“The Save” – will Stefan Frei pull it off again?

By on December 9, 2017 in Uncategorized

The 2017 MLS play-offs are here – and we’re back with an assessment of the MVP of last year’s finals, goalkeeper Stefan Frei. He crushed Toronto’s hopes with that impossible save against Jozy Altidore. Will he be able to pull it off again and help the Sounders win the championship two years in a row??

Unfortunately we haven’t compiled his full season yet – but there’s enough data for a snapshot of performance trends to date. Obviously Frei has had an outstanding season including an incredible play-off run. Bottom line? He’s not the best shot stopper in the league, but he is easily one of the top all-around keepers in the MLS that we have data on.

His distribution percentage is excellent across all categories (remember, Stopper counts a distribution event as successful if the ball is received by the target player and the play is continued) and he is very much a modern goalkeeper, as comfortable with his feet as his hands and providing deep support for his back line. The Sounders tend to play out of the back, largely because of their confidence in Frei.

When it comes to crosses, Frei dominates. His ability to win the ball in the air is outstanding. His timing and positioning is such that he rarely needs to punch the ball out – of 50 crosses recorded over 24 games, 68% were classified as high catches.

The Sounders defence has been so effective that it’s hard to get an accurate read on whether Frei has any left/right or high/low bias – although we can say conclusively that his aggressiveness and excellent positioning make him as hard to beat in 1v1 situations as in the air, with a Sv% of 0.86 in all recorded 1v1 scenarios. Our analysis suggest he is more likely to allow a low ball to his left than anywhere else, a finding supported by both penalty goals and the single free kick goal recorded in our partial season data. He’s so good in the air that beating him with a high shot, left or right, will be tough (as Altidore knows only too well!) It should be noted that Frei gets a higher percentage of shots on his left period, which is likely as much a function of a team defensive weakness than any flaw in his performance but regardless – Toronto, keep it low and to his left!

 

Updates to the Stopper app

By on October 22, 2017 in Uncategorized

We’ve been working on a number of updates to the app and now the latest release is available in the App Store and on Google Play. If you’re currently using Stopper, the biggest difference you’ll notice is the Game Summary screen: when saving a game, now you can mark it as a Win, Tie or Loss. Another new feature you’ll notice – on the same Game Summary screen – is a “Keeper Sentiment” rating, allowing the goalkeeper to rate their own performance.

We’ve also done lots of smaller UI updates, taken care of a couple of bugs and improved the scaling on the data capture interface to make it look and work better across a broader range of mobile screens. And for those of our users looking to track multiple players, or a single player across multiple clubs/teams – we’re working on it!  We’ve got some exciting new developments in the works that we’ll be updating you on in the next few months.

Stopper at Demo Camp Halifax

By on October 18, 2017 in Uncategorized

Our CEO Tyson Hynes took the stage to demo Stopper to a capacity audience last week as part of Halifax’s Demo Camp 2017. Watch the video for a live demo of how Stopper records goalkeeper game performance – and learn a surprising insight into Tim Howard’s strengths and weaknesses, thanks to Stopper’s unique game analytics.

 

Is there a better way to evaluate keepers?

By on March 31, 2017 in Uncategorized

Everyone agrees – standard goalkeeper stats like Goals Against and Save % vs Shots On Goal don’t tell anything meaningful about keeper ability. A great keeper on a weak team will have a high GAA; a mediocre keeper behind a strong team will have a lower GAA. Assessing a keeper often comes down to watching them play – and then going with your gut.

Does Stopper provide a more objective way to analyze keeper performance? To test and validate Stopper data for developing players, we’ve been tracking stats on several pro keepers. Today we’re going to share our data on the Vancouver Whitecaps keeper, David Ousted: a consistently strong player with over 10,000 minutes logged in MLS and a 2015 finalist for MLS Goalkeeper of the Year.

If you’re not familiar with Stopper, here’s a snapshot of the data we pull from a typical game – in this case, Whitecaps vs the Sounders on Sept. 17, 2016. The final score was 1-0 – but from a GK’s perspective, here’s the data that matters. (I should note here that we don’t really track the “Communication” metric for recorded games since its too noisy to hear what the keeper is saying!) Despite conceding a goal, Ousted made some solid saves including a nice reflex tip on an Ivanschitz ball at the 60′ mark.

Now let’s look at aggregated Saves for the Whitecaps 2016 season (below). As one would expect the majority of saves (24.5%) occur in the central quadrant of the goal. But over the course of the season, a strong trend emerged – even though Ousted made a similar amount of saves on both sides, he caught and held more shots on his left and tipped away more shots on his right.  From that we can infer that a) there may be a general tendency for right-footed strikers to aim wide on keeper right or b) Ousted is more comfortable catching and holding the ball on his left.

Here’s where it gets interesting! When reviewing the heat map of Goals Against, we can see an inverse correlation with saves. Of the 52 goals scored against Ousted in 2016, 32.7% were scored in the lower left quadrant of the net – almost double that of the lower right at 19.2%.

Obviously there are a lot of other factors that come into play in individual games (for example, a low far post goal might come from a header off a cross, or a shot from a surprising angle that catches the keeper off guard) but the aggregated season numbers seem to tell a clear story: the odds of scoring against Ousted on the lower left hand side are more than double when compared to the other quadrants of the net. And if comparing shots on goal vs goals allowed, overall scoring on the left hand side was almost twice as likely as on the right.

We’ll look at Distribution in a separate post as we wrap up our first analysis of a MLS keeper over a full season. Does Stopper provide a more objective way to analyze keeper performance? We’ll need to evaluate a few more pro keepers as we continue testing the app but early indicators are looking good: there may finally be a better benchmark for evaluating goalkeepers even at the professional level.

While the current version of Stopper is designed for developing keepers, we are working on a “professional” version that allows more detailed data capture and analysis, including video capture linked to Stopper data metrics. We’re looking for clubs or organizations that would like to do beta testing – drop me a line if you’re interested!

•Note that we have a few more “Saves” recorded than the official MLS stats – likely because of differences in accounting for Crosses vs Saves. If the ball was played as a cross but ended up being off-target with the potential to be a goal, we’ve recorded it as a save – OptaStat may record that kind of “accidental shot” differently?

Go long or go short? Part 3

By on March 3, 2017 in Uncategorized

Third and final post on Jared Young’s awesome analysis of MLS goal kicks on americansocceranalysis.com.

In our last posting we looked at how he show that long balls (i.e. goal kicks that went into the opposing half) statistically increase the opposing team’s goalscoring chances, while decreasing those of the goalkeepers team.

So now let’s look a short balls (i.e. balls received within the keeper’s 40% of field). The first possession of the offensive team results in a goal just 0.9% of the time. This is understandable given the team is beginning their possession further away from goal. If the defense gets the first possession they have an average chance of scoring of 1.2%. So far, advantage long ball.

But again it’s on the second possession where things get interesting. The offensive team scores 1.4% of the time and the opposition’s second possession results in a goal just 1.0% of the time. The results have flipped compared to the longer pass. Factoring in the percentage of first pass completion, then a shorter pass from the goalkeeper results in a goal 2.3% of the time over the next two possessions.

So if an average team decided to only have the keeper make passes within their own half and these results held, the team would improve their goal differential by 1.6 goals per season. That’s not a lot of goals – but in a tight season it could be the difference between top ten and top five in the table.

Jared’s hypothesis for why this shift occurs is due to the positioning of the opponent. In other words, the long pass allows the defense to get set deeper and shifts the possession odds in their favor. Meanwhile a shorter pass draws the defense forward and over the course of the two possessions increases the team’s chance of scoring – and let’s not forget nearly tripling the chances of winning the first possession. Here’s a quick chart that re-caps Jared’s long vs short ball analysis:

 

One of the things we set out to measure with Stopper was the success rate of goal kicks by the keeper. Even before we had the opportunity to read Jared’s excellent analysis, we decided to measure goal kick success as generating a second possession – which we now know leads to increased goal scoring opportunities.

 

Go long or go short? Part 2

By on February 11, 2017 in Uncategorized

And we’re back with our Part 2 review of Jared Young’s excellent analysis of goal kicks on americansocceranalysis.com

In Part 1, we looked at how it seems like based on scoring chances in the offensive end relative to other areas of the pitch, long balls make sense. In Part 2, we look at his analysis of pass completion of long vs short balls; and even more interestingly, how long vs short balls affect scoring chances on the second and third possession after winning the ball. Possession on goal kicks is a key metric for Stopper so we’ve had a lot of discussion over this!

For starters, Jared Young divided the field into two general zones – 1 being long balls, 2 being short. Over the course of the 2016 MLS season there were 3,200 long and 1,600 short goal kicks. The probability of completing a Zone 1 long ball is only 33.5%, significantly less than a Zone 2 short ball which comes in over 95%.

On a successful long ball, the offence scored on that possession 1.4% of the time, which is an above average rate. Conversely, if the defense gets possession, which it will 66.5% of the time, they have less than a 1% chance of scoring. So far, things still seem in favour of the long ball … but if you multiply the scoring odds by the probability of getting possession, the numbers actually come out in favor of the opponent 0.6% to 0.5%. Statistically, that’s not really significant except that the whole point of the long ball was to create greater overall value (e.g. scoring chances) for the team.

To quote Jared, “here is where things get a little crazy.”

Let’s say the long ball to the player is successful and the offence is able to pass the ball for a second possession. Despite being in a part of the field with a statistically better than average chance of scoring, the probability of scoring a goal on a second possession drops to just 0.7%. Conversely the second possession by the defending team is stronger than average at 1.5%.

On the third possession, despite the average position the opponent still has an elevated 1.6% chance of scoring while the original goal kick team gets back to the average of 1.2%.

Jared’s conclusion? When an MLS goalie begins a possession with a long ball, his own team subsequently has a 1.2% chance of scoring on the next two possessions – while the opponent chances jump to 2.1%, which is statistically equivalent to having possession at the top of the attacking third, a not insignificant advantage.

Long balls are a staple of goalkeeper play and hey, they look pretty cool. But based on the above analysis, are they the best option for a keeper? In our next segment, we’ll look at Jared’s analysis of short balls and how they stack up relative to the long goal kick.

 

Go long or go short? Part 1

By on January 15, 2017 in Uncategorized

Came across a great article by Jared Young on americansocceranalysis.com on the topic of goal kicks. One of the data points that Stopper captures is whether keepers turn over the ball on their goal kicks which leads to the question Jared asks: why is the long ball off the goalkick a staple of soccer at every level?

To quote him: “It’s long bothered me that goalkeepers always launch the ball into what seemed to be at best a coin flip proposition. A team has the ball (the most valuable thing on the field) and then they decide to just sling it up in the air to chance. Why does that make sense? The conventional wisdom is that 1) it’s better to get the ball as far up the field as possible and 2) even if the first attempt doesn’t work the next possession will still be closer to the goal. These two elements of said conventional wisdom turn out to be true, but it’s the coin flip that becomes the glaring issue.”

Let’s look at Jared’s analysis of that first statement today. The chart below is a result shows every possession in the 2016 MLS season as measured by having at least one pass or shot attempted, with the number showing the percentage of total goals scored starting from that position. For comparison, he broke the field down into 50 roughly equal zones – 5 vertical and 10 horizontal.

The probability of scoring on any possession in the 2016 season was 1.2%. It’s pretty clear that the closer you get to the opposing goal, the greater the probability of scoring. So on the face of it, conventional wisdom number one about goalkicks is true – assuming you can win possession and keep it.

We’ll walk you through Jared’s analysis of part two later this week – and find out if long kicks actually generate more goalscoring chances!

A Tale of Two Halves

By on November 23, 2016 in Uncategorized

In Tuesday’s back-and-forth match-up between Toronto FC side and the Montreal Impact, the first goal of the match was a stunning finish by Impact striker Dominic Oduro in the opening ten minutes of the game. That set the tone for the first half of the game with Toronto keeper Clint Irwin conceding two and Montreal’s Evan Bush with a clean sheet at the half. We’ll share stats on the full game soon – but first, a comparison of distribution showing the different keeper playing styles, even at the half.

16-11-23-tfcvsmi-2

Montreal used Bush to stretch the field, while Toronto barely played the ball back to Erwin at all. His one delivery off a back pass was turned over – and culminated with Oduro’s brilliant run down the flank. More game stats to follow soon!

 

 

Testing Stopper: Goal Kicks

By on November 10, 2016 in Uncategorized

How does Stopper measure goal kicks? As we get ready for an all-Canadian MLS conference show-down, we thought we’d use the break to look back at some Distribution results for David Ousted at the half way mark of the season.

If you look at the chart, you’ll see goal kicks measured as either successful or unsuccessful, with most recent games on the left and older games on the right. What makes a goal kick successful? Since the objective in football is to score goals, and to score goals you need possession of the ball, it follows that a goal kick that leads to continued possession is “positive” and one that turns the ball over is “negative”. Is the player able to receive the ball? Are they able to continue the play? Were they able to make a pass or hold the ball? If so, Stopper considers that a successful goal kick.

Obviously, the skill of the receiving player and of the opposing defence affect the outcome of a goal kick as well. But it’s up to the goalkeeper to weigh those considerations and make a decision that will generate goal-scoring opportunities. If the opposing defence is strong in the air, then continually sending long balls creates a low percentage environment. If they are weak in the air, then a long ball creates high percentage conversion opportunities.

For example, this chart shows that against the Galaxy in Game 5, Vancouver had significant ball turnover on goal kicks vs against Sporting Kansas in Game 9, where a higher percentage of goal kicks were converted to successful plays. In the same games, it was interesting to note successful distribution from a back pass was also significantly lower as a percentage vs the Galaxy than vs Sporting Kansas, with the latter match featuring both 100% conversion and more than double the number of back passes.

Next week we’ll look at Saves and Goals over the course of the Whitecaps season.

NOTE: 4th last game is a blank test game.

Testing Stopper: Halifax City

By on October 24, 2016 in Uncategorized

Thanks to all the keepers at Halifax City who have volunteered as Stopper testers! The Winter 2015/16 season is just getting started and we’ve got some great players signed up to Stopper. This will be the first full development season being tracked so we’re excited to see how our keepers use it to improve their game over the course of the next few months.

Special thanks to @CITYSoccerClub for all their support and encouragement!