We’ve known for a long, long time in soccer analytics terms (back in the stone age of 2013 at least, with this piece from Ben Pugsley) that game state changes how teams shoot. I haven’t seen many pieces on how it changes how teams pass, so we’ll take a short and sweet look at a few aspects today. This is called part 1 because there is a huge amount to dig into here and I plan on looking into it more at a later date, not because there are more lined up ready to go right after this. If only James paid the talent here at StatsBomb at Bill Simmons rates, we both have the same amount of HBO shows right now. Anyway, onto part one.

 

A reminder of what are becoming my ubiquitous and unübersichtlich (or confusing) zones.

zones

The data for this study comes from the 2014-15 French season. It’s the only place and year I had game-state data for all the passes.

 

Possession

Trailing teams possess the ball more. In this specific season we are looking at, trailing teams played 53% of the passes while leading teams played 47% of the passes. Unsurprisingly, this isn’t because low-possession teams take the lead more often, it’s because teams change their behavior after a score change: 15 teams increased their possession share when going behind while only 4 increased it when taking the lead.

 

Starting at the Back: Long Balls When Ahead

When teams go ahead, they start hitting a lot more long balls. Every single team in the study hit more long balls from zone 7 when leading compared to when tied. 17 of the 20 teams hit fewer long balls when losing compared to tied.

snip20161025_46

Field Tilt: Score or Time A Bigger Factor?

Field tilt here is measuring the share of completions in zones 1 and 2, so the more time spent in front of goal the higher the field tilt.

I was not surprised to find that losing teams have a higher share of their completions in advanced positions: teams average a 20% increase of the share of completions in zones 1 or 2 compared to tie games. I was kind of surprised to see winning teams see a 12% increase as well.

snip20161101_17

 

 

I can intuitively write an explanation backwards around that data (something like: When a team has a lead, games tend to see a lot more of the action played around the goalmouth as trailing teams push forward allowing for quick counters from the losing side), which I was ready to do until I looked at the time factor.

snip20161110_9

So this changes things. Whether your team is winning, tied or losing only makes a small difference until the final minutes but all 3 trend up nearly in lockstep as the game progresses. Whether this is a reflection of game theory where coaches and players purposefully play cautious early on or a reflection of fatigue can’t be determined. Putting on my pundit-hat (and we’ve seen in recent days that broad, sweeping conclusions from data journalist types based on small bits of data generally do not lead anyone wrong) I suspect it’s more of the first and less (though still some) of the second.

 

Up Front: Entering Zone 2 from Zone 3

This was surprising. While the league as a whole had a significantly higher completion % while winning (which is what I expected), only 11 of 20 teams actually had a higher completion % when leading compared to tied. 12 of 20 actually had a higher completion % while losing. This tells me that the difference in total comes from PSG and Lyon leading the most and being the best at this category. Game state itself doesn’t seem to be a pushing factor here, passes entering the danger zone are generally completed at a similarly slightly higher rate when losing or ahead compared to a tie.

snip20161101_19

Overall Totals here:

snip20161101_20

 

Time Factor

Here again we see time playing a big factor. 16 of the 20 teams have a higher completion % on passes from zone 3 to zone 2 in the final 15 minutes compared to the first 15 minutes. Here however we don’t see the steady increase from first 15 to final 15 but more of a sharp jump after the first 15 minutes.

snip20161110_11

 

In the second half you can see a sort of “reset” as the completion % on these entry drops to start the half though not close to the start of the game. Then it quickly rises again throughout the second half.

 

Conclusions And Further Questions

  1. Losing teams possess the ball more. I suspect this is mainly a move by the leading team to play more defensively but haven’t figured out a great way to try and test this.
  2. Losing teams generally play shorter passes out of the back while leading teams hit long balls. Presumably this has to do with the different levels of high pressing, but further investigation is needed for that to be the why.
  3. The field gets tilted toward more passes in the danger zone the longer the game goes on. Time seems to be the dominant factor here.
  4. Completion % entering the danger zone increases as the game goes along and “resets” a bit at halftime before increasing again.

 

The major question to me is if the time factor is most related teams scared of conceding at the start, why do they do this? If strategy changes later in the half, why are teams content to come out of the chute hesitant and “feeling each other out”? Maybe there is a good reason for so much of the game to be played in the midfield or own half to start the game, but the fact it changes so drastically the rest of the game indicates that there’s a chance that teams are not fully prepared so come out too conservative to open up games.

 

Anyway, this is a quick overview of an area that is open for further deep dives to try and reveal how the game chances as the clock ticks and scoreboard changes.

 

 

  • Ketiw93

    Don’t you think that the time factor can be related to better execution of defensive strategy at the beginnings of halves? As the time passes players are getting tired and less focused but at the beginnings they are just after listening instructions from managers.

    • Dustin Ward

      that is certainly possible, and I mentioned fatigue a little bit. not sure how that can be tested fully but it makes sense

  • Chris Mackeprang

    Seems like this analysis would be biased by the fact that not all teams are equally likely to be winning, since some teams are better than others. In an extreme case, if you had a league of only two teams and one of them won 90% of their games, then looking at “teams that are winning” would essentially involve looking only at data from the team that wins most of the time. You might be able to parse out game-state-dependent behavior by looking at individual teams’ behavior within each game (ie within a single game, PSG has say 10% fewer passes per minute during the minutes when they are winning versus during the minutes that game is tied) – kind of like the “same-store sales” concept.