Questions and answers

You play coach of a football team. Before a match you set the lineup of 11 players and a strategy (formation, playing style, two conditional substitutions). After the final whistle, Dobitka replays your version 5,000 times and shows how it could have unfolded. You get post-match commentary, stats and points in the fan ranking.

The current game is the 2026/27 club season: you pick a club from one of six leagues - Ekstraklasa, Premier League, La Liga, Bundesliga, Serie A or Ligue 1 (Dobitka's Top 6) - and lock a match plan before every round. World Cup 2026 is over - your picks, verdicts and tournament rankings stay in the archive.

No.You are not trying to guess the real score. You are trying to prove that your plan would have been better than the real coach's plan. Dobitka does not take deposits, does not pay cash prizes and is not financial advice. It is an analytical game for adult football fans.

Dobitka is completely free: the club season costs nothing and the World Cup 2026 archive is open to browse. No paid offer is active - if that changes (for example starting with a future season), terms will be shown clearly before purchase. The only payment option is voluntary support (buy us a coffee). The club season is live and match-plan saving is available. Current status on /pricing, coffee on /support.

You can pick your first club at any point in the available season, including after 31 August at 23:59. If you already lead a club, you can change your coaching bench once until that deadline. We remove unsettled plans for the old club, while settled results remain in your account history. The next change opens from 1 January through 1 February 2027. A new account currently saves one club; we do not sell an upgrade to that limit.

Yes - you choose a club from Dobitka's Top 6: the Polish Ekstraklasaand Europe's five biggest leagues (Premier League, La Liga, Bundesliga, Serie A, Ligue 1). The Champions League is not a separate pool of selectable clubs - a Champions League match involving your chosen club may appear in its fixture list. During onboarding, over 100 teams from these six leagues are available. World Cup 2026 (48 national teams) was a separate mode - today it is a results archive. Lower divisions and other countries are not supported yet.

The primary absence report comes from API-Football. The additional Sportmonks feed for Ekstraklasa is conditional: it runs only when explicitly enabled with a valid token. A disabled feed does not block the simulation and does not pretend that it supplied data.

When match records are available, we compare the report with minutes played: if a player listed as injured has been on the pitch since the injury date, the entry goes. Sitting on the bench is not enough - that is exactly where clubs park players easing back in. If the minutes check fails, we keep the provider warning and mark the refresh as partial instead of declaring the player fit.

When we have no report for your club at all, we say so in a banner rather than show a full squad and let you assume everyone is fit. Those are two different things and you should not have to guess which one you are looking at.

For Ekstraklasa yes, and we count them ourselves. Cards arrive with the player, the club and the minute, and the competition rules are unambiguous: the 4th yellow costs one match, the 8th another one, the 12th and every fourth after that cost two. Only league cards count - the domestic cup keeps its own tally. A red card means at least one match.

For someone picking a lineup a suspension is the same absence as an injury, except this one can be computed rather than guessed. We only show the certain cases: if a player picked up two yellows in one match the tally can be read more than one way, and we would rather claim nothing.

You get: the most common exact score across 5,000 replays, plus minutes, scorers and key moments from a representative iteration that ended with that score; stats (the 1X2 distribution, common scores and xG), match commentary in the style of your chosen commentator persona (Pub, Classic, Analytical, and for national-team matches - Anthem), a comparison with the real result and the real match xG plus, separately, the impact of your decisionsversus the model's reference plan (a comparison of two simulations, not the real match) and points in the ranking. All on one page with an option to share a screenshot with friends.

Nothing bad: a round without a plan is simply 0 points - no penalties, nothing negative. Only the rounds you actually play count; you can come back at any point of the season and chase the table.

In club mode the ranking is a table of managers of a single club. All FC Barcelona managers are in one table, all Lech managers in another, and so on. You earn points for your version of the match being better than what the club actually did - and still credible. The plan's quality against its reference and the impact of applied changes also count. The full formula is in Scoring - everyone can verify their result step by step. The table holds for the whole season.

You are not guessing the real result - this is not a prediction game. There are 100 points on the table, split into three baskets:

• How your match ended- the result of your version, a bonus for beating what the club actually did, and points for the club's real result.
• How good your plan was - how far your lineup and style raised the win chance above the plan we pre-fill for you. This is the part that settles the table.
• What you did in-game, and whether it holds up - changes that came on and genuinely helped, plus a check that such a match could have happened at all.

A settled plan earns at least 5 pts, and a round without a plan is simply zero - no penalties, nothing negative. All thresholds, examples and the exact formula: /punktacja. One-screen overview: How it works.

A decent score starts around half the pool, and anything above 70 needs a good match and a good plan at the same time. But the number on its own says less than the fraction next to it - for example 47/88. The denominator moves with what your club did on the pitch, so that is the honest reference point, not a round 100.

The quickest answer to “is this good?” is your club's manager table: everyone in it had exactly the same ceiling in that match, so the comparison is fair. Under the points breakdown on your result page you will also find one concrete sentence about where you left the most.

The current model shows the most common exact scoreacross 5,000 replays of your plan. It does not draw a rarer result as the official score.

Minutes, scorers and the timeline come from a representative iteration that ended with that score. Selection is deterministic: the same lineup and tactics replay the same result and timeline. Historical results retain the semantics of the version that stored them.

The result page holds two different comparisons, and each answers a different question:

• Your version vs the real match - the verdict, score points and the real-xG section put your simulation next to what the club actually did on the pitch.
• Your decisions vs the model's reference plan - the “Impact of your decisions” block and the “Key point” line compare two simulations: your plan and the lineup suggested at kick-off, under identical conditions. The real coach never played your lineup, so this part cannot be measured against the real match.

That is why your simulated xG can be lower than the real xG while your eleven still “helped”: simulated xG is the average of 5,000 replays of your plan, while real xG is one specific run of the match - a single match regularly lands above or below expectation.

Short version: a single match is chance - a miss from 5 metres, an injury in the 20th minute and that is it. Five thousand replays (Monte Carlo - we play the match over and over, each replay a different scenario produced by the model) tell you whether your lineup really made sense. You get: a percentage chance of win/draw/loss, the distribution of the most common scores and the most common exact score. Minutes, scorers and the timeline come from a representative iteration that ended with that score.

Under the hood: For each club, ClubElo is the strength baseline. If the data service provides enough current, valid results for that team, the model blends ClubElo with a fresh Dixon-Coles fit. If the fit is missing or fails validation, it uses ClubElo; if a ClubElo rating is also unavailable, it falls back to a conservative starting value. Poisson is the default goal distribution; bivariate Poisson is an optional variant. Each of the 5,000 replays plays out an alternative match scenario from the lineup, formation, playing style and plan B.

The model accounts for, among others:

• club attack and defence strength. For each club, ClubElo is the strength baseline. If the data service provides enough current, valid results for that team, the model blends ClubElo with a fresh Dixon-Coles fit. If the fit is missing or fails validation, it uses ClubElo; if a ClubElo rating is also unavailable, it falls back to a conservative starting value.
• xG and xA of each of the 11 players(xG = shot quality, xA = quality of passes leading to a shot) plus their role on the pitch - a creator like Pedri / De Bruyne / Modric, who does not score himself but drives the team's chances, is valued differently from an out-and-out striker, plus that player's form across their own last five matches (player form, not club form),
• home advantage (different weights for clubs and national teams),
• your formation (12 to choose from),
• playing style - 4 sliders: pressing, tempo, directness, defensive line,
• conditional substitutions (max 5 per FIFA: after a minute, when losing, after a red card, after an injury),
• plan B (changing formation or style during the match),
• reaction to a red card (keep the formation / park the bus / all-in),
• injury plan - who to bring off the bench on an injury (separate backup for GK / defender / midfielder / striker),
• the same match without your decisions - we run it a second time on the default setup, so you can see how much you moved yourself.

What we do not count: weather, kilometres on the coach, derby temperature. The engine can do all of it, but those corrections are copied from the literature and we have not yet checked on our own settled matches whether they help at all. Until we do, they stay off.

The exact weights, parameter calibration and internal model extensions remain our secret ;) - thanks to that your result cannot be reproduced with pen and paper.

The same lineup and the same strategy always give the same result - we show the modal score with a deterministically selected representative iteration. Refreshing the page or coming back next week changes nothing.

The current model uses 5,000 replays. That produces the full score distribution at a cost that fits the settlement path. A seed keeps the distribution and modal score repeatable for the same plan. The full result appears after the real score is confirmed; we do not promise a fixed time because it depends on the data source and settlement queue.

Refereeing style (red cards and penalties are randomised the same way, regardless of who officiates the match), motivation like “playing for promotion vs a comfortable mid-table”, the exact cohesion of the lineup (how many matches a given eleven has played together), style clashes (whether high pressing breaks a team that builds slowly from the back) or the stadium atmosphere. These things we are still working on. For now we deliberately simplify the model - it is meant to be stable and repeatable. See how often the model is right and where it misses.

We tune them, but never retroactively. Every plan is stamped with a rules version the moment it locks, and it is settled under that version even if we switch on a newer one later. Your settled points are never recalculated - and a link from an old result opens exactly the formula we used for it. The /punktacja page shows the version active for new plans by default.

Matches, lineups, stats:API-Football v3 (Dobitka's Top 6 fixtures: Ekstraklasa, Premier League, La Liga, Bundesliga, Serie A and Ligue 1; Champions League matches for selected clubs; World Cup 2026). Team form, xG, xA, ratings: our own statistical model built on FBref data (a public football stats database) and historical matches. Match commentary: the Claude model with a separate voice for each persona. Simulation: Monte Carlo. For each club, ClubElo is the strength baseline. If the data service provides enough current, valid results for that team, the model blends ClubElo with a fresh Dixon-Coles fit. If the fit is missing or fails validation, it uses ClubElo; if a ClubElo rating is also unavailable, it falls back to a conservative starting value. Poisson is the default goal distribution; bivariate Poisson is optional. The current season is settled from the model's raw 1X2 distribution, without an additional calibration layer changing points or the ranking.

Yes. We do not store passwords or card data. We process your email (for code-based login), nickname, favourite club, your picks, setups and simulation results, plus minimal technical data needed for operation and security. We do not collect tracking cookies without consent, do not sell data, do not pass it to third parties for marketing. The app and the database run in the EU - Vercel functions in Dublin, the Supabase database in Ireland (the hosting provider is Vercel Inc., USA, with transfers under the EU-US Data Privacy Framework). Connection over HTTPS, passwords do not exist (login via an email code or Google OAuth). Full description in the privacy policy. We delete accounts on request: write to reklamacjedobitka.pl from the email linked to your account and we will remove your data as described in the terms (section 5 of the Polish version).

Email kontaktdobitka.pl - we reply within 24 working hours. All legal matters (complaints, withdrawal, GDPR) are described in the terms.

The tournament ended with the final on 19 July 2026 and is now an archive. New picks, lineups and World Cup leagues can no longer be saved - what remains is a full read-only view: the bracket with results, the final predictions table and your verdicts. Nothing you scored disappears. Enter via the /mundial page.

The World Cup had 3 separate rankings: Picks (tournament predictions, one general table), Team fans (a table per national team) and Neutrals (a separate global ranking). All of them are settled and frozen.

• Final tables: /mundial/leaderboard
• Bracket with results: /mundial/drabinka
• Your picks (view only): /mundial/picks

From two sources. Tournament predictions (locked on 18 June, after the first group round): World Cup champion 20 pts, top scorer 15, best goalkeeper 10, group winner 5 each (×12), 2nd and 3rd place in a group 3 and 2, play-off bracket 2 pts per correct slot in the round of 16 and beyond.

Match plans - points for every locked plan, analogous to club scoring: better than the real coach + a sensible setup + courage + substitutions. The simulation was based on Elo (less accurate than club - no individual xG for all 736 World Cup players). Results are settled and frozen in the archive rankings.