Strength of schedule update: Sounders with easy run-in

Hello again everyone!

Let’s get it out of the way now. We lost last week, twice, and the streak is over. Wednesday was painful, but we were OK. Sunday though, well that was just painful. Matters then got worse when we learned that Harry Shipp and Brad Smith may be out for a few weeks.

What’s done is done though, and the Sounders need to turn it around to keep playoff hopes, and preferably first-round home game hopes, alive. Fortunately, it seems that the team still has their eye on the finish line, and this weekend is the perfect opportunity to bounce back.

That’s because on Saturday we’ll be hosting this season’s second-worst away team. In fact, things stay fairly sunny after that too. The rest of this season is shaping up to be pretty strongly in our favor, and our main competitions’ seasons are not. So let’s return to this post I wrote three games ago and see how things are shaking out. (see below for the original post)

The bottom line

Incorporating the latest results from around the league, the Sounders are still on pace for 4th place in the West, but now they’ve moved to the bottom of the strength-of-schedule table (in a good way).

The Sounders now have the easiest schedule remaining in the West on paper, with the most difficult fixture being October 21st at Houston. LAFC still have a relatively easy schedule, but the rest of our immediate competition (RSL, Portland, Vancouver and LA Galaxy) have some serious challenges ahead. That’s good for us. Here’s a look at the remaining schedule:

• 2018-09-29 hosting Colorado (0.43 PPG away, 2nd worst in the league)
• 2018-10-08 hosting Houston (0.57 PPG away, 5th worst in the league)
• 2018-10-17 at Orlando (1.27 PPG at home, 4th worst in the league)
• 2018-10-21 at Houston (1.6 PPG at home, 8th worst in the league)
• 2018-10-27 hosting San Jose (0.60 PPG away, 6th worst in the league)

In summary, that’s three very bad away teams coming to town, and two medium-bad home teams hosting us. Here’s how it compares to other teams in the West:

Difficulty of each remaining game (probability based on PPG home vs away)

Or to put it another way, we can aggregate together the PPG for each team’s remaining schedule, and it makes a good summary of strength of schedule they have yet to face:

Strength of remaining schedule for each team based on opponent’s PPG (home and away)

Despite our recent losses, not much has changed since my original post in terms of expected points and standings. Using the simulation approach defined below, we’re now estimated to end with 54 points, which should put us in 4th place. Dallas retains the top spot, as my model predicts they are most likely to finish first (though it’s still very close).

Expected points and standing according to simulations

Unfortunately, the first round bye now appears out of reach for the Sounders. Only 2% of my simulations had us finish in the top 2 spots. On the other hand, only 2% of my simulations showed us missing the playoffs entirely, so we’re just as likely to continue our run of playoff berths.

But what happens if we drop the ball this weekend? I know it’s a favorable match-up, but I’m now nervous enough that I looked up this bad scenario. It turns out, in the 11% of simulations where my model predicted that happening, the Sounders still went on to finish 5th in the West with 52 points, just behind Portland. If we lose the next two? Amazingly, the strength of schedule facing Vancouver and LA Galaxy still leave us ahead of them in the modeled standings, just barely making the red line with 50 points on average. It would in fact require three more losses for us to statistically be more likely to miss the playoffs than make them, but let’s hope it doesn’t get to that.

It is nice to know though that we do actually have a little bit of cushion, thanks to the odds elsewhere in the conference.

---

Original Post and Methodology

If you’re anything like me, you’re feeling much better about our fate in this 2018 season than you were a few months ago. It’s amazing the difference an eight-game winning streak can make, isn’t it?

As supporters, weekends have gone from worrisome to wonderful again as we’ve watched our team scrape out win after win, and it feels like a weight is lifting off the shoulders of the players too.

If you’re like me though, you’re also anxiously watching the standings and results from around the Western Conference as our competition bounces around above and below us. Unfortunately, they’re not really helping us out lately; many of them pulling together impressive runs of their own.

Well I’m here to deliver some good news (or at least expectations) on that front, in the form of updates to the strength-of-schedule rankings. Let’s look ahead to our remaining eight games and the points per game (PPG) of the opposition:

• September 15 at Vancouver. Vancouver averages 1.77 PPG at home
• September 19 against Philadelphia. Philly averages 1.21 PPG away
• September 23 at LA Galaxy. LAG averages 1.57 PPG at home
• September 29 against Colorado. Colorado averages 0.5 PPG away
• October 9 against Houston. Houston averages 0.54 PPG away
• October 17 at Orlando. Orlando averages 1.29 PPG at home
• October 21 at Houston. Houston averages 1.5 PPG at home
• October 28 against San Jose. San Jose averages 0.64 PPG away

There are some tough ones in there for sure, but there are also some more promising match-ups against the weaker teams this season, as you can see. To be clear, there’s no such thing as an easy win in MLS, and none of these should be taken as a given. This just shows how each team has fared so far. As we know, huge turnarounds are possible.

However, our opponents’ collective PPG are pretty favorable for us, especially when compared to the strength of schedule for everyone else around the league:

This figure shows the average PPG each team still has to face as of this point in the season (keeping track of whether they’re playing at home or away). The teams are ranked in order of difficulty.

As you can see, Colorado has the most difficult remaining schedule (opponents’ PPG: 1.7), and LAFC has the easiest (opponents’ PPG: 1.07). Third from the bottom however are the Sounders (opponents’ PPG: 1.13), with a similar strength-of-schedule similar to Atlanta and DC (1.14 and 1.09 respectively). Also note that we have games in hand on most teams, with 8 remaining where most have 6 or 7.

By the numbers, things are looking good. Our closest competition in the standings, LAG, Portland and Vancouver, have relatively challenging remaining games ahead of them. They each fall right in the middle of the pack in terms of strength of schedule. Unfortunately, the next three teams above us, RSL, SKC and LAFC, have easy-ish schedules too, setting up for a tight race for the crucial first-round bye. FC Dallas may be poised to drop some points though; they have the hardest remaining schedule among Western Conference teams currently in playoff position.

That’s good to see, but where, exactly does it leave our prospects? Where are we most likely to end up? How likely is it that we’ll make the playoffs?

While there are sources (like 538) out there who provide such answers (and are much better at it than my rogue analytical musings), I went ahead and tried to figure it out.

I put together the simplest prediction model that should still pass the sniff test: PPG of the home team, PPG of the away team and the disparity between the home team and the away team.

[Warning: statistics ahead. Skip this paragraph if you want]

I threw the three of those variables (representing the “disparity” as an interaction term) into a multinomial logistic regression with win/draw/loss as the possible outcomes and fit it to the 2018 season’s results so far. Doing so yielded remarkably accurate predictions for such a parity league. Across all games in 2018, the regression correctly predicted the winner 61% of the time, and most (58%) of its incorrect guesses were ties that were just barely expected to be wins. Of course there are lots of limitations. This is in-sample fit and anything can happen out-of-sample. It also doesn’t account for momentum (i.e. correlated residuals), history (it’s only 2018’s data), days of rest, xG and a host of other things that predict wins. But problems with extrapolation notwithstanding, I’m willing to assume that this model yields reasonably-reliable predictions for the rest of the season, especially when you take into account confidence intervals, which I do.

OK back to the results. Here’s what the prediction model says are the probabilities of winning for the remaining games in the Western conference:

You can see that the bad teams have a low probability of winning in each remaining game, and also that the bad away teams (like RSL) have huge swings from game to game, depending on whether they’re at home. The Sounders and LAFC have a reasonably-high probability in each game, reflecting the relatively easy schedules. Our lowest-probability game is the next one against Vancouver, who are pretty good at home, and our highest-probability game is Sept. 29 against Colorado, who are very bad on the road. You could have inferred that from the list above, but it’s encouraging to see that this prediction model I made says the same thing.

So, I simulated the rest of the season for every team. In the simulation, I allowed for a realistic amount of randomness of unexpected wins and losses, all according to said prediction model. Then, I did it 999 more times because computers are cool and can do that. I then tallied up the total points each team ended with in each of the 1,000 simulations and pieced together the standings. The results look like this:

You can see from the graphs roughly how often (among all simulated seasons) each team ends above and below the red line (which I drew at 6.5). Sorry, but I didn’t do anything to resolve ties in this simulation because I’m lazy and this was already pretty complicated, so you can see small bars at 1.5, 2.5 etc.

Anyway, this puts the Sounders at an average of 59 points (four higher than the 538 analysis predicts), which is good for an average position of 4th place (same as 538). There’s a fair amount of uncertainty though, as the 90th percentile of their finishes was 2nd place (64 points), and the 10th percentile was 6th place (54 points). The teams ahead of us are expected to be Dallas, LAFC and SKC, though who will be in each position is really anyone’s guess, as you can see.

Here’s the final table of averages across all simulations for the Western Conference:

So that’s good news! In fact, the Sounders made the playoffs in 95.5% of the simulations. We made the top 2 in 15% of simulations, and finished first in the West in 4.5%. Sadly, there were no simulations in which the Sounders won the Supporters’ Shield, but I think we kind of knew that already.

What does this mean in plain English? With eight games remaining, and considering everyone else’s schedule, the Sounders just need to win the games they’re expected to win and we’ll be alright. We could even afford maybe one unexpected loss. Of course, with even two unexpected losses, we might be right back to anxiously monitoring the weekly standings again. I suppose I will anyway, but let’s hope it doesn’t come to that and we ride this wave to the third straight MLS Cup Final.

...I’ll leave you with some other fun things we can do with all these simulated finishes to the 2018 season.

For one example, if you’re a fan (like I am) of Sounder-at-Heart Talgrath’s Doom Counter, you may be interested to know that 5.7% of simulations included a team making the playoffs with 47 points or less. There, it’s settled. It’s not impossible, but it’s unlikely.

For another interesting tidbit though, we can check how long this win streak is expected to continue (according to the 1,000 simulated ways the season might finish). Among those simulations, the Sounders beat the Whitecaps, and thus extended the winning streak, in 399/1000 simulations (I estimate a 39.9% probability of win, 538 estimates 33%). On top of that, we extended the streak to 10 games in 18.2% of simulations, assuming they’re independent events (see limitations section above). In fact, there were actually 8 simulations where the Sounders did the impossible and just won out the rest of the season. I can honestly say I didn’t expect any, purely from a statistical point of view. I’m not saying it’s going to happen, but it’s fun to know that there’s actually a universe in which it does!