“Many clubs operate the way Liverpool did before 2012 and that is fine,” Ian Graham tells Sky Sports. Fine because he believes that works too? Not quite. “You always need a sucker in the market. If everyone is using data brilliantly, I am out of a job, aren’t I?”
Graham was Liverpool’s director of research for over a decade before leaving last year. During his time, they won the Champions League and finally lifted the Premier League trophy. It was all achieved with the help of their in-house analytics department.
His ‘possession value’ model, designed to calculate how much each player improves their team’s chances of scoring with every touch of the ball, was revolutionary. It was part of Liverpool’s recruitment strategy when signing the key men behind their success.
Jurgen Klopp leaned towards Mario Gotze. The data favoured Sadio Mane. When Julian Brandt, another talented German, was Klopp’s first choice in 2017, the club convinced him to agree to the signing of Mohamed Salah instead. The rest is Liverpool history.
“We had missed a couple of opportunities when we sold Luis Suarez and then Raheem Sterling. That money should have been reinvested differently but I think we learned from those lessons. When we eventually did it correctly, that is what led to the success.”
In conversation with Sky Sports, Graham, who has written a new book entitled How to Win the Premier League, expands on some of these stories and reveals some of those lessons learned from helping to embed data into Liverpool’s decision-making process.
You need buy-in from the coach
Before Klopp, there was Brendan Rodgers. His desire for control over recruitment led to an uneasy relationship with Liverpool’s fledgling analytics department. Those were the days when the club’s notorious ‘transfer committee’ found itself ridiculed in the media.
As Graham explains, the problem was not his work but that his work was ignored. After some difficult experiences with Harry Redknapp at Tottenham, he found Rodgers reluctant too, insisting on the signings of Joe Allen and, frustratingly, Christian Benteke.
“I think with younger managers in the Premier League, because they have grown up with data earlier in their careers, we are seeing that culture slowly change. What makes Jurgen different is that the older managers did not often think the same way he did.”
“It was not so much that Jurgen saw some data analysis and thought it was brilliant. The owners explained the structure and he was open-minded enough to appreciate that different viewpoints can only add to the quality of the decision-making.
“Jurgen had more control than he would at a German club, but he was very used to that collaborative way of working. Whereas the previous manager, the traditional manager model in England, was of the mind that they had to be in sole charge of all decisions.”
Try to be certain the style fits
Benteke struggled at Liverpool, just as Graham had anticipated he would, not because he was a terrible player but because his style of play was so ill-suited to Liverpool. This was a target man being pursued by a club that did not really require a target man.
“The most frustrating times were when a player looked good but we just did not play that style. For example, a player adding a lot of value through crossing does not work at a team that does not have anyone to receive those crosses. That is the style argument.”
Don’t just make the stats fit
“There is this great quote by Marilyn vos Savant that says statistics can be used to support or undercut any argument,” says Graham. “And that is definitely true in football data.” One of the beauties of Graham’s possession-value metric was its clarity.
In a world with so many such metrics, Liverpool knew which one mattered to them. The key was not to be seduced by the other stats available. “A player you like will look good on one of the metrics even if they are a really bad player. That is the danger,” he explains.
“You can kind of retrofit the statistics and say, ‘Well, this metric that I have only just heard of is the one I really care about if the guy I like looks good on it.’ I have seen that happen at a couple of clubs where on the surface it does appear they are using data.
“But they are not using it in an honest way. If you are only using data to support what you thought anyway, you do not need it. If it is just giving the same view as your video analyst or the eye of the coach, why bother if you are only going to use it to support that view?”
Players need to play to improve
“There were never any good left-backs or left-sided centre-backs on the market. They were always in short supply.” But Liverpool solved the problem on the left side of their defence when they were able to partner Virgil van Dijk with Andy Robertson from 2018.
“When a great left-sided defender came up the following season, we knew he was not going to play for us so we could not really sign him. That was really frustrating. In any other season, we would have bought them but that year we did not need them.”
Eventually, Liverpool became so strong that they were recruiting for squad depth. “You are spending money on an insurance policy, basically. Most teams would kill to have that problem.” But when building that great team, there had been certain principles.
“The owners were happy to spend but they wanted evidence that money would lead to better performances. So, the first rule was if we are going to spend on a player, they have to play, they have to be on the pitch in order for them to make a difference.
“The way that young players improve is through relevant minutes playing against decent quality opposition. And so you want to be able to be guaranteeing at least 1500 minutes to your young players every season. That is all about squad management.”
Parking players will not develop them. Will that be a problem for Chelsea? “I cannot speak to what Chelsea’s plan is but I think they have got a large number of extremely good young players. It is how that squad is managed that is the question,” says Graham.
“When you sign a young player you either want performance today or you want to lock in that young player’s improvement, so that by the time they are 25, they have got a chance of being a world-class player. If they are not playing, they just don’t get that opportunity.”
Sign players at the right time
Some young players will be ready quicker. Klopp called off the search for a right-back when he worked with a teenage Trent Alexander-Arnold in training, allowing him the space in the squad in which to grow. Others are helped by arriving at the right moment.
Liverpool mastered that skill, buying many of their future superstars just as they were about to hit their peak years. “I think it was Julian Ward who came up with the phrase that they need to have a career before having their Liverpool career,” says Graham.
“If you look at the successful players that were signed, they were 23 or 24, maybe one or two 25-year-olds. All of the big successes were in that age range. The risk versus reward trade-off meant that we were happy to pay a higher price for a lower risk player.”
Graham cites two examples. “We were beaten to Mo Salah by Chelsea in 2014 but the player we got in 2017 was much closer to the finished product.” He wanted to sign Mane straight from Red Bull Salzburg too. “I was always on the risky side,” he admits.
“Southampton paid £12 million for Mane. Two years later, we paid Southampton £30 million and we were supposed to be the smart team! But that difference in price was a lower risk because Mane had already proven he was a very good Premier League player.”
Even the stats don’t tell you it all
“In the early days when other people in the industry asked whether Liverpool used data to sign players, the answer was no. Liverpool used data to help sign players. Michael Edwards and the scouting department were curious and asked very difficult questions.
“Especially for centre-backs, where off-ball positioning is really important, we made the best estimate that we could using the information that we had. But we would admit that it was not perfect and it was really important that they understood it was not perfect.”
They were operating before tracking data had been integrated into their model – and there were always details that showed up in the video that could not be seen in the data. In the case of two high-profile signings, data would later reveal they were even better.
“Van Dijk was signed before we had an operational tracking model. Tracking data gave us more confidence to rate players because you could see the off-the-ball positioning.” That information confirmed what they already suspected – Van Dijk was exceptional.
Alisson Becker is another. “We could not look at Alisson using our goalkeeping model because we did not have tracking data for Serie A games, where you could see exactly what the goalkeeper’s positioning was doing to the expected goals,” says Graham.
“Fans would say, ‘I could have saved that.’ But you couldn’t. A better goalkeeper can make a save look routine, whereas a worse goalkeeper whose positioning is slightly off or whose reactions are slightly slower, they can make a save look spectacular.
“There were a couple of goalkeepers who we said were good shot-stoppers and we were right. But they did not have to be good shot-stoppers if they were sensible with their positioning. They just put themselves in trouble and were good at getting out of it.”
Who is using data well now?
As Graham learned over a decade ago now, clubs could be doing good work but find that it is not impacting the decision-making. He suspects that Barcelona’s analysis is of high quality but has had little influence on some of the players they have actually signed.
Maybe something has changed at Arsenal? “You can never tell from the outside, but I think they had really good analysis. Arsene Wenger probably didn’t use it that much. In recent seasons, Arsenal’s signings have been very much in that Liverpool mode.
“I think they are reaping the benefits of that. Again, I do not know for certain whether it is data analysis that has caused Arsenal’s recruitment to go from questionable to very good, but I would guess that it is involved somehow. I think things have changed there.”
A team that he is more sure of is Brentford – because their signings had often already been identified as potential Liverpool targets. “Whenever we picked up on a player who was 21 or 22 in a second-tier league, it was a case of, are Brentford going to sign them?”
Graham adds: “Brentford did sign some players who we did not know about but they signed a lot of players that did not surprise us. It was great to watch Brentford from the outside. I was sort of a bit jealous not to be signing those really, really unknown players.
“Liverpool helped to show it could be done at the highest level, but I think Brentford and Brighton certainly showed that data could be used to take clubs that did not have particularly high budgets at the time, from League One to the Premier League as well.”
What is next for data?
Graham is now consulting with a number of clubs, exploring various new ideas. “We did something on predicting positioning at corners,” he reveals. Even now, not everyone wants to pay for it. “Data is a big investment for clubs,” he acknowledges.
“Typically, all revenue goes on player wages and transfer fees, and there is not too much money left to invest in other things. But I think the case has been made that it will save you money in the long run. Not making mistakes is the key and data helps with that.”
He is intrigued by the potential of artificial intelligence but, despite offers, less keen on repeating what he calls the Liverpool trick. “I have kind of done that already.” One suspects there are still enough suckers in the market for him to succeed if he so chose.