Supercomputer updates Rangers & Celtic’s title chances after pivotal weekend

Start with what happened over the weekend, because the context explains why the probabilities moved. Rangers arrived at Dens Park having won seven consecutive matches, and a victory would have taken them to the top of the table, at least until Celtic played on Sunday. Instead they drew 2-2, dropping points for the first time in eight games, after errors in defence and midfield allowed Dundee to score twice. Then Celtic produced one of the most emphatic results of their season, beating Hearts 7-1 at Celtic Park, with Callum Osmand reported to have scored twice. That result did more than add three points to Celtic’s total. It also wrecked the hopes of the club that had been expected to be Rangers’ main rival, and it highlighted the strength of a team that Opta still regards as the clear favourite.

The probabilities themselves are worth unpacking. A 74.88 per cent chance for Celtic means the model expects them to win the league in roughly three out of every four simulated seasons. Rangers’ 22.92 per cent works out at around one in four, or slightly less. Hearts’ 2.14 per cent is a long shot, and it is notable that the three figures add up to 99.94 per cent, which leaves almost nothing for any other club. That tells us that the model sees the title as a straight fight between the Old Firm, with Hearts clinging to a faint hope. Before the season started, Opta had put Celtic at 55.36 per cent, Rangers at 26.35 per cent and Hearts at 11.86 per cent. Hearts have therefore lost nearly all of their chance, Celtic have gained a great deal, and Rangers have drifted down only slightly.

That comparison is a useful corrective to the gloomy tone that often accompanies a dropped-points headline. Rangers began the campaign with a 26 per cent chance and now have 23 per cent, which is a modest decline given everything that has happened in between. It is also worth remembering how low their figure fell during the early weeks of the season. After a slow start under Derek McInnes, in which the club collected one point from its first two league games and lost at home to Hibernian, the picture looked bleak, and another model rated their chances at just under 17 per cent at that stage. Four weeks ago, after Rangers beat Celtic 1-0 at Celtic Park, Opta put their chance at 9.95 per cent, which is strikingly low for a team that had just won the most important match of the season so far. By that measure, the present figure is a considerable improvement.

Why would a model give Rangers such a low chance after a big win over their rivals? The answer lies in how these systems work. A supercomputer of this type does not simply look at the table. It estimates the strength of each team using a mix of results, underlying performance data and market information, and then simulates the rest of the season thousands of times, drawing random outcomes for each remaining fixture according to those strength ratings. The title probability is simply the share of simulations in which a given team finishes top. That means the model is heavily influenced by points already banked, by the quality of the remaining schedule and by the gap in underlying strength between the contenders. A team that starts the season badly has to win more often than expected to catch up, and the model reflects that.

This explains one of the most striking details in the update. Based on expected probabilities, Opta projects Rangers to finish the season on 66 points and Celtic on 72. By comparison, Rangers collected 72 points last season and finished third, while Celtic won the title with 82. The model therefore expects Rangers to gain fewer points than they did last year, which may sound surprising for a team that has just won seven games in a row. But there are good reasons for it. The early-season stumbles have already cost points that cannot be recovered, and the model does not give heavy weight to short winning streaks. It tends to regress performance towards an underlying estimate of team quality, and for Rangers that estimate is lower than the recent results might suggest.

The projection for Celtic is just as interesting, because it points to a very different kind of season from the one they enjoyed last year. A total of 72 points would be ten fewer than the 82 they achieved as champions, and it would be a modest return for a team that has dominated Scottish football for so long. That implies that the model sees a tighter league, one in which the champions are not able to run away from the pack. It also suggests that Celtic are not expected to be quite as strong as in previous seasons, which fits with a campaign in which they have already shown vulnerability, including a defeat to Rangers in September. The fact that the model still gives them a 75 per cent chance reflects the size of the gap in expected points, not an expectation that they will be flawless.

For Rangers, then, the gap of six projected points is the real story. A season in which they reach 66 points would likely leave them second, but it would not be enough to win the title unless Celtic collapsed. To have a realistic chance, Rangers need to outperform the model’s expectations by a considerable margin, and that means winning the matches in which they are favourites, and winning some of the ones where they are not. McInnes’s side have already shown that they can do that, because the winning run that preceded the Dundee draw was built on victories over a range of opponents. The question is whether they can sustain that standard for the remainder of the season, and whether the squad is deep enough to cope with injuries, suspensions and fixture congestion.

It is important to be honest about the limits of the numbers. A supercomputer cannot account for a manager’s tactical changes, a key player’s injury in January, or the psychological effect of a high-profile result. It treats every match as a probabilistic event drawn from a distribution, and while that approach is reasonable over a long season, it cannot capture the human elements that decide close races. It also depends on the quality of its inputs. If a team’s underlying strength is underestimated, the model will keep undervaluing it until enough evidence accumulates to correct the estimate. Rangers’ recent form may be a sign of real improvement that the model has not yet fully absorbed, or it may be a temporary run of good results. Only time will tell which.

The Dundee draw illustrates the point well. By most measures, Rangers were the better side over the ninety minutes, with around sixty per cent of the possession and more than twenty shots. They also made two defensive errors that cost them two points, one in the build-up to Bradley Fink’s goal and the other when Dan Neil’s misplaced pass allowed Callum Jones to equalise. A model based on expected goals might see that match as evidence of strong underlying performance, while a model based on results would simply register a dropped two points. The truth is somewhere in between, and the supercomputer’s modest reduction in Rangers’ chances suggests that it treats the draw as a small setback rather than a disaster. That seems fair.

Celtic’s performance against Hearts has a different significance. A 7-1 victory is not just three points. It is a statement of intent, and it also has a mathematical effect on the table through goal difference, which could matter in a tight race. It is also a sign that Celtic can produce overwhelming performances when they are at their best, which makes them a dangerous opponent for any team. For Hearts, the result was a heavy blow, and it explains why their title chances have fallen so far. The Edinburgh club spent much of last season at the top of the table and came close to winning the league, but the model now sees them as an outside bet. A defeat of that scale can damage confidence, and it will test the resilience of Tynecastle’s squad.

By my reckoning, the result also leaves Rangers four points behind Celtic in the table, assuming both teams have played the same number of matches. That is a manageable gap at this stage of the season, particularly with the two sides still to meet again at Ibrox in the new year. A single result can change the picture dramatically, and the model will update accordingly. But it also means that Rangers no longer have the luxury of dropping more points against teams they should beat. Every match from now on carries extra weight, and the next fixture, a home game against St Johnstone, is an opportunity to reassert the momentum that was broken at Dens Park. A win would reduce the pressure, while another draw would raise questions about the consistency of the team.

What can McInnes do to improve the odds? The most important thing is to eliminate the avoidable mistakes that have cost points in recent matches. Rangers have the quality to beat most teams in the league, but they have shown a tendency to concede goals from errors rather than from skilful opposition play. Fixing that is partly a matter of concentration and partly a matter of selection, and it will require the manager to make firm decisions about who plays and where. The attacking side of the game also needs attention, because the team has relied too heavily on individual brilliance for its goals. Bouanani’s curling finish and Raskin’s thunderbolt were spectacular, but a title-winning side needs to create chances through collective play as well as individual quality.

The January transfer window is another factor that the model cannot predict. The club’s chairman, Andrew Cavenagh, has urged patience and indicated that transfer business will be limited as the club works towards financial stability, which means that McInnes is unlikely to be given the resources to make major changes to his squad. That puts a premium on getting the best out of the players he already has, and on managing injuries carefully. It also means that any weaknesses that emerge in the first half of the season will have to be solved from within. For a manager with a reputation for resourcefulness, that is not an impossible task, but it does narrow the margin for error.

Celtic have their own challenges, though they are of a different kind. The club has a European commitment to manage, with a Europa League tie against Celta Vigo approaching and several injuries to key players, including Cameron Carter-Vickers, Carl Starfelt, Yang Hyun-jun and Alistair Johnston. Managing a squad across domestic and European competition is a demanding exercise, and it can lead to dropped points if the rotation is not handled well. The supercomputer takes some account of these factors through its strength ratings, but it cannot fully capture the physical and mental toll of playing every three or four days. If Celtic’s European campaign becomes demanding, there may be opportunities for Rangers to close the gap in the league.

It is also worth thinking about what a 22.92 per cent chance really means in practice. It is not a prediction that Rangers will lose the title. It is a statement that, in roughly one out of every four or five simulated seasons, they come out on top. Those are not bad odds for a team that began the season in difficulty and has had to recover from a poor start. They are also higher than the odds for most clubs in most leagues, and they reflect the fact that Rangers remain a serious contender. The danger lies in treating the figure as either a certainty or a verdict. It is simply an estimate of the current situation, and it will change with every match that is played.

For supporters, the most sensible approach is to treat the model as a useful guide and not a prophecy. It provides a reasonable summary of where the title race stands, based on the information available. It also provides a challenge, because it shows how much work Rangers still have to do. The gap between 22.92 and 74.88 is not impossible to close, but it will require a sustained run of results and a few slips from Celtic. The recent history of the league suggests that such slips do happen, particularly over the course of a long season. The league has been won and lost on narrow margins before, and there is no reason to think that this year will be any different.

In the end, the supercomputer’s update is a reminder that the title race is far from over. Celtic remain the favourites, and their 7-1 win over Hearts showed why. Rangers have been knocked back by the draw at Dundee, but they are still in a stronger position than they were four weeks ago, and they have the chance to improve it. The model projects a gap of six points by the end of the season, but projections are not results, and the gap will be decided on the pitch. If McInnes can fix the mistakes that cost points at Dens Park and maintain the standard that produced the winning run, the numbers will shift in his favour. If not, the supercomputer will be proved right, and Celtic will lift another trophy. That is the nature of the race, and it is what makes it worth watchingl