From Microsoft to the MLS: Ravi Ramineni crunches data to help the Sounders win

GeekWire

Having worked at Microsoft for seven years on projects like Bing and fraud management, Ravi Ramineni understands how to extract relevant information from massive datasets. He also happens to be a huge soccer fan and spent much of his down time away from Redmond analyzing soccer-related data and analytics on his blog AnalyseFootball.

So it makes sense why Ramineni calls his current role with the Seattle Sounders FC a dream job.

The 38-year-old is now a sports scientist and performance analyst for Seattle’s professional soccer team, a career he took on after leaving Microsoft in late 2012. The data geek was recruited by Sports Science and Performance Manager Dave Tenney, who had already spent three years with the Sounders tracking player performance but needed technical help figuring out how that data could be better used to ultimately win more matches and keep players fresh.

ince joining the Sounders three years ago, Ramineni’s role has grown each season. The team is already widely-known for its innovative utilization of wearable devices, but Ramineni has helped it crunch and visualize the data more effectively to help the coaching staff answer more questions.

This has arguably helped the Sounders gain an edge, as the team finished with the league’s best record last year and notched the most regular season wins (20) in franchise history.

“It’s taken some time, but I’ve shown I can help them,” said Ramieni, who moved to Seattle in 2007 from India to work for Microsoft. “I was able to slowly build trust and now I can answer questions from coaches, who are trusting the numbers and also me.”

We stopped by the Sounders practice facility in Tukwila, Wash., last week to meet Ramineni, who attends every practice with a laptop that monitors live tracking data from the devices the players wear. Read on for edited excerpts from our conversation with Ramineni, who shared more insight into his day-to-day work with the Sounders.

GeekWire: Thanks for meeting with us, Ravi. I noticed you standing on the sidelines at practice today with your laptop and a few gadgets. Tell me more about your day-to-day work here.

Ravi Ramineni: “The guys are wearing two types of devices: A heart rate monitor on their chest and a GPS receiver unit that sits between their shoulder blades. The heart rate monitor tells us how hard a player is working. The GPS tracks distance covered in different velocity zones, total distance covered during a session, and speed. It give us about 300 to 350 metrics per player, per session. It also tells us a lot about accelerations, decelerations, change of direction, those type of things.

We know what the demands of the match are, and the trainings are tailored to make sure they get enough stimulus on what they have to do in the game. For example, midfielders have to do a certain amount of high-speed running in a given session and have to cover a certain amount of total distance.

I have my computer on with my antennas just to see what the players are doing in real time and how hard they are progressing. There are always some guys rehabbing, and for them we have specific targets to reach certain max speeds, total distance, etc. We check every now and then during a session to see if they reached those targets. This also happens with healthy guys — we want to see them hit certain numbers. If they don’t, Dave [Tenney] will take them out and do extra work or specific drills with them so we get what we want.”

GeekWire: That’s cool. What do you do when practice is over? What do you do with the data?

Ramineni: “We collect all this data on the field and my job after that is to bring back all the units and download the data into my computer and then process it. We try to be very precise in terms of every specific training drill, whether it’s playing full field 11-on-11 or something else, to really figure out how much work it was. I break the data down based on specific drills. I also make notes while I am watching practice — sometimes people get hurt and sit out three minutes and go back in. These are things you need to note down so I can come back and say, OK, he only played from 11:31 to 11:42 and after that, something happened.

This is important when you talk about taking out the noise. When I’m watching practice, I know exactly who did what and how big the size of field was for a certain drill. Sometimes I measure the grids myself. All this makes the data more powerful. You’re adding more and more context.

So after a session, I’ll come back and splice the data out for each player based on what they’ve done. Then, I upload all of that into the SQL server-based platform I’ve built that is like a central database. Once the data is uploaded, I add all the metadata to match each dataset with a specific drill. This lets us go back and see how intense a particular drill was for each player.

Then, the next step is to send out training reports for the day. We use Tableau Software to generate the report based on what the coaches need. There is a bevy of reports that go to different coaches, some based on specific requests.

As of late, we’re also doing the same for match data — we have about 250,000 rows of data per match now. This helps us understand game demands a lot more and also shows us more insight into the tactical side since we can look at every pass and each movement during a given match.”

GeekWire: How much impact does this data really have on what happens at practice or during matches?

Ramineni: “It’s not easy to establish that link. You can say that’s the holy grail. I always look to quantify it in terms of how much our training has changed, and how much we look at the data and how we use it to make decisions. The way it works is, the training reports are sent to Dave, and he makes a decision to change how much high-speed running each player does the next day, for example. Maybe the underloaded guys get some extra work, while the overloaded guys do a little less. Those are all decisions you make using this data.

As far as it helping us win more matches, last year I look at Clint Dempsey coming back from the World Cup. It was a long trip for him and once he came back, we decided to give him some extra rest. We saw other players from the World Cup come back and play right away. You could see that they either weren’t as good or promptly got hurt or missed a few games after. We were able to rest Clint and he stayed healthy for the rest of the season.

It’s hard to quantify that, but if we rushed him into the lineup, then maybe he couldn’t have played the final couple of matches that were key to us winning the Supporters Shield.”

GeekWire: What about this season? There have been a lot of injuries lately.

Ramineni: “People won’t believe it, but we’re still roughly at the same pace of last year as far as total number missed days due to injury issues. Everything happened in a small time frame this season, though.

But we’ve definitely had some issues this year and we are investigating those, doing some deep dives. Anytime something new pops up, you want to be able to prevent getting bitten by the same thing again. What we are aiming for is where we have an early warning system. I don’t want to say predicting injuries, but an early warning system that flags when there might be something going on that we need to look into. We have enough data now that we can do this and we’ve been adding contextual information to the data as we go. That’s where we want to go and we are making progress, but there is always room for improvement.”

GeekWire: How much has your job changed since you first got here?

Ramineni: “A lot. When I first joined, I had nothing to show because I was building the platform. When I came here, there were just Excel spreadsheets. When I built the platform and started building reports, we only had two or three reports for each training session. Now we have around 20.

There’s a lot more usage of the information and we’ve worked on presenting it in a way for coaches where it’s easy for them to understand. They have so little time and we try to be as brief as possible but provide them with useful information. We’ve improved on that.

We’ve also just added more reports because there were more questions coming from the coaches. I think one of the most important things about data is that it should generate discussion, and that should create more questions. We’ve generated the discussion enough that now we’ve actually made progress in terms of providing different types of reports with more data and we have gotten to stage now that if we don’t send the reports early enough in the afternoon or evening, we get emails asking for them.

I think this is another way of saying, while someone may not use all the data to make a decision, they don’t want to make a decision without looking at the data.”

GeekWire: How does your job with the Sounders compare to your seven years with Microsoft?

Ramineni: “There are definitely some similarities. When I was working on Bing, I had to process all the data and mine it. We collected 15 terabytes of data every day, and we had to sort through that and write queries and analyze it. That helped me a lot with what I’m doing now. The data volume is much lower, but it’s the same principle and idea.

The other thing with Microsoft, it’s a big company with a lot of people. For me now, this is like a startup. There are five of us in the department, and just myself and an intern on the data analysis side.

And when I was at Microsoft, everyone was like me. Now I’m like the guy that is so different. The first few months here, there were always guys who’d look at me and say, ‘What is this Microsoft guy doing here?’ When you’re going into a field where you are a completely different bird with different types of experience and a different background, it’s easy to get overwhelmed and easy for people to think, ‘he’s not going to help us because, what does he know about soccer?’

It’s taken some time, but I’ve shown I can help them. I was able to slowly build trust and now I can answer questions from coaches, who are trusting the numbers and also me.”

undefined