How to read PitchIQ

How PitchIQ works

For each match PitchIQ gathers fixtures, tables, results, lineups, injuries, referees, weather and bookmaker prices from several sources (see Where the data comes from). On top of those facts it builds four independent reads:

• The risk index: how likely the favourite is to slip up, from fourteen rule-based factors.
• The PitchIQ model: its own chances for each result and for goals, from team strength carried across seasons.
• The data verdict: whether the model disagrees with the bookmakers, and whether chance quality and form back that disagreement up.
• The market finder: which markets came in more often than expected in the most similar matches of the last 25 years.

Before kickoff every read is archived unchanged, then scored after full time. That is how PitchIQ tests whether its reads hold up (see PitchIQ Model).

The match list

PitchIQ opens in light mode, easy to read outdoors; switch to dark mode with ☾ in the top bar (and back with ☀). Your choice is remembered on that device.

The header strip

Four shortcut cards for the day: the lowest-risk match still to come, the strongest favourite, the number of matches live now (or the next kick-off), and the day's totals. Tap any card to open it.

• Lowest risk: the lowest risk index among matches not yet started, from PitchIQ's latest read. Only a Low read (0–30) qualifies, and reads with most factors known come first (a provisional read is low only for lack of data). When none is left on the day you are viewing, the card looks ahead to the next days (up to three) and shows the day and time.
• Strongest favourite: the biggest win chance the bookmakers give anyone among matches not yet started, if it is 60% or more; it looks ahead to the next days the same way.
• The card record (the ✓ badge): every match that was on one of these cards when it kicked off counts as a pick. PitchIQ re-reads it in the 45 minutes before kickoff (team news and lineups arrive late), then settles it after full time: did the side the prices favoured win? A draw counts as not coming in. The badge shows the share that came in once ten picks have settled; hover it for the count and the average chance the bookmakers gave those favourites, which is what to compare it with. Picks before live tracking began were rebuilt from reads archived before each kickoff. Days run on UTC.
• More about a pick: the round ⌄ button on the lowest-risk and strongest-favourite cards opens the card across the strip: kick-off and competition, why the match was picked, its risk and the bookmakers' chance for the favourite, the card's record in words, and its latest picks as ✓ (came in) or ✗ (did not), newest first. Tap the card itself, or Open match, to go to the match.

Filtering and sorting

• Competition: the dropdown lists every competition playing that day with its match count.
• Live card: appears while matches are in play, previews their scores, and filters the list to them with one tap.
• Show: all matches, only those not started, not started and live, live only, or finished only (remembered on this device).
• Sort: by competition (grouped), or across all competitions by kick-off, highest or lowest risk, strongest favourite, closest contest, strongest data verdict, most market-finder picks, most or fewest goals expected, or lowest home or draw price. Matches still to play come first, then live, then finished. Risk, verdict, picks and goals come from PitchIQ's latest archived read; matches without one yet are listed last.

A match row

• Prices: home, draw, away decimal odds. The lowest (the favourite) is highlighted, and the star marks a clear favourite. Where the bookmakers with a strong African presence price the match, their best price is shown.
• The bar under the prices shows the chances those prices imply (green home, grey draw, blue away). With no prices yet, the row shows PitchIQ model percentages instead, in blue.
• Risk tag: the risk index from PitchIQ's latest read of the match, coloured by level (hover for when it was read). Opening the match works it out fresh and the row updates to match; a tag marked * and dashed comes from a read made under earlier rules and will be refreshed.
• After full time: chips compare the result with what was expected: As expected, Favourite held (a draw), Upset, more or fewer goals than expected, and whether the risk level, the verdict and the market finder's picks were right.

League character

Each competition header shows its character in one line: goals per game, draws, over 2.5 goals and how often favourites slip up. For the competitions in the 25-year archive these come from its last three seasons; for the others (MLS, Brasileirão, the cups and more) from this season's and last season's results, where "favourites slip" cannot be measured without past prices, so home wins are shown instead.

The risk index

The risk index measures how likely an upset is, not who wins. Upset here means a clear favourite (at least 8 points ahead on the bookmakers' win chances) failing to win.

Fourteen factors are each read as clear, partial or triggered by fixed rules. Each has a weight. A triggered factor adds its whole weight, a partial or unknown one half of it, and the total is shown as a share of all the weights (37), from 0 to 100.

The points beside each factor are what it adds to the index, so they add up to it. In the bar under the level, each coloured part is one factor's points and the green remainder is the clear share that no factor adds. When factors carrying at least 40% of the weight are unknown, the index is marked provisional.

Where the weights come from

For the factors the archive can measure, the weight is set by how much more often favourites failed to win when the factor triggered, fitted on 2000/01–2015/16 and checked on the seasons since: one point of weight per 1.5 points of extra upset rate. Market moves are measured on the seasons with closing prices (since 2019/20): a favourite drifting 5+ points failed to win 53-55% of the time, one shortening 3+ points only 39-42%. Team news, weather, rest and xG are not in the archive, so their weights follow their direct impact: missing key players weigh most.

The newest factor, the outsider's big wins, counts the outsider's league wins this season over sides level with or above today's favourite in the table. In the archive, favourites failed to win 42% of the time against outsiders with no such wins, 47% with one and 51% with two or more, the biggest gap of any factor. Wins over top-half or top-quarter sides that sit below the favourite made no difference (46.9% against 46.6%), so they are listed but not counted. The effect holds among heavy favourites too: where the outsider was priced under 20%, favourites failed 33% of the time with no big wins and 40% with two or more.

Most signals you might expect, such as a favourite below the outsider in the table, poor form or weak away results, raise the upset rate mainly because the prices are already closer in those games. The risk index shows how often the expected result fails. It does not claim an edge over the prices: the PitchIQ Model page tracks that separately.

The fourteen factors

Tap a factor on a match page to see the detail behind its reading and the numbers it used. This table is read from the rules PitchIQ is running, so it is always current.

Why so few unknowns

PitchIQ fills each factor from the best source available: referee records from the results archive or, for other leagues, from the referee's last 30 games on API-Football; xG from the archive or from recent games in all competitions; league character from the archive or from this season's and last season's results; price movement from a bookmaker's opening line or PitchIQ's own earlier reads. Knockout cup ties have no table, so season phase and motivation simply do not apply. What can still be unknown is what is genuinely not published yet, mostly team news, which clubs release close to kickoff.

Data verdict

A verdict is a value signal, not a prediction of the winner. It can point to an outsider the model rates a little higher than its price while the favourite is still the likelier winner. When it makes no call, the card says No call and shows the biggest gap only for reference.

The verdict looks at the market where PitchIQ's model and the bookmakers disagree most (home win, away win, over or under 2.5 goals). The model must rate it at least 5 points above the bookmakers. Then it checks two more signals: chance quality (xG, or shots on target in older seasons) and form; for goal markets, the league's scoring profile too. Where football-data.co.uk has no league results, both are read from each team's last five games on API-Football (all competitions, at least three with figures).

Two cautions each lower a verdict one level: a long shot (priced under 30%, where the model's edges have come in least often against the price) and the early season (under eight games). The card shows how each verdict level did in the learning run: how often it came in against the chance the bookmakers priced.

Market finder

The finder takes the match's pre-match numbers (prices, the model's chances and expected goals, chance creation, form, the league's profile), finds the 400 most similar matches in 25 years of results, and counts how often each of about 70 markets came in: goals lines, both teams to score, team goals, halves, double chance, handicaps and more.

A market is suggested when it came in clearly more often than expected (at least 3 points above, statistically solid) and in at least half of those matches. "Expected" is the bookmakers' price where there is one, otherwise the model, otherwise the league average. Only market groups that beat the usual expectation on seasons the finder had not learned from can be suggested.

In the full table, Fair is the price at which a market breaks even given how often it came in, and Price is the best price on offer at a bookmaker with an African presence (or the average). A price above the fair price is where the similar matches say the market is generous.

Leagues outside the archive (for example the Brasileirão, MLS or Eliteserien) are compared in the same way: their matches get the same fingerprint, built from the season model, recent games on API-Football and the league's own goals and draw rates, and the similar matches come from all 22 archive leagues. One archive league stands in for league-level figures: the top division closest in character (goals per game, over 2.5, both teams to score, draws, home wins and favourite upsets), with nearer countries preferred when the character is close, so Denmark borrows the Eredivisie and Norway the Bundesliga. The card names the stand-in. Domestic cups would borrow a league from their own country, but knockout games have no league table for the model, so the finder usually has nothing to compare.

PitchIQ model

The model rates every team's attack and defence from results, carrying strength across seasons (partly reset each summer), and turns two teams' ratings into expected goals for each side. From those it works out the chance of every scoreline, and so of each result and goals market. Its settings were fitted on 2000/01–2011/12 and tested on the seasons since.

Leagues outside that 25-year archive (for example Eliteserien, Allsvenskan or the J1 League) use a simpler season model built from this season's home and away records in the live table. It pulls every team towards the league average, so it tends to rate outsiders a little higher than the bookmakers do; verdicts built on it say so and have no learning-run record.

On a match page the bars compare the model (green) with the bookmakers (blue, margin removed), and the ± number is the gap in points. The bookmakers are still slightly more accurate overall, which is why the verdict asks for other signals to agree before backing the model's view.

Market and bookmakers

The market card shows the prices and the chances they imply once the bookmaker's margin is removed. The margin comes off long shots more than favourites, as bookmakers price it (see Margin in the glossary), so a strong favourite's chance reads a point or two higher than a plain share of the prices would give, and an outsider's a little lower. The bookmaker table compares every bookmaker PitchIQ has a price from, with the best price per outcome highlighted and those with a strong African presence (1xBet, Betway, Betano, Bet365, Sportingbet and others) listed first and tagged.

Opening line → now shows how the chances have moved since the bookmaker opened its market; big moves often mean team news.

Second opinion

API-Football's own prediction, shown beside PitchIQ's: its advice, win percentages and a comparison of form, attack, defence and head-to-head. Its percentages are coarse (often 45/45/10), so read the direction, and note whether it agrees with PitchIQ's model and the bookmakers.

Team news, lineups and conditions

• Team news: club-reported injuries and suspensions. Premier League news comes from Fantasy Premier League, with each player's chance of playing; other leagues from API-Football. KEY marks a regular starter. No reports at all is treated as unknown, never as a fully fit squad.
• Lineups: the confirmed starting elevens once published, usually about an hour before kickoff.
• Conditions: the kickoff-hour forecast at the venue (temperature, rain, wind), the referee and any derby or rivalry.

Form, table and team stats

• Form: the last five games in all competitions, oldest to newest, with points and goals. Hover a chip for the opponent, score and competition.
• League this season: every league game so far, the same games the table counts, so the two always match.
• Expected goals (xG): the quality of chances created and allowed in the last league games with xG, and across all competitions where published.
• Table: the current standings with title, European or promotion, and relegation places.
• Team stats: home and away splits against the league average (green better, red worse), clean sheets, both teams scoring, scoring first and what happened next, goal timing by 15-minute period, streaks and "match facts" that stand out.
• Head to head: recent meetings in every competition. The bar on top counts wins for each side and draws (home side green, away blue), with the edge when one side has won at least two more; each meeting is labelled with its winner. Only the last three seasons count towards the risk index.

Profiles and league character

Long-run facts from the archive: the league's last three seasons (home wins, draws and how many the bookmakers priced, goals, over 2.5, both teams scoring, favourites slipping), each team's home and away record and habits (holding half-time leads, comebacks), and the referee's career card record. Each comes with an "in short" line in everyday words.

Data check

PitchIQ combines several sources, and they can disagree when one lags or misses a game. The data check compares them for both teams: the table against the league results, the results archive against the table, and recent form against the results. When everything agrees you see a tick; when something differs it says exactly what, and which figures the page uses. For example: "football-data.co.uk has 5 of 6 league games so far: xG, the model and the verdict use those 5".

The Analyst and chat

What should I make of this? asks the Analyst to read the whole briefing and give the bottom line, what the evidence points to, what could undo it and how much confidence it deserves, plus a check that the data hangs together. The chat bubble answers questions about the open match or the day's matches, using only PitchIQ's data. Its reviews and answers are written automatically (AI-generated) from that data and cannot change the risk index; both are rate-limited.

Half-time read

Once a match kicks off, a Live read card appears at the top of its page and updates every minute; at the break it becomes the Half-time read. It re-works the result chances from the score:

• It starts from the goals each side was expected to score before kickoff (PitchIQ's model, or the bookmakers' prices where there is no model).
• It scales them to the time left: about 55% of goals come after half-time, adjusted towards each team's own habit (a side that scores most of its goals after the break gets more of them in the second half; the adjustment is pulled halfway back to the norm so a small sample cannot dominate).
• It nudges them by how the game has gone: expected goals (xG) created so far, shrunk halfway towards the pre-match view so one lucky half does not dominate.
• It allows for the game state: a trailing side pushes and scores about 15% more, the leader about 10% less.
• It adjusts for red cards: a side down to ten men scores about 30% less and concedes about 30% more.

The card shows each result's chance from here against before kickoff, the chance of over (and under) 1.5, 2.5 and 3.5 goals from here, both teams to score, the goals still expected, and plain notes that read the lead in context:

• An outsider in front is only called "an upset is likely" when it is now over 50% to win; otherwise the note says it is far from settled and how often the favourite still avoids defeat.
• The teams' habits: a trailing side that scores most of its goals after half-time, a side that often comes back, a leader that has let half-time leads slip (or rarely does).
• Red cards, and what they did to each side's chances and to the goals.
• Pre-match goals calls (the data verdict, and the market finder's goals picks from the archived pre-match read): marked "in danger" when their chance has fallen 15+ points, for example an "under 3.5" after a red card and an early goal.

If you saved the match and turned reminders on, the half-time read is also sent to you.

Track record: your bets

Sign in with your phone number (a code by text message), Google or email to keep a private record of your bets. The same screen creates an account the first time. To log a bet, tap ＋ Log a bet on a saved match (or on its match page) and tap what you bet on: one market for a single, or one per market group for a same-match combi. More markets opens draw no bet, more goal lines, half-time and second-half markets, each team's goals and clean sheets, handicaps, a correct score you type in, and your own pick in your own words (which you settle yourself). The bookmakers' price is filled in (for a combi, the prices multiplied, which bookmakers usually shorten, so check the price you got); the stake is optional and remembered. PitchIQ's read at that moment is kept with the bet. Anything else you can add by hand with ＋ Add a bet on the Track record page. Bets on PitchIQ matches settle themselves from the 90-minute score, as bookmakers settle; cancelled matches are void. Anything else you settle yourself, including half results and cash-outs.

The breakdowns show where you win and lose: by market, by price range, by the match's risk level when you bet, by whether PitchIQ's verdict or finder backed the bet, by stake size and by competition. Insights appear once at least eight bets have settled. Export everything as CSV at any time, or import a CSV in the same format.

Predictions

Test your own reading of a match, with no money involved. Before kickoff, tap ＋ Log a prediction on a saved match or its match page and pick the outcomes you expect, one per market (result, double chance, goals, both teams to score, team goals), and how sure you are. More markets adds draw no bet, more goal lines, the half-time result, first- and second-half goals, each team's goals and clean sheet, handicaps and a correct score you type in. Each team's goals count as separate markets, as do the half-time result and first-half goals. For anything not listed (a scorer, a red card, a penalty) write your own pick: PitchIQ cannot check it from the score, so after full time the prediction waits for you to mark whether it came in. Own picks have no bookmaker price, so they count in your hit rate but not in your edge. You can change or delete a prediction until kickoff.

After full time each pick is checked against the score. Track record → My predictions shows how often your picks came in against how often the bookmakers' prices (margin removed) said they would; the difference is your edge. It also breaks your record down by market, by how sure you were, and by whether you picked the same as PitchIQ's data verdict when it made a call. A record says little before about 30 settled picks.

Reminders

Save a match with ☆ and, once reminders are on, PitchIQ messages you twice: at half-time with the half-time read, and about an hour before kickoff, as soon as team sheets are out. It analyses the match again first, so the message carries the fresh read: risk and what drives it, key absences and any firm verdict. If lineups are still missing 40 minutes before kickoff, the reminder comes anyway and says so.

The number on the Saved tab counts every match you have saved, finished or not, until you unsave it (☆) or clear finished ones; results of finished matches update when you open Saved.

Turn reminders on from the Saved page. Notifications on your phone or computer work now. On iPhone, first add PitchIQ to your home screen (Share → Add to Home Screen) and open it from there. Telegram and email are on the way.

Your data and feedback

Before signing in: you can open 5 match analyses, ask for 3 Analyst reviews and send 1 message to the chat. After that PitchIQ asks you to sign in (free), and these limits no longer apply. Opening the same match again, or asking for its Analyst review again, does not use up another free one.

Contact: Contact in the top menu and at the bottom of the page opens a short form for questions, problems, data or privacy requests and anything else. Replies come by email, or on WhatsApp if you leave a number.

Feedback (from 29 October 2026, once you have had a few days with PitchIQ): tap the glowing Feedback button in the bottom-left corner. Three quick questions, all optional: how useful PitchIQ is for you (1 to 5), what it is about, and anything you want to add. It says which page or match you were on, so you only need to describe what happened. Signed in, it is linked to your account; otherwise leave an email or phone number if you want a reply.

Usage counts: while PitchIQ is being tested it counts how often each feature is used each day (match pages opened, Analyst reviews, chats, saves, bets added) and, for signed-in people, which days they used it. Nothing is logged page by page, and visitors who are not signed in only count towards the totals.

Deleting your account: on the Track record page, Delete account removes your account, saved matches, bets, reminder settings and your sign-in, at once. Export your bets as CSV first if you want to keep them.

The privacy notice lists everything PitchIQ keeps and who helps run it; the terms cover who can use PitchIQ (18 and over) and what it is not.

PitchIQ Model and learning run

The learning run replays every match since 2000/01 in 22 competitions. For each, the model sees only what was known before kickoff, then it is scored against the result and the bookmakers. Settings are fitted on early seasons and judged only on later ones, so the results are honest about how the method does on matches it has not seen.

The live record scores the reads PitchIQ archived before real kickoffs: whether the risk index's upset rates rise from low to high, and how accurate the model is against the market. It needs around 100 settled matches to say much.

Learning from finished matches. Every finished match is reviewed against PitchIQ's last scheduled read before kickoff: the prices, the fourteen factors, the verdict, the model and fourteen pre-match signals (such as the favourite's chance, price movement, the table gap, form, rest, xG and the gap between the model and the price), set against what happened. The PitchIQ Model page shows what the reviews add up to:

• Risk factors, live: how much more often favourites failed when each factor triggered, against the archive's figure. Once both sides have 150+ matches a factor is marked holding up, stronger or weaker live; factors the archive cannot measure (team news, rest, weather) are judged on live matches alone.
• Trend scan: for each signal, whether favourites failed more or less often than their prices said at its high end than at its low end. Because many signals are watched at once, chance alone throws up differences, so a trend is marked worth testing only with 150+ matches at each end and a difference at least 2.5 standard errors wide.
• Verdicts, live and competitions against the price, and the misses worth a look from the last 30 days: upsets rated low risk, strong and moderate verdicts that missed, and the results the prices least expected.

Nothing changes by itself. A trend worth testing is checked on the 25-year archive, the way the outsider's big wins factor was, and becomes a factor or changes the model only if it holds up on seasons it was not fitted on.

Brier score measures how close win/draw/loss chances were to what happened: 0 is perfect, lower is better. Calibration checks that things the model calls 60% happen about 60% of the time.

Where the data comes from

Every figure on a match page carries a source line: where it came from, whether it is observed, reported, derived or a forecast, and when it was fetched. "Stale copy" means a source was briefly unreachable and the last good copy is shown.

Glossary

Decimal odds

What a 1-unit winning bet returns, stake included. 2.50 returns 2.50 (1.50 profit).

Implied chance

The chance a price represents: 1 ÷ odds. 2.00 implies 50%.

Margin

The bookmaker's built-in edge: the implied chances of all outcomes add up to more than 100%. PitchIQ removes it before comparing, taking more of it off long shots than off favourites (the power method), because that is where bookmakers load it: in 198,000 archived matches, spreading the margin evenly made outcomes look 7.6% likely that came in 5.2% of the time; the power method shows them at 7.1% (6.7% came in), and its chances match results in every price band.

Favourite / outsider

The side the bookmakers rate more / less likely to win. A clear favourite is 8+ points ahead.

Upset

A clear favourite failing to win (a draw counts).

xG (expected goals)

The quality of chances: how many goals the shots taken would produce on average. A team scoring far above its xG may be due a correction.

Fair price

The odds at which a market breaks even given how often it comes in: 1 ÷ its chance.

Edge

How many points PitchIQ's model rates an outcome above the bookmakers.

BTTS

Both teams to score.

Asian handicap

A goal head start or deficit. On whole lines (−1, +1) a tie returns the stake.

Draw no bet

Stake returned if the match is drawn.

Double chance

Two of the three results, for example home or draw.

ROI

Return on investment: profit ÷ total staked.

Brier score

Accuracy of probability forecasts; lower is better.

AUC

How well a score separates two groups (upsets from non-upsets): 0.5 is chance, 1.0 perfect.

Bet responsibly

PitchIQ explains matches; it does not promise outcomes, and no data removes the bookmaker's margin. Decide a budget you can afford to lose, keep stakes steady, never chase losses, and take a break when betting stops being fun. Most bookmakers let you set deposit limits, time-outs and self-exclusion. If gambling is causing harm, seek help from a local support service.

← Back to matches