2017 Projecting By The Numbers and Chasing A Narrative

[mgarbowski]

One-quarter into the season, and NYCFC is off to its best start ever. 16 points in 9 games. Third place in the conference. Sixth place overall, and fifth place overall in Points Per Game. It is by far the team’s best first quarter, and leaves me wondering what the narrative thread of this thread will be this year. In 2015 it was the unlikely and ultimately unsuccessful quest for the playoffs. Last year it started as a playoff quest: After 9 games last year, six of which were at home, the team had only 2 wins, and 10 points in this 9 games for 1.11 PPG. The playoffs were very much an uncertain destination. On June 2 the team had only 18 points after 15 games for a measly 1.2 PPG and a projected 41 points for the year. It was not until a 4-game winning streak that finished on July 6 that the playoffs became a likelihood, and the rest of the season was spent figuring out whether the team could attain the Supporters Shield, First place in the East, or a CCL berth. It failed in all three, but it went down to the last few games of the season before that became certain.

This year the playoffs seem highly likely, but it is also a bit early to tell where anyone stands for first place in the East or overall. Also, due to the CCL format changes, there is no CCL berth to be won during the regular season. So I’m not sure what to measure for in terms go achieving goals, but that won’t stop me going on anyway.

Let’s start with the basics of the East Conference PPG standings:


NYC is third here just as in the regular standings, with a projected 60-61 points. It’s a bit astonishing that this puts NYC only in third, but TFC and Orlando got off to very strong starts, and TFC at least figures to challenge for 1st place as well as the shield. Over in the West, by the way, Dalls has a 2.25 PPG that projects out to a cool 76.5 points at season’s end.

The East conference playoff line stands at 43-45 points, which seems close enough. I trust this more than I do the first place projections, because the latter is a product of one team as an outlier while the playoff line is a product of 11 teams combined. NYC will really have to falter if it is to cause any worry about making the playoffs. There are 25 games left and this is how various finishes play out. All lines are in standard MLS W-L-T format:

14 6 5 — 1.88 63 pts
12 7 6 — 1.68 58 pts
11 8 6 — 1.56 55 pts
10 8 7 — 1.48 53 pts
9 9 7 — 1.36 50 pts
8 10 7 — 1.24 47 pts
9 13 3 — 1.20 46 pts
9 15 1 — 1.12 44 pts
8 14 3 — 1.08 43 its

For contrast, here is the same list after 9 games last year.
14-6-5 1.88 57
13-6-6 1.80 55
12-9-4 1.60 50
11-8-6 1.56 49
11-6-8 1.64 51
10-8-7 1.48 47
9-9-7 1.36 44
9-11-5 1.28 42
8-11-6 1.20 40

Now is the time for fun with charts. You might remember that 2016 looked better, but still very similar to 2015 for much of the season. Not this year.

Here is a new one:



Finishing up with assorted points:

• Remember when Portland and Atlanta ruled the GD table because of piling it on against Minnesota in the first two games? They’re both down to +5, tied for 6th overall, and below our 7+.
• Orlando’s GD is down to 0. They have made very efficient use of their goals, winning close and losing big.
• Some weeks ago someone else, I can neither remember nor find who, noted that NYC was the only team in top 5 in the league for both Goals scored and Goals Against. I have followed that and we are still in that club at spot #4 on both sides. But Toronto has also joined us in that Double Top 5 club.
• As mentioned, last year at this point NYCFC was 2-3-4 with 10 points in 9 games. One notable cause of the disappointing start was a tendency to give up goals on set pieces. Through 9 games NYC conceded 3 such goals, all on free kicks: Perquis for Toronto, Tierney for New England, and Oduro with a heartbreaker for Montreal at minute 90. Every one of those came in a tie, so if you indulge the fallacy that everything else happens exactly the same, NYC would have had 6 more points and a record of 5-3-1, with 16 points in the first 9 games. That is exactly the actual team record in 2017. As better as the team has looked overall, you could argue the difference in results is solely to not giving up any set piece goals. I don't think that explains everything, but it is something.

 

[Midas Mulligan]

Set pieces are weird. For example, and I use what I know here, Liverpool is lambasted for not defending set pieces well. Yet, the numbers I have heard quoted say that statistically, a club should give up 15-18 set piece goals per year. So really, they are just about where they should be.

Let's take some % of those chances as a given. The bigger issue to concern yourself with, I think, is whether or not you "put the opponent to the sword". Set piece goals given up are part of the game. The distribution of those and chance conversion can indeed produce anomalous results.

The weird thing is that I don't think soccer necessarily gives a good and meaningful sample when you look at one club over one season. I intuitively believe there is indeed the greater possibility in this game that you're just really unlucky (or lucky) a whole year. The way to overcome it to be a top team in your league is to be so far outside the mean elsewhere that the variance of outcomes is within what I'll malappropriate as the team's margin of error.

Even Barca and Real occasionally lose to shit teams. But it's so rare because they are so often 2-3 xG better. It's really rare, but almost certain, to be really unlucky in defense and attack sometimes. BUT it's almost impossible to be super, super unlucky in both defense and attack enough to significantly change where you finish in the table.

Again, this is my late night, intuitive interpretation. Subject to both error and revision.

 

[adam]

Set pieces are weird. For example, and I use what I know here, Liverpool is lambasted for not defending set pieces well. Yet, the numbers I have heard quoted say that statistically, a club should give up 15-18 set piece goals per year. So really, they are just about where they should be.

Let's take some % of those chances as a given. The bigger issue to concern yourself with, I think, is whether or not you "put the opponent to the sword". Set piece goals given up are part of the game. The distribution of those and chance conversion can indeed produce anomalous results.

The weird thing is that I don't think soccer necessarily gives a good and meaningful sample when you look at one club over one season. I intuitively believe there is indeed the greater possibility in this game that you're just really unlucky (or lucky) a whole year. The way to overcome it to be a top team in your league is to be so far outside the mean elsewhere that the variance of outcomes is within what I'll malappropriate as the team's margin of error.

Even Barca and Real occasionally lose to shit teams. But it's so rare because they are so often 2-3 xG better. It's really rare, but almost certain, to be really unlucky in defense and attack sometimes. BUT it's almost impossible to be super, super unlucky in both defense and attack enough to significantly change where you finish in the table.

Again, this is my late night, intuitive interpretation. Subject to both error and revision.

I agree with your train of though. But finally having a backline and mids over 6' probably help us somewhat stay below your expected tally compared with last year.

 

[adam]

One-quarter into the season, and NYCFC is off to its best start ever. 16 points in 9 games. Third place in the conference. Sixth place overall, and fifth place overall in Points Per Game. It is by far the team’s best first quarter, and leaves me wondering what the narrative thread of this thread will be this year. In 2015 it was the unlikely and ultimately unsuccessful quest for the playoffs. Last year it started as a playoff quest: After 9 games last year, six of which were at home, the team had only 2 wins, and 10 points in this 9 games for 1.11 PPG. The playoffs were very much an uncertain destination. On June 2 the team had only 18 points after 15 games for a measly 1.2 PPG and a projected 41 points for the year. It was not until a 4-game winning streak that finished on July 6 that the playoffs became a likelihood, and the rest of the season was spent figuring out whether the team could attain the Supporters Shield, First place in the East, or a CCL berth. It failed in all three, but it went down to the last few games of the season before that became certain.

This year the playoffs seem highly likely, but it is also a bit early to tell where anyone stands for first place in the East or overall. Also, due to the CCL format changes, there is no CCL berth to be won during the regular season. So I’m not sure what to measure for in terms go achieving goals, but that won’t stop me going on anyway.

Let’s start with the basics of the East Conference PPG standings:
View attachment 6988

NYC is third here just as in the regular standings, with a projected 60-61 points. It’s a bit astonishing that this puts NYC only in third, but TFC and Orlando got off to very strong starts, and TFC at least figures to challenge for 1st place as well as the shield. Over in the West, but the way, Dalls has a 2.25 PPG that projects out to a cool 76.5 points at season’s end.

The East conference playoff line stands at 43-45 points, which seems close enough. I trust this more than I do the first place projections, because the latter is a product of one team as an outlier while the playoff line is a product of 11 teams combined. NYC will really have to falter if it is to cause any worry about making the playoffs. There are 25 games left and this is how various finishes play out. All lines are in standard MLS W-L-T format:

14 6 5 — 1.88 63 pts
12 7 6 — 1.68 58 pts
11 8 6 — 1.56 55 pts
10 8 7 — 1.48 53 pts
9 9 7 — 1.36 50 pts
8 10 7 — 1.24 47 pts
9 13 3 — 1.20 46 pts
9 15 1 — 1.12 44 pts
8 14 3 — 1.08 43 its

For contrast, here is the same list after 9 games last year.
14-6-5 1.88 57
13-6-6 1.80 55
12-9-4 1.60 50
11-8-6 1.56 49
11-6-8 1.64 51
10-8-7 1.48 47
9-9-7 1.36 44
9-11-5 1.28 42
8-11-6 1.20 40

Now is the time for fun with charts. You might remember that 2016 looked better, but still very similar to 2015 for much of the season. Not this year.
View attachment 6989
Here is a new one:
View attachment 6990


Finishing up with assorted points:

• Remember when Portland and Atlanta ruled the GD table because of piling it on against Minnesota in the first two games? They’re both down to +5, tied for 6th overall, and below our 7+.
• Orlando’s GD is down to 0. They have made very efficient use of their goals, winning close and losing big.
• Some weeks ago someone else, I can neither remember nor find who, noted that NYC was the only team in top 5 in the league for both Goals scored and Goals Against. I have followed that and we are still in that club at spot #4 on both sides. But Toronto has also joined us in that Double Top 5 club.
• As mentioned, last year at this point NYCFC was 2-3-4 with 10 points in 9 games. One notable cause of the disappointing start was a tendency to give up goals on set pieces. Through 9 games NYC conceded 3 such goals, all on free kicks: Perquis for Toronto, Tierney for New England, and Oruro with a heartbreaker for Montreal at minute 90. Every one of those came in a tie, so if you indulge the fallacy that everything else happens exactly the same, NYC would have had 6 more points and a record of 5-3-1, with 16 points in the first 9 games. That is exactly the actual team record in 2017. As better as the team has looked overall, you could argue the difference in results is solely to not giving up any set piece goals. I don't think that explains everything, but it is something.

Great work as always. Who needs a pro writer to cover the team when we have post like this? Should have brought it into particof's office, along with the charts. His jaw would have dropped.

 

[Sabo]

Leading the MLS in passes completed, possession and shots on goal (although I think about 10 of the shots on goal were in the first minute of the second half of the Atlanta game).

According to the SPI used by Nate Silver we have the best offense in the league. https://projects.fivethirtyeight.com/soccer-predictions/mls/?ex_cid=rrpromo

During May we will see whether we have the depth to win the Shield -- the way Ben Sweat has stepped up in Mata's absence is encouraging.

 

[Keith Putnam]

The way to overcome it to be a top team in your league is to be so far outside the mean elsewhere that the variance of outcomes is within what I'll malappropriate as the team's margin of error.

I think you would really enjoy The Double Pivot podcast, if you're not already a listener.

 

[Tom in Fairfield CT]

Leading the MLS in passes completed, possession and shots on goal (although I think about 10 of the shots on goal were in the first minute of the second half of the Atlanta game).

According to the SPI used by Nate Silver we have the best offense in the league. https://projects.fivethirtyeight.com/soccer-predictions/mls/?ex_cid=rrpromo

During May we will see whether we have the depth to win the Shield -- the way Ben Sweat has stepped up in Mata's absence is encouraging.

They have us as one of the worst defenses in the league too. Must use prior years data or something. Only TO has given up fewer goals in the east.

 

[Midas Mulligan]

I think you would really enjoy The Double Pivot podcast, if you're not already a listener.

Thanks for the tip. I actually subscribe and listen when I'm out of other stuff. I want to hear what they say, I just don't like the packaging and delivery a lot of the time. Their voices are sort of annoying, too. But it is full of good stuff.

 

[Gotham Gator]

Leading the MLS in passes completed, possession and shots on goal (although I think about 10 of the shots on goal were in the first minute of the second half of the Atlanta game).

According to the SPI used by Nate Silver we have the best offense in the league. https://projects.fivethirtyeight.com/soccer-predictions/mls/?ex_cid=rrpromo

During May we will see whether we have the depth to win the Shield -- the way Ben Sweat has stepped up in Mata's absence is encouraging.

They have us as one of the worst defenses in the league too. Must use prior years data or something. Only TO has given up fewer goals in the east.

538's ratings typically start the year using the prior season's rating, adjusted for offseason moves and regression to the mean. Then, over the course of the year, that season's results are factored in, and at some point, the prior season and offseason stuff falls out completely. Looking at their methodology for soccer, linked below, it appears to be much the same. My guess is that stuff unrelated to 2017 is at least half the rating at the moment.

Their ratings for soccer are interesting. They rightly note that the score of games alone provides only very limited information. So, they use an equal mix of actual goals scored, goals adjusted for things like time of game and being up/down a man, expected goals by shooting situation, and expected goals by other stats.

https://fivethirtyeight.com/features/how-our-club-soccer-projections-work/

 

[Keith Putnam]

I just don't like the packaging and delivery a lot of the time.

Yeah, their production value is annoyingly inconsistent and often poor, which is a shame because most of the other Howler podcasts sound pretty good. But I always find it interesting. And the guys are endearing. Apparently Mike Goodman lost his job at ESPN Insider due to layoffs just recently.

...okay, that's enough OT from me.

 

[roxfontaine]

Sweet mother I love this forum! I'll be back after I digest all of this.

 

[Gotham Gator]

By the way, I checked 538's predictions for our upcoming games, and they are not bullish to say the least. Dallas has a 53% chance of an outright win on Sunday, we and RSL are equal to win at 37% each on Wednesday, and Orlando has a 45% chance to win (vs. 30% for us) the following weekend.

Their predictions take into account home field advantage as well as days of rest, so both of those work against us.

I ran the various scenarios in a quick spreadsheet. Our expected points from the next three games are 3.43, with the most likely point total being 4 points (20.6%). We are only 2.4% to win all 3 games, 21.0% to win at least 2 games, and 65.6% to win at least 1 game.

 

[Gotham Gator]

Also, I checked this site's measurement of Expected Goals. We are 3rd in the league at Expected Goals/game at 1.53, and tied for first in the league at Expected Goals Allowed/game at 0.9.

 

[Tom in Fairfield CT]

By the way, I checked 538's predictions for our upcoming games, and they are not bullish to say the least. Dallas has a 53% chance of an outright win on Sunday, we and RSL are equal to win at 37% each on Wednesday, and Orlando has a 45% chance to win (vs. 30% for us) the following weekend.

Their predictions take into account home field advantage as well as days of rest, so both of those work against us.

I ran the various scenarios in a quick spreadsheet. Our expected points from the next three games are 3.43, with the most likely point total being 4 points (20.6%). We are only 2.4% to win all 3 games, 21.0% to win at least 2 games, and 65.6% to win at least 1 game.

I think Dallas is good, they've been good and have good players. however, their results this year are not that impressive.

Their wins are against the shitty Galaxy, the shitty revolution, the shitty minnesota united, and the shitty RSL, with one legit win vs SKC in there.

They've drawn the three good teams they've played, SKC, Portland and San Jose.

Its nice to not lose. But I'll take our wins vs atlanta, SJ and columbus over their wins. Its a winnable game based on results for sure.

 

[mgarbowski]

Set pieces are weird. For example, and I use what I know here, Liverpool is lambasted for not defending set pieces well. Yet, the numbers I have heard quoted say that statistically, a club should give up 15-18 set piece goals per year. So really, they are just about where they should be.

Let's take some % of those chances as a given. The bigger issue to concern yourself with, I think, is whether or not you "put the opponent to the sword". Set piece goals given up are part of the game. The distribution of those and chance conversion can indeed produce anomalous results.

The weird thing is that I don't think soccer necessarily gives a good and meaningful sample when you look at one club over one season. I intuitively believe there is indeed the greater possibility in this game that you're just really unlucky (or lucky) a whole year. The way to overcome it to be a top team in your league is to be so far outside the mean elsewhere that the variance of outcomes is within what I'll malappropriate as the team's margin of error.

Even Barca and Real occasionally lose to shit teams. But it's so rare because they are so often 2-3 xG better. It's really rare, but almost certain, to be really unlucky in defense and attack sometimes. BUT it's almost impossible to be super, super unlucky in both defense and attack enough to significantly change where you finish in the table.

Again, this is my late night, intuitive interpretation. Subject to both error and revision.

I mostly agree. In another thread we discussed the MLS article about xGD, in which NYC leads the league. One reason Pythagorean works in baseball is the volume of games. You can run the Pythagorean on a 70 game sample when you're less than halfway through the season. Baseball calls the MLS 34-game season a busy month. I think both games share a high chance factor, which baseball cancels out in the regular season with game volume, but chance rules the baseball playoffs. In MLS, it is a big factor in the comparatively short season and wildly amplifies in the playoffs. As you said, the only way to overcome it is to be so much better you are effectively outside the margin of error, and I don't think that's possible in MLS.
I could also be convinced that giving up set piece goals is equally (or close to equally) a function of chance as it is effort or skill. It could also be that the most effective way to reduce set piece goals against is to reduce conceded corners and free kicks in dangerous situations.

 

[Midas Mulligan]

I mostly agree. In another thread we discussed the MLS article about xGD, in which NYC leads the league. One reason Pythagorean works in baseball is the volume of games. You can run the Pythagorean on a 70 game sample when you're less than halfway through the season. Baseball calls the MLS 34-game season a busy month. I think both games share a high chance factor, which baseball cancels out in the regular season with game volume, but chance rules the baseball playoffs. In MLS, it is a big factor in the comparatively short season and wildly amplifies in the playoffs. As you said, the only way to overcome it is to be so much better you are effectively outside the margin of error, and I don't think that's possible in MLS.
I could also be convinced that giving up set piece goals is equally (or close to equally) a function of chance as it is effort or skill. It could also be that the most effective way to reduce set piece goals against is to reduce conceded corners and free kicks in dangerous situations.

MLB parity is often described as everyone starts the season with 50 wins and 5o losses in the bag. It's the other 62 that matter. (Yes, there's a lot wrong with that empirically, but just making the point I agree with you that regular season there is not nearly so much chance.)

As for the set piece goals, I don't see how your assessment couldn't be right that less chances equals less conversions, unless there's some weird experience value that comes into play in the tail (seems unlikely).

For that matter, I think conceding on corners can be luck, too. So yeah, long winded agreement.

 

[mgarbowski]

538's ratings typically start the year using the prior season's rating, adjusted for offseason moves and regression to the mean. Then, over the course of the year, that season's results are factored in, and at some point, the prior season and offseason stuff falls out completely. Looking at their methodology for soccer, linked below, it appears to be much the same. My guess is that stuff unrelated to 2017 is at least half the rating at the moment.

By the way, I checked 538's predictions for our upcoming games, and they are not bullish to say the least. Dallas has a 53% chance of an outright win on Sunday, we and RSL are equal to win at 37% each on Wednesday, and Orlando has a 45% chance to win (vs. 30% for us) the following weekend.

Their predictions take into account home field advantage as well as days of rest, so both of those work against us.

I've been tracking the 538 rankings for NYCFC. They jumped this week more than ever before across assorted stats such as projected wins, GD and likelihood of making the playoffs, but it's hard to tell if that's more because NYC finally won 2 in a row and 2 rivals, Orlando and RB, lost twice in one week, or whether this year's data is supplanting last year's. I also could never figure out why they ranked NYC so low at the start of this season when the team finished 4th in the league overall last year.
ETA: And 538 will almost never pick an Away team to win for the reasons you state. There are occasional exceptions.

 

[Sabo]

I've been tracking the 538 rankings for NYCFC. They jumped this week more than ever before across assorted stats such as projected wins, GD and likelihood of making the playoffs, but it's hard to tell if that's more because NYC finally won 2 in a row and 2 rivals, Orlando and RB, lost twice in one week, or whether this year's data is supplanting last year's. I also could never figure out why they ranked NYC so low at the start of this season when the team finished 4th in the league overall last year.
ETA: And 538 will almost never pick an Away team to win for the reasons you state. There are occasional exceptions.

We ranked so low because of our defense -- after allowing the most goals and fourth most goals (or whatever the figure was) over the last 2 seasons, we started the season with as bad a defensive rating as anyone in the league. Its gone from 2.0 to 1.9 during the season, but the rating has to be based more on last year than this year. The 538 model has NYCFC as one of the worst defenses in the league and San Jose as one of the best with both allowing just 10 goals this season.

The 538 model relying heavily on prior season performances at the start of a new season should work better in the established European leagues where there is less parity and you can predict who will be top teams, middle teams and bottom teams pretty accurately based on past season performances -- the same isn't true for MLS. NYCFC defense is so much better this year than any previous year that our performance in the current 538 model should be much worse than what will happen this season.

The big changes in the NYCFC rankings were based on our point total compared to our rivals. If we beat Dallas this weekend we will probably jump up to 3rd or 4th in the league. Around midseason, the 538 rankings will become much more accurate than they are now.

 

[adam]

We ranked so low because of our defense -- after allowing the most goals and fourth most goals (or whatever the figure was) over the last 2 seasons, we started the season with as bad a defensive rating as anyone in the league. Its gone from 2.0 to 1.9 during the season, but the rating has to be based more on last year than this year. The 538 model has NYCFC as one of the worst defenses in the league and San Jose as one of the best with both allowing just 10 goals this season.

The 538 model relying heavily on prior season performances at the start of a new season should work better in the established European leagues where there is less parity and you can predict who will be top teams, middle teams and bottom teams pretty accurately based on past season performances -- the same isn't true for MLS. NYCFC defense is so much better this year than any previous year that our performance in the current 538 model should be much worse than what will happen this season.

The big changes in the NYCFC rankings were based on our point total compared to our rivals. If we beat Dallas this weekend we will probably jump up to 3rd or 4th in the league. Around midseason, the 538 rankings will become much more accurate than they are now.

Also roster turnover skews our league stats more than others. With a capped league, we have more turnover from transfer window to window than other leagues (or at least it feels that way, maybe I just pulled it out of my ass). I mean look who's on our roster compared to last season or 2 seasons ago. It's a huge difference.

 

[SoupInNYC]

Also roster turnover skews our league stats more than others. With a capped league, we have more turnover from transfer window to window than other leagues (or at least it feels that way, maybe I just pulled it out of my ass). I mean look who's on our roster compared to last season or 2 seasons ago. It's a huge difference.

I think a lot of that has to do with the cap continuing to increase, but moreso the additional TAM/GAM being used by teams. Its allowing teams around the league to spend additional $ to improve their rosters.

This has two effects: (1) some teams are spending this money and others aren't, causing a shift in quality of teams (2) teams that are spending this money, some are making acquisitions that are turning out better than others (i.e., Mena vs Callens), causing a shift in quality).

I would imagine this "issue" will continue for several years to come as the ability to spend increases throughout the league.