Thorns FC: 2017 By the Numbers

By John Lawes Nov 22, 2017, 10:12am PST

The 2017 NWSL Championship already seems like a lifetime ago, doesn’t it?

Before we leave 2017 behind, though, let’s take one last look at the Thorns’ 2017 title run through the lens of the plus-minus ratings (“PMR”) system.

Remember how the PMRs work? If not, the long-form explanation is here.

Here’s a table presenting the PMRs for the entire Thorns FC for the full 2017 season. The table presents the player’s plus rating in the green shaded column under the player’s name, the minus rating in the next column over, and the net - the sum of the pluses and minuses - in the third column. Note that net negatives are shown in red.

To make this a little easier to visualize, I’ve made a “block and whisker” plot for this same dataset.

Ever used one of these before? It’s a nice way of showing the combination of the average value and the spread around that average. Here’s a graphic explanation of how to read the block-and-tail icons. I’ve used Emily Sonnet’s data as an example.

Emily Sonnett was a particularly useful player for this example because her play was so up-and-down in 2017; she was the Queen of the Outlier. The fat part of her bar is fairly short, meaning that she had a lot of games where her net PMR was close to her mean - which was also close to the team’s net season average of +6.0.

But Sonnett had a game (Boston here on Matchday 7) where she went +23/-5 for a net +18 and a game where she was rated +2/-13 (Boston away on Matchday 23) where she was a net -11. Those are the dots, and she has more of them than any other Thorn.

We’ll talk about the individual players in a bit; but, first, let’s look at how the PMRs document the Thorns’ play over the course of the season.

As a team, the Thorns’ aggregate net PMR ranged from a high of +13.5 (for Matchday 7 against Boston) to a low of +0.8 (Seattle away on Matchday 12).

What the data shows is that the Thorns went through roughly three distinct periods of play during 2017.

April to Mid June

During this time, the team’s net PMRs suggest what was going on on the pitch; a team still finding itself. Tough losses, like North Carolina away, alternated with impressive wins such as Boston and Chicago here. The team’s position on the table jumped from as high as second to as low as sixth. Players, too, went from a good outing to a poor one; ups-and-downs rather than consistently good or poor form.

Mid-June to Mid-July

Beginning with the 1-nil away loss to Washington on June 24th, the team slumped over three of four straight games with poor performances at Washington and Houston, and a total collapse in Seattle (the worst-played match of 2017). These matches are the ones highlighted in yellow and orange in the plot above.

Despite the result, the 3-nil FCKC home win is within this group. Christine Sinclair, Hayley Raso, Amandine Henry, Allie Long, and Dagny Brynjarsdottir had very good games and carried the team. Several other starters weren’t doing so well or were outright poor. The result was good, but the team and many players were still struggling.

Let’s go back to the data and isolate those four games to take a deeper look at who those players were. From this we can identify four trouble spots.

Nadia Nadim was missing for much of the time at the UEFA Championship finals. Her replacements were Ashleigh Sykes, who picked those matches to faceplant on the pitch, and Tyler Lussi, who was unimpressive in her single match,
Long had a series of below-par to outright awful games,
Shim, who was called to fill in midfield, was a trainwreck, and
The defense had a series of pretty awful games as a group, and as individuals, Sonnett and Klingenberg were both not in form.

This could have been a breakpoint. From here, the Thorns might have slid out of contention. They didn’t, and one reason was North Carolina came here in July riding a midseason wave and the Thorns rallied, defeating the Courage 1-nil.

Mid-July to the End of the Season

After that win, Thorns FC went on a tear, going 9-1-1 through the end of the regular season.

One reason was many of the players crushed it individually, in particular during the two-week stretch between August 19th and September 2nd which included three wins by an aggregate score of 8-2.

(Mind you, it didn’t hurt that two of the three victims were Washington and Houston and the third was Seattle without Megan Rapinoe.)

The PMRs suggest the obvious: that the team was playing well during this time. The only loss was a poor group performance in Kansas City. The only other games where the PMRs suggest the Thorns played poorly were both wins; a scrappy 2-1 result over Washington here in July where the Spirit beat Portland to the basement, and the truly awful performance in Boston that was rescued by an against-the-run-of-play Raso miracle goal.

Good is good. But, sometimes, lucky is good, too.

Here’s what I think the data shows helped make that run happen.

Moving Backwards to Go Forwards

Tactically, Mark Parsons moved Sinclair back into midfield for good. That helped the forwards, too. Here’s the forwards’ net PMRs in the early season.

Here they are after mid-July.

All the players with significant minutes are looking better; Sinclair’s mean is up just a bit, but Raso’s and Sykes’ performances are significantly improved. Only Nadim is lagging, and I suspect that was the effect of her long European trip.

I think this general improvement had a lot to do with Sinclair pulling the strings from the top of midfield. The forwards, including Sinc, had been starved for service much of the early season. With the not-so-big-boned Gal from Southern British Columbia behind them, though, the passes started to arrive, and the goals began to pour in.

Midfield Maneuvers, the Great Horan, and the Mystery of the Missing Long

In the midfield, here’s the changes between the early...

...and the late season.

One thing jumps out from these two plots: the Great Horan emerged in midseason. That alone was a key to the title drive. Horan was the single most-improved player in the second half of the season.

But Parsons also made several tactical adjustments in midfield. He moved Mana Shim - first to the bench, then off the squad. That cut out a ton of minuses, and a lot of trouble on the pitch.

He also began subbing out Amandine Henry, who was fading in late minutes, for Brynjarsdottir in the last half hour or so. You can see how that helped Brynjarsdottir, and (despite the plot showing Henry’s numbers dropping in late season) that boost helped the midfield overall.

Allie Long was also largely relegated to the bench.

I want to talk more about Long in the future. I think her past season and her next - whether with Portland or elsewhere - deserve a full discussion. But for the sake of this inquiry, and purely based on her PMRs, I can say just this - something happened to Long around the time of the Matchday 9 1-3 home loss to Sky Blue.

Although her work was as up-and-down in the early season as her club’s, Long’s average net rating was well above the team mean.

But, after playing well in the Sky Blue loss, she had her worst match of the season in the 1-nil away loss to Washington and never regained her early season form. By the Final, she was reduced to playing a single minute.

The narrative in 2017 was that Long was going to have trouble seeing minutes because Henry and Horan were the better midfield options. But I think a large part of her shrinking field time was not so much her teammates’ positives, but her own negatives. Long suddenly wasn’t having a good season and I think Parsons saw that and did what he had to do.

We may never know the why of what happened to Long. But her rating numbers clearly show that something was wrong enough that her coach believed his only recourse was to remove her from the pitch. I personally hate that had to happen to Long, but looking at her numbers, I understand why it did.

The Kat That Came Back, and the Backline

The Thorns defense stepped up noticeably in the late season. The return of Kat Reynolds to right back may have been a factor, as well as the the defenders in general finished working out their early season jitters. Here’s the net PMRs for the backline and Franch up until mid-July.

And here they are from that time to the end of the season.

What I find interesting about the idea that Reynolds’ return made the difference is that Reynolds’ ratings aren’t better than Boureille’s, the player she replaced. On the contrary; Boureille’s season average net of +5.2 is more than half again better than Reynolds’ +2.9.

One possible explanation is that having their familiar teammate back on the pitch helped Sonnett, Menges, and Klingenberg relax and work together better. All their mean ratings are higher after Reynolds’ return, and the spread of their endpoints is much lower, indicating that they’re playing not just better but more consistently better.

Another possibility is this is simply correlation and not causation - the other defenders worked out their issues about the same time Reynolds got back on the pitch.

Whatever the reason, Sonnett made a huge step up in the latter half of the season, and Menges and Klingenberg did better, as well.

Franch, as we know, turned out to be the Goalkeeper of the Year.

Primary Numbers - the Best and Worst of 2017

Which player had the best individual game of 2017?

Lindsey Horan, with a net +25 in the 2-3 win in Chicago, August 12th. Here’s what I said at the time:

“Horan (+16/-5 : +14/-0 : +30/-5) Woman of the Match for the second game in a row. Simply sensational. Her passing was outstanding and, more importantly, when Chicago started seriously pressing in the second half she tightened up her precision and stinginess with the ball. All five of her “minuses” were either tackles or passes for loss in the first half – in the second? None.”

I’d argue that her contribution to the 2-2 draw in Boston back in May was almost equally important, but purely by the numbers that was the best Thorns performance of the season.

Who had the worst individual game of 2017?

Before I looked at the table my thought was Brynjarsdottir in Seattle, just based on how bad she looked that day. But having her lunch eaten by Rapinoe was only the second-worst beatdown suffered by a Thorn in 2017. Nope, the worst outing of the season was Sonnet in Boston in August.

“Sonnett (+1/-7 : +2/-6 : +2/-13) Oh. My. God. If there was any question how critical Emily Menges is to the Thorns’ central defense this match conclusively answered it. Without her Wall of Emily partner Sonnett shambled around in a welter of poor positioning and questionable decision-making. The worst I’ve seen from Sonnett all season, and there’s no real excuse; you should and can be better than this, Sonnett. Fortunate not to have been punished worse.”

Ouch.

Who had the best overall season in 2017?

Several players were consistently good in 2017. By average net PMR, the top performers of the season were:

Player Average net PMR Comments

Horan +10.5 Her post-matchday 14 average is +14.

Sinclair +9.7

Henry +9.5

Raso +9.2

As noted above, Horan emerged as a great player at the end of this season. She was good before, but she was a force of nature late in the year, and, as such, is also the team’s top performer for the season.

Which of the regulars had the worst season in 2017?

Two players with more than 10 starts had average net PMRs below +5.0 for the season; Klingenberg, at +4.2, and Brynjarsdottir at +3.2. I consider both to have “extenuating circumstances”.

Klingenberg had some very shaky performances very early in the season as she recovered from her back problems. And she struggled with her positioning all season; Parsons, like most modern coaches, expects his outside backs to get upfield. Klingenberg, in particular, provides excellent service from the left wing. But her pace is not what it was (and she was never the speediest of defenders to begin with) so she tended to end up with a lot of +11/-7 sorts of games.

But her 6 assists put her up near the top of the league, and were twice her nearest teammates’ (Nadim and Raso both had 3). She had a rough season, yes, but she also provided those critical assists and finished the season strongly when the team needed her most.

Brynjarsdottir’s numbers are hammered by her shellacking in Seattle. If you take that match out of the table, her net PMR goes up to 4.4. And she played a huge part in sealing the win in the Final; her +13 in 36 minutes is second only to Nadim’s +15 in 21 minutes.

So the regular player who had the toughest season by my count was Kat Reynolds with a average net PMR of 2.9 for the season. As with Brynjarsdottir, that owes largely to two very poor matches; a net -3 in the August away loss to FCKC, and a -4 in the win at Boston in September. But even taking away those two results her net PMR is still in the low fours.

My hope for Reynolds is that her tough 2017 was due to a late start and a long injury layoff, and that with full health and a full preseason camp she’ll be back in form for 2018.

Which player had the wildest ride, the biggest swing in form, according to their PMRs?

That would be Sonnett, who went from a +8 in the Washington match here on September 2nd, to a -11 against Boston away, to a +7 in the scoreless draw in Orlando on September 23rd. A total of -19 and +18 over the space of three weeks. Howdy, folks, please keep hands and arms inside the train and remain seated at all times.

In terms of standard deviation from the mean, it was Henry; her high of +22 and low of -4 are part of the reason that her standard deviation of 8 is the highest on the team.

Who was the most consistent?

Interestingly enough, Klingberg’s standard deviation of 4.5 is the lowest of the regulars. That makes me think the number is an artifact of Kling’s many +11/-7 games, because my impression of her during the season was always either providing an outstanding service or a timely tackle, or getting skinned in an ugly fashion. But her statistic suggests that despite appearances her highs and lows weren’t that high or that low.

Other than Kling, Long’s standard deviation is 4.4, not surprising for a veteran player of her ability. As we’ve discussed, Long’s problems weren’t consistency but overall level of play dropping precipitously in midseason.

Which one of the players (who was not a regular starter) looked the most promising?

I don’t think anyone will be surprised to see that Ashleigh Sykes’ average PMR of 5.8 was the best of the non-starters.

But even that masks her true quality. Sykes’ late start meant she had to scramble through several early matches trying to find her feet. During that time, she had three tough outings that pull her average down; Washington, Seattle, and Houston away (matches, it’s worth noting, that the team kinda stunk up, too). If you restrict her analysis to just the last ten matchdays her average jumps up to 8.3, putting her right up there with Raso and comfortably above Nadim.

Much has been made in the sporting press of Thorns FC’s needing to find a replacement for Nadim. But based on Sykes’ late-season form, I’m not sure we haven’t already found her.

Is there anyone on the fringe who looks particularly vulnerable?

Not anymore, really. Meghan Cox was the least productive of the subs and is no longer with the squad. Mallory Weber might be considered a bit of an at-risk fringe player, but she is dirt cheap as well as versatile enough to be slotted into several positions as a sub or even a spot-starter, so she seems likely to be around next April.

The Sum of the Numbers

The 2017 PMR numbers reflect what we saw on the field last season; that after some initial difficulty finding their way, and a tough four-game stretch in midseason, the Thorns turned out to be - if by no other measure than the results on the pitch - the best team in the NWSL.

Obviously, we can still hope for even more going forward; that Morris’ (and, it seems, Heath’s) injuries resolve well. That Lussi and Jordan fulfill their promise. That the front office can draft, trade for, or develop, a player who can replace Henry.

But the core of that solid 2017 team is going to be here in April 2018, and by the numbers that bodes very well for next season.

Update 11/24:

Richard Hamje (in the comments) has a good question; is there a way to assess whether Reynold’s return to the XI was the reason (causation) that the defender’s aggregate PMRs improved late in the season, or just a coincidence (correlation)?

He suggested we look at the net PMRs by the positional units, so here they are:

I think this makes the case the Reynolds’ return and the defensive improvements are related. If we add trendlines to the defender’s average net PMR some trends jump out. Up-and-down but relatively consistent early in the season leading up to a big drop down to Matchday 10 (Washington away and the beginning of the “slump”). A slow, steady improvement through the slump and into the beginning of the “late season crush”. Reynolds returns to the starting XI on Matchday 15, and two games later the defensive unit’s improvement jumps up to a rate five times the midseason trend.

When you look at the data this way Reynolds looks like the straw that stirred the defenders’ drinks.

What might tend to contradict the Reynolds-as-defensive-savior would be a general improvement in the team’s play; if the whole team showed a sudden improvement from Matchday 16 to the Final then it would be much more likely that the defenders were just part of a general upward trend.

The data doesn’t show that. Here’s the average net PMRs for the units after Matchday 9:

Matchday 10 through 13:

Forwards 2.6

Midfielders 2.1

Defenders 1.1

Matchday 14 through 16:

Forwards 12.6

Midfielders 9.9

Defenders 2.2

Matchday 17 through 24 and the playoffs:

Forwards 7.7

Midfielders 8.3

Defenders 6.4

So the data is pretty conclusive; Kat Reynolds, Defensive Superstar!