By Mark Asher Goodman (@rapidsrabbi)
For a guy with an awful lot to brag about, Fran Taylor doesn’t like to say much. “Fran’s kinda cagey, isn’t he?” I said to Ryan Madden, the Colorado Rapids Director of Communications. “Yup. And it’s not just you. He’s like that with everybody.”
In my infinite quest to find the best and most advanced numbers to help find the best players that would craft winning football teams, I had come to Commerce City, Colorado to speak to Fran. I had hoped that in our conversations, Fran would divulge some trade secret or some new metric that would let me (and other sabermetric nerds that read American Soccer Analysis) understand the game on a deeper level. Spoiler alert: Fran did not share the ‘one number to rule them all.’ But he did tell me a lot of really cool stuff.
Fran’s ‘cagey-ness’ and humility go hand in hand with a backstory that makes him a candidate for ‘most interesting soccer analyst in the world.’ In terms of humility, he would describe his job as nothing more than “taking things off (Rapids General Manager) Padraig Smith’s plate”. In terms of his qualities as an ‘international man of mystery’, Fran is likely the only person on the planet who played in the Laotian Soccer League and has also worked for Arsenal Football Club.
Taylor, the 29-year-old Assistant General Manager for the Colorado Rapids, joined the team in January of this year to bolster the data-driven team of Padraig Smith in trying to use numbers to build a better soccer club. While baseball and basketball teams have adopted advanced metrics to try and divulge information that will give their club a competitive edge in their respective sports, soccer is still finding its way in the moneyball era. Newly discovered numbers are illuminating the game, but only in the hands of those who know what to look for.
“They called me ‘Kao’ - Frog”
Fran’s road to Major League Soccer is unique, to say the least. After graduating from Occidental College in Los Angeles in 2009, Fran and a colleague, Jaeson Rosenfeld, created a soccer data analytics company called ‘Stat DNA.’ Fran did it less because of his economics degree, and more so because of his love for soccer, a game he played all his life, continuing all the way to a standout career for the Occidental Tigers, where he was twice named to the All-Conference Best XI as Oxy’s starting striker.
While Rosenfeld based in Chicago, Taylor went to the Southeast Asian nation of Laos in order to do the grunt work of data tagging and analytics. Why Laos? “Because if you are going to pay someone in the US to do 25 hours of work to collect data on one game, it’s going to be too expensive to be sustainable” says Taylor.
This is not Opta data. This is much more detailed stuff. Because Opta tags events in soccer in real time, you can only tag and denote so much - mostly things that happen on the ball, like dribbles, tackles, passes and shots. But what about off-the-ball movement of attacking players? What about the position of the defense when a ball carrier is in the attack? That’s where Stat DNA came in.
“What that allows you to do is have a level of accuracy particularly with timestamps, and also add additional layers of data on top. So we could do things like pressure, and the number of guys behind the ball. That allowed you to do a lot more than what Opta could do. The big difference with us and Opta, with regards to event tagging was time-stamps. If you’re going to build a model that looks at a goalkeepers ability to claim a ball in their area, you need to know the exact time of when the goalkeeper claimed the ball or when it passed through his area, and when it was delivered. Because you’re looking at how much distance a keeper can cover in any certain period of time to claim a ball in the box.”
At the time Rosenfeld and Taylor created Stat DNA, nobody was doing that kind of work. Either your ginormous mega-team was collecting data on its own, or you just couldn’t afford it. Taylor’s company started with an eclectic mix of clients, recording their games and the games of the opponents in their league or potential players they wanted scouted, and sending them to Laos. Back then, the games were send via digitally FTP to the local telecom company, where they took all night to be downloaded onto a hard drive, and then Fran would pick them up at five AM and take them to the office for the day’s work. Then the data analysts would tag all the events, and Fran would then stick all that data into a computer model to give him useful player information. Taylor and Rosenfeld had clients like the University of North Carolina women’s team and Brazilian club Botafogo, sending them tape of their own players and team as well as opponents they wanted scouted.
In his free time, Fran would play soccer in the Laotian Premier League. “They liked to make fun of me a lot and tried to convince me that any person of Asian descent was actually Lao, like Manny Pacquiao. They called me “Kop”- or “Frog”. I have no idea why. It wasn’t a very serious league- I think they paid me 400,000 kip a month or the equivalent of 50 USD.”
After a few years of watching all manner of soccer matches from around the globe, a bigger client came calling in 2013: perennial English Premier League heavyweight Arsenal Football Club. At first they were Stat DNA’s main client. Then they became their sole client. Finally, Arsenal were so happy with the results from the hard-working crew in Laos, they bought StatDNA, lock stock and barrel. That soon required Fran to relocate to Boston, and then London. Which was a bit of a change from spending his time playing in Laotian first division games on the weekend and scouting Clemson’s women’s team all week.
“I mean it was night and day.”
Getting an Education at Arsenal
Taylor spent his days at Arsenal’s facility, looking back at every movement and detail of each past Arsenal game and upcoming opponent, along with a team of sabermetric wiz kid analysts and under the direction of the Manager Arsene Wenger’s technical staff. But as much as Fran was contributing to Arsenal’s week-to-week preparation, Arsenal was training Fran about what to look for.
“When Arsenal bought us, things really changed at least for me in my role. Or at least what I was able to learn. Because there’s only so much you can learn watching footage by yourself, purely from the lens of an analytics person. Some of the things that I remember thinking about the game back then are kind of scary to me. And naive. And it took me working in a club environment with people who knew the game intimately to kind of teach me a lot of those things.”
In its essence, Fran began the important work of separating the ability to measure something with the ability to find measurable things that really mattered. That only happened because he was sitting with Arsenal coaches and watching the games, instead of sitting with office drones clicking their mouse buttons to input match events. Numbers are just numbers, and useless ones, until they can be converting into information that describes good soccer as opposed to not-good soccer. Fran described the analysis room at Arsenal as “a kind of hangout spot” for the technical staff.
“That’s where (Arsenal Assistant Coaches) Steve Bould, Neil Banfield would roll in if they wanted to blow off steam or watch parts of the game back. And then the analysts there are top notch too. Because they’re the ones that are working with the technical staff pre-match, briefing the players on the strength of the opposition.”
“I remember one time we went through all of the goals that we had conceded that season . We were kind of trying to build a grading rubric for defenders. And we’re sitting around watching all the goals we conceded. And I think we had Andre Santos on the team at the team ( left back for Arsenal, 2011 to 2013). And he was involved in some of these, and he made some pretty colossal errors for Arsenal. Having everyone sitting around there, Bouldie was there, Neil was there, some of the analysts were there; not just kind of a incredible learning moment for me where a veil was lifted of ‘holy crap there’s this whole wealth of soccer knowledge that I haven’t really been exposed to’ but also, we had a really great time just laughing about it.”
The other cognitive leap at Arsenal for Taylor was beginning to discern soccer movements as a collective ballet of players as opposed to a set of independent actions. To be able to see the difference between a discrete failure a player makes as opposed to a breakdown that occurred 40 yards further down-field.
“A lot of time like center backs will be one-v-one like in a crossing situation somewhere in the width of the goalmouth. To have that? Something must have broken down before then . You almost always want someone bracketing a center forward, someone on the near post side, and someone a touch tighter to them defending any kind of movement towards them or towards the back post or towards the goal if you don’t have that person in front, goal side, protecting a run across the defender that’s very hard to defend and usually teams allow for that and if they’re not faced with that ; if there’s a one-v-one situation it means that that is a sub-optimal defensive situation.”
This is where Arsenal’s coaches shifted the paradigm: before Arsenal, Fran was mostly looking at objective measures. Soccer event as truth. Throw data at it till it describes the thing you want to know about a player. Arsenal shifted his perspective to understanding the need for subjective analysis. Not ‘was that a tackle?’, but perhaps instead ‘was the player in the correct position to make the tackle?’ I say perhaps because Fran is a constant poker-player with specific measures and details of the ‘what’ and the ‘how’. He won’t let you see his cards, even if you slip him a blue poker chip.
“I see a lot is near-side centerbacks drift outside the near post, right? And for me, when they’ve done that, they’ve left their team at some kind of loss. But you have to evaluate ‘Why is that near-side center-back drawn out?’ ‘Was it a good decision?’ ‘Did it look like his fullback was vulnerable, was that what dragged him on?’ ‘Or did he do it because he’s not paying attention?’ ‘Or is he not aware of his role?’ You have to do a layer of subjective work before you come up with a number for that . But we came up with numbers for it.”
What that number is, Fran won’t say. He’s cagey like that.
The Next Frontier
What Fran didn’t have at Arsenal was decision making power, or as he put it “a seat at the table.” “I would be in a meeting with Arsene Wenger… twice… a year. I would come with insights, it would go to my boss. My boss would take it to Arsene. Arsene might act on it.”
So when Padraig Smith, General Manager of the Colorado Rapids, came calling with the job of Assistant General Manager, Fran jumped. “This is just a much more interesting and fun environment for me. A little more influence, a little more responsibility. And it’s a new challenge, you know? I spent eight years doing things one way. And it’s nice to vary that.”
Now Fran has the ear of the General Manager and the Coach, Anthony Hudson. When he does opposition scouting, it’s after direct consultation with the Rapids technical staff, assessing the things that have been gleaned from watching the last three or four matches, and then filling in the rest by creating data to measure something they still want to understand. Taylor will talk to the coaching staff and try to discern useful information for an upcoming game. Take, for example, a recent Rapids match against Portland Timbers. Fran could take the needs of the coaching staff plus his subjective understanding of what they want to look for to help give Colorado any edge he can find.
“There is a camera tracking system on players that records 25 frames per second. And there are things that we’ll code off of that, and then there are things that data providers will provide for us. So a lot of the ‘distance covered’ - ‘velocity’ - ‘high intensity runs’ - that kind of stuff, will be provided for us directly, and then there’ll be a lot of cool stuff that we do on our own based off of that. What’s important is the position of those 22 players provides a ton of context about the game that allows you to build metrics that are a little more true to certain soccer concepts.
Because against Portland (which often play a 4-3-2-1), if you play a 40 yard diagonal you may not have eliminated any players, whereas if they’re playing a 4-2-3-1, that might actually be different. You might have eliminated a few guys with that.”
For Fran, and for fans and statheads, there is one question that stands out as the critical one in his work: can this new world of digital tracking and big data and regression plots help a smart team to win? The answer is certainly yes, but only in the hands of a clever craftsman who can separate the meaningful from the irrelevant; the signal from the noise.
“If it’s not readily apparent why a certain metric is valuable, I don’t put a lot of stock in it.”
It has to pass the eye test of a technical director and the math test of a data nerd. There are only a few metrics that Fran trusts. Expected goals (xG) is one of them.
“Expected goals - you know what that means. It’s the volume of chances that a player gets in the course of a match. Now there can be issues with the expected goals model… not taking into account where defenders are, for example. Or pressure on the ball, or who is obstructing the shooter and goalies view of the target. It’s flawed, but in the end, you can live with that.”
Another number that Fran thinks can tell you a lot about the quality and ability of a player or players is duels; sometimes categorized into ‘tackles won/lost’, ‘Interceptions’ and ‘aerial duels won/lost’.
“I think ball-winning stats tell you how a player is at ball-winning. It doesn't tell you how good a player is defensively, but if a player is intercepting a lot of passes, it tells you he is at least reading and anticipating passing angles. If you have a large enough data set, if someone has a lot of tackles it tells you someone is actively going and getting about the pitch and trying to break up play. You can look at their challenge success rate, which is about their physical capabilities and the angles they’ll take into a challenge. So those I put some good weight on.“
Taylor knows, despite all the thinking and analysis he has done over the past eight years of soccer players trying to stop an attack, that defense and the ability to find a clear and consistent measure of good defending is still yet to be unlocked.
“That’s beyond the next frontier. The level of programming and development that would need to go into actually being able to evaluate a defenders decision-making based on x-y coordinate data would be... it’s a huge problem.”
“I’m very good at where the art form of soccer and the numbers intersect, but at some point when the numbers component is too difficult to understand I can no longer bring those two things together. I don’t think it’s impossible. Anything a human understands, a computer can eventually understand. But we’re not quite there yet. Computers can drive cars. They can’t evaluate defenders. Yet.”
The added piece to this is the elephant in the room of this article: the Colorado Rapids are not a good football team this year. An enraged fan or old-school soccer Luddite could easily decide that this advanced metrics stuff is all voodoo and nonsense, since Smith and Taylor used it to build a Colorado Rapids team that are currently 21st out of 23 MLS teams.
But it would be best to give this team some slack, and a little bit more time. Smith’s only been GM for eight months; and Fran has only been in Colorado since January. Getting the right players, and the right system, and the right metrics that help discover those right players for that system will take a little time. Rome wasn’t built in a day, or as Taylor puts it “I hate to quote (Sixers GM) Sam Hinkie, but there’s an element of ‘Trust the process’ that’s important here in what we’re trying to do.”
The next frontier for an assistant GM is, of course, to become a General Manager. Fran isn’t ready to make that leap, at least yet anyways. “I’ve been in this role (Asst. GM) for five months. I need to prove to myself first that I can get to the next level. You know?” If Fran continues to help analytics progress to the next level, then perhaps analytics will propel Fran to the next level, too.
Mark Asher Goodman writes for Around MLS and Pittsburgh Soccer Now. He primarily covers the Colorado Rapids and the USL’s Pittsburgh Riverhounds SC. When he’s not writing about soccer or watching soccer, he’s a rabbi. No, really, like he’s an actual rabbi. You can find him on twitter at @rapidsrabbi and @riverhoundrabbi. His last article for ASA was ‘The Elusive Advance Defensive Metric’, which you might also like.
Photographs care of John Babiak, @Photog_JohnB