent per round with win chances) via the player drawer;
• intra-team matchups: when your own selections can meet, and how likely;
• interactive editing: swap players, change roles, live rule validation and expected points, plus re-optimisation around your locked players;
• side-by-side comparison of hard-court vs overall Elo models;
• saved teams (SQLite) and one-click export of your selection.

Frontend development (requires Node): cd web && npm install && npm run dev (proxies /api to localhost:8000), build with npm run build.

The CLI shares the pipeline with the web UI.

Game rules

• A team consists of 15 players: 13 regular players, one joker and one kluns.
• Each player has an amount of "black points" (0-5), derived from his seed: seeds 1-2 → 5, 3-4 → 4, 5-8 → 3, 9-16 → 2, 17-32 → 1, unseeded → 0.
• The non-kluns players may use at most 20 black points combined; the kluns's black points don't count and are added to the budget instead (the "kluns trick").
• Each win earns 10 - bp points, doubled from the fourth round onward.
• The tournament winner earns 50 bonus points.
• The joker earns a one-time bonus of 50 - 5 * bp for reaching round 4.
• The kluns earns nothing but subtracts 10 points per round he advances (capped at 50 if he wins the tournament).

Installation

pip install git+https://github.com/bartcode/sportpools-tennis.git

Simulation mode (default)

Fetch the draw from Wikipedia and the Elo ratings from Tennis Abstract, simulate and optimise:

sportpools --tournament us-open --year 2026 --surface hard

Recommend --surface hard for the US Open: hard-court Elo is the better predictor for a hard-court slam. Use --surface all for overall Elo as a sensitivity check.

Fetched pages are cached in .cache/ for 6 hours (adjust with --cache-ttl, set 0 to always refetch); if a fetch fails, a stale cached copy is used when available.

The report contains:

• the optimal 15 with each player's expected points and round-by-round chances (P(reach R4), P(QF), P(SF), P(F), P(title)) to tune the risk profile;
• a joker analysis: the bonus formula and the best joker candidates in the team;
• a kluns analysis: exit-round chances, expected penalty, recycled black points, and the best kluns alternatives with the total team value for each (the whole team is re-optimised for every candidate kluns);
• reserve suggestions.

Alternative inputs:

• --draw-url <url> fetches the draw from any URL (Wikipedia article or TennisTemple draw page; the format is auto-detected).
• --draw-url <file> reads a draw saved to disk. TennisTemple is behind Cloudflare, so for a TennisTemple draw save the page from your browser (HTML only) and point --draw-url at the file.
• --ratings-data <file> reads Elo ratings from a saved copy of the Elo report page instead of fetching it live.
• --rating-method dr computes Dominance Ratio ratings from Tennis Abstract JS match logs (pass the log files via --ratings-data) instead of using Elo.

The output lists the optimal team with each player's expected points, the joker and kluns choices, the black points budget usage, and reserve suggestions.

Forecast mode (legacy)

1. Go to TennisAbstract and select the forecast of the upcoming Grand Slam.
2. Save the page as HTML-only and save it to your disk.
3. Run sportpools -f ./page.htm.

In this mode you choose the kluns yourself with -l; the joker is suggested afterwards.

Notes

• The number of available black points is 20 by default, adjustable with -b.
• The simulator converts Elo-based best-of-3 win probabilities to best-of-5 for Grand Slams.
• Players missing from the Elo table (e.g. wild cards) get a conservative ATP-rank-based estimate; the run logs the rating coverage and any unmatched players.