How it works
The method behind the forecasts, and how to read what the rest of the site shows.
Overview
The Rugby Tournament Predictor uses a Monte Carlo simulation approach to forecast tournament outcomes. It simulates the entire tournament thousands of times and reports how often each outcome occurred, so what it publishes is a probability for each outcome.
The pipeline works in three steps:
- Rate every team using historical match data and the Glicko-2 rating system
- Simulate every match in the tournament using those ratings to generate win probabilities and scores
- Repeat thousands of times and aggregate the results into the tables and charts you see
Glicko-2 ratings
Every team has a Glicko-2 rating, an evolution of the Elo system used in chess. Each team's rating has two components:
- Rating — the team's estimated strength. Higher is better. A team starts at 1500, except 25 established nations, which start from a rating set by their tier: 1550, 1400 or 1250.
- Rating deviation — how uncertain the system is about that rating. A team that has played few matches, or none recently, carries a larger deviation.
Ratings update after every match. A surprise upset causes a bigger shift than an expected result, and home advantage is factored in — the next section sets out how much each result moves a rating. The rankings also show an SG Rating — Scaled Glicko — which is the same number on a 1–99 range. The rescale is linear, so it orders teams identically and changes nothing but the unit.
It runs 1–99 rather than 0–100 on purpose. The two ends are left empty because they would say something the scale cannot: 100 reads as “nothing better is possible”, when all that is true is that nothing better has been recorded. The range is set from the ranked field's own history with headroom above it, so the strongest rating ever held here comes out in the mid-nineties and the rest of the scale is room for a side better than any yet seen. Reserving 0 does a second job: a team the table does not place shows a dash, and a zero would be impossible to tell apart from it.
Club and international rugby are rated separately, from separate match records, and the two scales have no bridge between them. The site shows one at a time so it cannot imply a comparison the data does not support.
How a rating moves
A rating changes only when the team plays. Before each match the two ratings say how likely each side is to win; afterwards the result is set against that, and both ratings move by how far it departed from what was expected. A result the ratings saw coming moves them a little, and an upset moves them a lot.
Four things set the size of the move:
- The venue
- — The update expects more of the home side. It gives the home team a head start of 120 rating points in a Test and 260 in club rugby, so between two evenly rated sides the home side is the expected winner 66% of the time in a Test and 81% in club rugby. A home win is close to what was expected and moves a rating little; an away win counts as an upset and moves it a lot. The same works in reverse for a defeat, which costs far more at home than away.
- The margin
- — A bigger win moves a rating further, with diminishing returns: winning by 20 counts about four times as much as winning by one, and winning by 40 about five times. A heavy favourite gets less credit for running up the score.
- The opponent
- — Beating a side rated above you moves a rating further than beating one rated below. A result against a side the engine knows little about counts for less, because its rating says less.
- How well the team is known
- — A team with a thin record carries a wider deviation, and a wider deviation moves further on each result. The deviation also widens the longer a team goes without playing.
The venue is the one that most often surprises. A side that has just played several matches away from home, and won some of them, can climb well above where its league position suggests, and a side that has just played at home can sit below it while winning. Over a season it largely evens out, because a side plays about as many matches away as at home.
That expectation belongs to the rating update. It is not the forecast published for the match, which measures home advantage competition by competition and adds the adjustments described further down.
Two more things to keep in mind when reading the table:
- Nothing resets between seasons. A rating describes how strong a side is now on everything it has done, which is a different question from where it stands in this season's table. The league column on the club rankings is a single season's finish, and says which.
- Every rating carries a deviation, shown beside it as ±. Two sides whose ratings are closer together than that are not separated by the rating, whatever order the table puts them in.
Each team's page lists what every recent result did to its rating, so a move can be traced result by result, and Ask answers why any team's rating has moved.
When a team is not ranked
Some teams are listed without a rank. There are four reasons, and none of them is a judgement about the team.
- Provisional
- — Too few matches on record for the rating to be pinned down. A handful of results can put a rating almost anywhere, so ranking it against sides with hundreds of matches would overstate what is known. The forecasts treat it the same way: a high rating built on a few results is read lower before anything is predicted from it, by more the thinner the record. A low one is left as it is, because pulling weak sides up made the forecasts worse when it was tested.
- Not comparable
- — No recent chain of results connects the team to the ranked field, so its rating sits on a scale that merely shares a number line with theirs. Inside club rugby this separates the provincial competitions, and Super Rugby Pacific, from the northern leagues. Each such group is ranked in a table of its own.
- No recent matches
- — The team has not played recently enough to be ranked beside sides playing now, so its rating describes it as it last played.
- Select side
- — The British & Irish Lions are picked from England, Ireland, Scotland and Wales, which are already ranked, so ranking the Lions beside them would count the same players twice. They are still rated, because the teams they tour against are rated on those Tests.
This ranking and World Rugby's
The rankings page shows World Rugby's official position beside this model's. They are different systems built for different jobs — theirs is the sport's official standing, this one exists to feed a simulation — so where they disagree, the reason is usually in how the two are built.
| Aspect | This model | World Rugby |
|---|---|---|
| Engine | Glicko-2. Each team carries a rating, an uncertainty around it, and a volatility describing how erratic its results are. | A points exchange, derived from Elo. Whatever one side gains, the other loses. |
| Built from | Every senior men's international on record, replayed in date order from the 1870s. The only thing carried in is a starting rating for 25 established nations, set by tier. | Every Test since the system launched in October 2003, starting from the ratings allocated then. A new member union enters on 30 points. |
| Margin of victory | Scales with the log of the winning margin, damped where the winner was already far ahead. A forty-point win moves a rating further than a sixteen-point one. | A win by more than 15 points multiplies the exchange by 1.5. Past that threshold, the size of the win makes no further difference. |
| Match importance | Every match weighs the same. Competition effects are measured separately and applied when a fixture is predicted, never written into the rating. | Rugby World Cup matches exchange double points. Everything else weighs the same, friendly or competitive. |
| Home advantage | The home side is handicapped inside the update by an amount measured from the record, so what comes out describes neutral-venue strength. | The home side is handicapped by 3 rating points, for the same reason. |
| Uncertainty | Published beside every rating, and load-bearing: it decides whether a team is ranked at all, and how far one result can move it. | Not published. A new union's rating is provisional until it has played ten Tests, and the size of an exchange does not otherwise depend on how well known a side is. |
| Scale | Glicko-2, rescaled to 1–99 for reading from a span of −700 to 2,800 internationally. The span covers the ranked field's own history, so the rescale is a change of unit and nothing more. | 0–100 by design, with the leading sides above 90. |
| Who gets a rank | Every rated team is listed; it is placed unless one of the four reasons above applies. | Every member union is placed. |
World Rugby's side of that table is from its own explanation of the rankings.
Why the two can differ
- How much of a win counts
- — Beyond fifteen points World Rugby stops reading the scoreboard and this model keeps reading it. A side that wins narrowly and often tends to sit higher in the official table; a side that wins by fifty tends to sit higher here.
- What a match is worth
- — A World Cup pool match moves an official rating twice as far as a July tour match does. Here they count the same, so a good or a bad World Cup shows up more sharply in the official table than in this one.
- How far back the record goes
- — Every official rating traces back through an unbroken chain of exchanges to the allocations made in 2003. This model rebuilds every rating from the matches themselves each time it runs, so a change to the record — a missing fixture recovered, a duplicate removed — is reflected all the way through rather than only from here on.
- How confident the system is
- — Two teams on the same rating are not equally well known here, and the same result moves a rarely-seen side much further than a settled one. The official exchange does not vary that way.
- What the ranking is for
- — An official standing has to place every member union. This one declines to place a team the record cannot pin down, which is why dozens of sides carry an official rank and no rank here.
Neither table is a correction of the other, and the official ranking is never an input to anything on this site. That is what makes the comparison worth publishing: two orderings built from the same results by different rules can disagree, and one built from the other could not. Near the top of the table they agree almost exactly, which is a check the model could have failed.
Monte Carlo simulation
A single simulation plays out every match in the tournament using the teams' current ratings. Each match result is randomly sampled from the predicted probabilities, so the outcome varies each time. After all matches are played, standings are computed with bonus points and tiebreakers.
This process is repeated (typically 5,000 times). The final output reports how often each outcome occurred across all simulations. For example, if Ireland finished first in 3,200 out of 5,000 simulations, their Win % is 64.0%.
A fixture that has been played is “locked in”: every simulation replays its real score instead of drawing a new one, so the forecast is conditional on everything that has actually happened and narrows as the tournament goes on. Results are matched to fixtures by the two teams involved and the window the tournament is played in.
One consequence to keep in mind while reading a table: a locked fixture's probability is 100% for the result it had. That is not the model having been certain, and it is why a tournament's Results tab does not grade the forecast — scoring a prediction means predicting from what was known beforehand, which the model performance page does across every match on record.
What the model reads
A forecast is built in two layers. The rating comes first and carries most of the weight: it already contains everything a team has done, at the strength of the opposition it did it against. A machine-learning layer then sits on top of it and adjusts, reading 28 features — the rating gap, recent form weighted by the quality of the opposition, the head-to-head record between those two sides, rest days since each last played, and which way each side's form is trending. Home advantage sits in the rating layer beneath it, outside the 28.
The win probability is adjusted by gradient boosting, and so is the club margin. The international record is about 9,313 matches, which is a small table of numbers by the standards of machine learning. Tree-based models handle that well: they cope with features on completely different scales, they do not assume any relationship is a straight line, and they handle a missing value without inventing one to fill the gap. A neural network wants far more examples than this sport has ever produced. The international margin is the exception: a plain ridge regression adjusts it, because that measured better there.
The first question is whether the machine learning earns its place at all. Trained before a cutoff and scored on 1,397 matches it had never seen, it calls four fewer winners than the rating alone, 1,074 of them (76.9%) against 1,078. What it improves is the probability attached to each call, where log loss falls from 0.496 to 0.492, and the margin, where the typical error falls from 22.1 to 21.2 points. It does nothing at all for predicted totals. That is the size of it, and it is why the adjustment is capped: the model nudges the rating, it does not overrule it. The current figures are on model performance.
The second question is which learner should do it, and it turned out to matter little. In August 2026 six alternatives were tried against the gradient boosting that ships — plain regularised regression, and five other tree-based methods — with everything else held identical. The two strongest were then run head to head over about four thousand matches, each one predicted by a model trained only on the matches before it.
They finished level. On international rugby they called the same proportion of winners to five decimal places, and in each half of the app the four measures split two apiece between them.
Once the rating has done its work and the adjustment is capped, the choice of algorithm is close to a free one — which is the same lesson as the section below, arriving from the other direction.
Why so simple a model works
The obvious objection to all of this is that the model is too thin. It knows almost nothing a knowledgeable supporter knows, such as who is fit or who has just changed coach, and it still calls about seven matches in ten. The reason is the single idea the whole design rests on.
A rating is not a measurement of one thing. It is a running summary of results, and a result is the joint product of everything a team is: its depth, its coaching, its set-piece, its conditioning, the players it can call on. Anything that persistently helps a side win shows up as a persistently higher rating, without ever having been named as a variable. The rating does not know Ireland have a good lineout. It knows Ireland keep beating people, which is the thing the lineout was supposed to explain.
That makes the test for a new input much sharper than “does this matter?” Almost everything matters. The question is whether it tells you something the results have not already told you, and whether you can know it before kick-off. Most of what a supporter would reach for fails one of those, and usually the first.
Home advantage is the exception that shows the rule. It is deliberately not in the rating, because it is a property of the fixture rather than of the team — so it is applied separately. Internationally: Playing at home is worth 120 rating points, or roughly 6 points of winning margin. It counts only when one side is at home. To a club it depends on the competition: measured from each one's own results, the Top 14's is well over twice the NPC's. A team's rating travels with it; the ground does not.
The numbers bear the argument out. Rating alone calls 73.6% of the 9,313 international matches the backtest scores. On matches it had never seen, the machine-learning layer on top of it calls four fewer winners than the rating does, and what it adds is a better-sized probability on each call and a margin that lands closer. The second layer is real, and the rating does most of the work.
Accuracy is the weaker evidence, though. The stronger evidence is calibration across the fixture list: the rating engine is checked in ten groups running from the closest fixtures to the most lopsided, and on international rugby four of the ten fall outside their margin of error, each with the favourite winning more often than the rating said, and the other six lean the same way: the rating alone is too cautious across the whole list. The published forecast, which adds the second layer, is checked on matches it had never seen in five bands of stated confidence, and all five land inside their margin of error. Club rugby is held to the same checks, and model performance shows the groups for both halves, each with its margin of error beside it.
The cost of this design is that a summary of what has happened is slow to notice a change. A new coach, a wave of injuries, a side in transition — the rating only learns about any of it from results, so it lags a real step change by several matches. More features would not fix that, because the missing information is not in the record yet either.
The simplicity is chosen. An extra column buys some signal and pays for it in noise, and on a few thousand matches that trade is often a losing one: the noise shows up as probabilities stated more confidently than the results bear out.
Surely that matters?
Several things a supporter would expect to count are deliberately not read. Each was left out on a measurement, and where a thing was tested and dropped, the test is described below.
Team selection
A forecast does not move when a side names its XV, for two reasons.
The practical reason is that the player data describes players as they are now — caps, age, club, World Cup pedigree, kept current each week — and nothing records what they were in 2015. Applying today's values to a fixture from then would credit the match to a squad that did not exist at the time, which would flatter the model against its own history.
The substantive reason is that it was tested directly. For 795 players with enough matches both in and out of the matchday squad, the difference their presence made was measured against what the team's rating expected. Not one of them stands out beyond what chance produces across that many comparisons — the working is on the squads page. Players matter, and a squad's strength is already inside the rating, because the rating was built from the results those players produced.
Weather
Weather has a real and measurable effect on scoring. Historical conditions are on file for most grounds the record can place, and when this was measured in August 2026, across about three thousand internationals carrying both a reading and a score, wind cost about 1.6 points off the match total per 10 km/h while temperature added about 2.1 points per 10 °C. Both are measured against the era the match was played in, without which the long rise in scoring is attributed to the weather instead.
It is still switched off, because the effect was fed into the model end to end and the forecasts came out worse. Scoring patterns have changed enormously since 1871, so the model weights recent matches far more heavily than old ones — which leaves a much smaller effective training sample than the raw match count suggests. Three more columns cost more in added noise than they return in signal.
There is a second reason that would bite even if the first did not: nothing can supply the weather for a match six months away, and a tournament forecast is made months ahead.
Altitude
Also absent, and internationally the reason is that the record is too thin to settle it either way. A ground the record can place carries a terrain elevation, and in November 2026 only 203 internationals on record had been played at 1,200 metres or above, and 90 at 1,500 — a sample small enough that an effect of the size usually claimed for altitude cannot be told apart from chance.
The club record can carry the same question. Club grounds are placed and hold elevations, and at the same date 507 club matches had been played at 1,200 metres or above and 164 at 1,500 — a larger high-altitude sample than the international one. Nothing has measured it, so nothing is claimed from it here.
Note that none of this contradicts the folklore about touring sides at Ellis Park. Whatever advantage a side has at its own altitude is already in its rating, having been earned in results at that ground. What is not established is a separate effect on top of that, and this record is not large enough to establish one.
Possession, territory and tackle counts
Match statistics are published on each team's profile and are never an input. The first reason is timing: they are products of the match rather than facts available before it. Nobody knows who will win possession until they have won it, so a statistic like that cannot help predict the result — it can only describe it afterwards.
The obvious repair is to use a team's historical averages instead, and that mostly restates the rating. A side that wins collisions is a side that wins matches, and the rating already counts the wins. Coverage is also partial: World Rugby publishes statistics for about a fifth of the international fixtures on record, concentrated on the major competitions.
Long-haul travel
Distance travelled and time-zone shift were computed for 99% of the match record as it stood in August 2026, and offered to the model. The winner call did not move, and the margin error came out slightly worse, so neither was kept. All the model carries is a crude flag for which hemisphere each side comes from, and it barely registers.
Home advantage itself is very much used, and is one of the larger factors in any prediction. It is the long-haul part specifically that does not survive measurement.
News, rumours and official rankings
The news and video pages are for reading, and nothing on them reaches a forecast. World Rugby's own ranking is published beside this model's on the rankings page, and is deliberately not an input of any kind. The value of showing it is that the two can disagree, which they cannot do if one is built from the other.
Reading a probability
A published probability carries real error, and the model performance page is where that is measured rather than asserted. Three things worth knowing when reading it:
- Accuracy
- — The share of matches whose winner was called correctly, over the whole match record. That record starts in 1871, and the professional era is far higher-scoring, so an all-time margin error is not the error to expect on a modern fixture. Each figure names the range it covers.
- Calibration
- — Whether a stated 70% happens 70% of the time. A gap only counts as a fault if it is larger than the margin of error of the group it sits in, so that margin is shown beside it. The same check is repeated across the fixture list, because a model can be right on average and still wrong on close games or on mismatches.
- Log loss and Brier score
- — Two ways of scoring probabilities rather than calls. Both are better when lower, and both punish a confident wrong answer harder than a hesitant one.
The headline accuracy figure comes from a walk-forward test: every match is predicted from ratings that stop the day before it, so no forecast has ever seen its own result. It scores the rating engine alone. The holdout table below it scores the blended forecast that is actually published, on matches the model was never trained on.
Reading the standings table
The standings table is how the competition is expected to finish. Each team's figures are its average across all simulations, written in whole matches and points: a side that wins in a little over half of its simulated matches is shown with the nearer whole number of wins rather than a fraction of a game. The rounding is done across the whole table at once, so every row adds up the way a real table does, wins balance losses, and the order is still the expected one.
- Win %
- — Probability of winning the tournament (finishing first overall).
- P / W / D / L
- — Matches played, won, drawn and lost.
- PD
- — Points difference: points scored minus points conceded.
- BP and Pts
- — Bonus points, and competition points under that competition's system (see Scoring below). Points are four for a win and two for a draw plus bonus points, so the row adds up. Bonus points absorb what the rounding leaves over, so they can be up to two away from the average.
- GS %
- — Grand Slam probability — the chance of winning every match in the tournament. Only shown for the Six Nations.
Position distribution
Below the standings is a heatmap showing what percentage of simulations each team finished in each position. Darker cells indicate a higher likelihood. This reveals the full range of possible outcomes — a team might be most likely to finish 2nd but still has a realistic chance of finishing 1st or 4th.
Pools and knockout brackets
A World Cup is simulated the same way, with one difference: the pool stage decides who enters the bracket, and the bracket is then played out tie by tie. Every simulation produces its own set of qualifiers, so no knockout tie has fixed participants — who plays whom is an outcome rather than a fixture.
That changes how three parts of the page read:
- Pool tables
- — The numbered columns are how often the team finished in that place in its own pool, not in one overall table of every entrant. Pools are listed in expected finishing order.
- How far each team gets
- — The share of simulations in which the team reached each round. 3rd is winning the bronze final, which is a branch off the semi-finals rather than a step towards the trophy, so it does not sit between the semi-final and the final.
- Bracket ties
- — Each tie lists the teams that turned up in it. The first figure is how often the team occupied one of the two slots; the second, shaded darker, is how often it occupied a slot and won. The second figure can never exceed the first.
The Likely path view draws the same bracket as a conventional tree. To do that it has to fix one reading — each tie settled in favour of whichever side comes out of it most often — so it shows one likely path rather than a likely bracket. The odds against any exact draw are long.
Two details of a knockout are conventions rather than published rules. A tie that simulates as a draw is settled by sampling the two win probabilities renormalised, standing in for extra time and kicks. And where the published bracket allows more than one way to seat the four best third-placed qualifiers, the first eligible seating is taken — World Rugby has not released the table that resolves those cases. It moves individual paths, not the shape of the draw.
Match predictions
Each match card shows:
- Average score — where the simulated results are centred, not a scoreline the model expects
- Probability bar — the blue segment is the home team's chance of winning, the orange segment the away team's, and the grey segment between them a draw. The draw probability is shown beneath when significant.
- Margin chart — how far apart the two sides finished across the runs, positive toward the home team
The home team is always shown on the left. The three answer different questions: the bar says who wins, the chart says by how much, and the score says where those two are centred.
How a score is arrived at
Each simulated match is two separate draws. One sets the total — how much scoring the fixture produces — and one sets the margin, the gap between the sides. The two are drawn independently and then split down the middle: the winner takes half the total plus half the gap, the loser the rest.
So a seven-point margin is not a scoreline. Against a low-scoring afternoon it is 17–10; against an open one, 27–20; against a shootout, 59–52. That is why the margin chart carries a note about the range of totals underneath it — one bar of it covers every one of those results.
It also means moving along that chart moves both sides. Two points of extra margin is a point onto the winner and a point off the loser, not two points onto one team.
Why the average score is a score nobody expects
The gap shown is an average over both outcomes, and it describes neither of them. Take a fixture where the home side are around 60% to win. When they win, they typically win by something like 17 points; when they lose, they typically lose by something like 12. Averaging a comfortable home win against a comfortable away one leaves a gap of about seven — a margin that describes the fixture as a whole and not either way it can go.
The scoreline that comes out of it turns up in roughly one simulated match in a thousand. No scoreline does much better, because results are spread far too widely for any single one to be likely, and across the fixtures checked the commonest of all sits under 1%. A predicted score is a centre of gravity: useful to know, and a different thing from a prediction of the result.
Read the probability bar for who wins and the margin chart for how close it will be. The score is there to give those a scale.
Not every fixture on the front page carries a tournament forecast. A July tour or a domestic cup is not simulated as a competition, so those fixtures show a one-off prediction instead where both teams are covered — the same answer the match predictor gives. No history is kept for it, so it has no movement to show. That is a limit of what is recorded rather than of what could be: a one-off would move for the same reason a tournament fixture does.
What moves a forecast
Mostly the two sides' ratings, drifting as they and their opponents play other matches. The prediction reads the two ratings, the venue, and form, head-to-head and rest figures built from the same results. It does not read team selection, which is switched off, or weather. The date it is measured from is the day after the newest match on record, so a day with no results moves nothing at all.
The other thing is the Monday retrain. The adjustment that sits on top of the rating is refitted every week on the record as it then stands, and that moves a forecast a little even in a week nobody played — typically by less than a point. It is fitted five times over and averaged, which keeps that small.
Which means nothing is learned about a fixture itself as it draws closer. There is no team news to arrive and no conditions to firm up. If neither side plays between now and kick-off, the forecast should barely move, and outside the Monday retrain it does not.
So a moving line is quieter than it looks. Measured across the snapshots on file in August 2026, the middle move was under a point of win probability. The large jumps were mostly the model itself changing — a retrain, a recalibration, a fixed bug — and the largest, eighty-five points inside half an hour, was a correction to which side was at home. Updates records what each run changed, which is what tells the two apart.
What if
The What if tab answers a conditional question: given that a match goes a particular way, what happens to everything downstream. Nothing is re-simulated for it. Each simulation's results are recorded, so assuming a result means keeping the simulations where it happened and reading the outcome off those.
That has two consequences worth knowing. The figures are exact rather than approximate — they are the same runs behind the published forecast, counted a second way — and they always agree with the headline numbers when nothing is assumed. But a combination the simulation never produced has no answer at all: pin five unlikely results together and there may be a handful of surviving runs, or none. The number of runs every figure rests on is shown above them, and the page says when that number is thin.
A scenario lives in the address bar, so a link carries the assumptions with it.
Scoring system
Most competitions here use the standard World Rugby bonus point system:
- Win: 4 points
- Draw: 2 points
- Loss: 0 points
- Losing bonus: +1 point for losing by 7 or fewer
- Try bonus: +1 point for scoring four or more tries
Bonus rules are read per competition. The Top 14 awards its losing bonus at five points. The Rugby Championship, the Top 14, Super Rugby Pacific and both European cups award the attacking bonus for scoring three more tries than the opponent, where the others ask for four.
Tries are simulated alongside the score in international rugby, from the relationship between the two in matches where both were recorded. In club rugby the tries of a played match are on file — counted from ESPN's play-by-play, or taken from World Rugby's match statistics — but no club try model has been fitted yet. So a club match already played awards the attacking bonus on the tries recorded for it, where there are any, and one still to be played awards it on a points threshold. Club tables carry no tries column, because a total counting only the matches already played would understate every side.
Tiebreakers are applied in this order:
- Tournament points
- Point differential
- Points scored
- Head-to-head record
Both European cups and Super Rugby Pacific separate level sides on wins first, and their tables are ordered that way. The NPC puts the head-to-head result ahead of points difference, which is not modelled, so an exact tie there follows the order above. That moves a placing in a few simulated seasons and rarely a qualification.
Results
The Results tab lists the fixtures of an edition that have been played, and scores each one against the forecast it had beforehand. That forecast is recorded while the fixture is still to come and never rewritten afterwards: once a match is played every simulation replays its real score, so the probability published beside it becomes 100% for what happened, and scoring that would only ever agree with itself.
A fixture played before that record began cannot be scored and says so. A handful of matches is too small a sample to read as an accuracy rate in any case — model performance measures the same thing across every match on record.
Where the data comes from
Every source behind these numbers is listed on about and sources, along with what this app is and is not.