Seeding and "Meaning"

I posted this stat on a couple threads elsewhere, but I think it deserves its own thread, because it has implications fans don't usually think about.

Basically, there's been another "no-playoffs" or "8-of-10 is too many" go around where people have been bringing up the concept of "meaningful games."

I personally think most people use that as a term without definition. There's a reasonably sound way of determining what constitutes a meaningful game and what doesn't, in my view, and I don't see too many people using it. So I'm gonna give it a try.

My simple methodology starts with "a game is meaningful if it affects your chances of winning the title." My next assumption is that to win the title, first you get to the title match, and then it's basically anyone's game. I already knew home field in the playoffs was important in MLS, as I once ran the numbers for a couple seasons (playoffs and regular season) and found out that the home team in MLS took about 62% of the points (a draw is 50%, and a win is 100%). [Note: strangely enough, that basically wasn't affected by the shootout. Nor by crowd size, as it applied just as much on weekdays as weekends.]

So, I looked at who has gotten to the title game the past eight years (we're starting to have a decent sample to work with). Here are the MLS Cup Finalists by conference (or division) when we had them.

1 - 75% (12 of 16)
2 - 19% (3 of 16)
3 - 0% (0 of 16)
4 - 6% (1 of 16)

I look at the '97 'Pids as a partial/possible statistical anomaly. That was seven years ago now, and took place in the league's second season. MLS has come a long way since then. So I figure the odds of making the Final are above 75% if you're a 1 seed, above 20% if you're a 2 seed, and below 5% if you're anything else.

So given that, playoff seeding is pretty meaningful, as long as you're not choosing between seeds 3 and 4.

This leads to some surprising conclusions. While columnists like Luis Bueno of the Press Enterprise sees a bunch of meaningless games (http://www.pe.com/columns/bueno/stories/PE_Sports_Local_soccer_column_01.a04b4.html), I see 4 out of 5 being meaningful (you have to stretch it a little to think Burn-Rev is). And where a lot of people might see this weekend's DC-Metro as being nothing more than two teams trying to get in form for the playoffs (and DC still trying to lock up the playoffs), I see a virtual must-win for DC (since they have no shot at the 2 seed if they lose), and a "really ought to win" for Metro (since they'd nearly lock it up if they won, especially if Chicage didn't win). Ditto SJ-Col, in the West.

I also note that in terms of a chance to win the championship, if there were no playoffs, only two games would be very meaningful this week, those involving Kansas City and Columbus (I haven't tried any calculations, but the rest of the league is 5 pts back of those two, or 4 pts and a game in hand, with only 3 to play. You might pass one team that way, but the odds of passing two have got to be pretty small). That means the Fire-DCU game that strikes me as pretty darn meaningful would be useless without a playoff.

On the other hand, note that "meaningfullness" of games would hardly change a whit (this year, and most seasons) if only 4 teams made the playoffs instead of 8. NE would be eliminated instead of "all-but-eliminated-with-virtually-no-chance-even-if-they-made-it."

--

We have a perception problem here. NFL fans tend to perceive seeding to be important. We don't. I think that's a bias on our part that probably comes from being used to no playoff leagues, and it's frightfully divorced from reality, in my view.


Maybe 4 of 10 would be better, as DC and SJ would percieve their games as the "must-wins" they basically are. The cost there, that NE-Dal is meaningless, might be worth the gain. (4 of 10 is also an NFL-like proportion of teams, which might have something to do with all of our perceptions).

But the real difference is not in the regular season, but in an almost wasted first round of the playoffs.

Regardless, any time we have this discussion about "meaningful" games, we ought to think more about what that word means than we do now.

--

Well, there's my theory. Take aim now. . .

Stan

 

Couple ammendments here.

1. Actually, I think (without going into the math too deeply), Metro would actually clinch the #2 if they won this weekend.

2. The dynamics change after this week if you're going 4 of 10, a bunch of teams would be eliminated this week if the results broke wrong for them. Still, you're talking about teams being dead and, in the words of the Crow, being dead but not knowing it yet.

 

Couple other points I just thought of:

1. For anyone who says, "the higher seed only wins because theywee the better team," and not because they have the home field, remember the home team takes 62% of the points. Columbus has only taken 61% of theirs this year, meaning that if you took a hypothetical .500 team at home against CLB, the odds ever so slightly favor the .500 team. And CLB has the best record in the league from that perspective. [KC has taken 57% of the points in their matches]. And a relatively meager 4% difference is gonna be pretty common among close seeds. So you have to suspect that the home team factor is the bigger deal. Soccer has a lot of one-off results.

2. I don't know how the home and away thing will work out yet. Could dilute the results, but probably not completely, since the series tiebreaker is still at the higher seed's field.

Stan

 

That's basically what I was getting at. Letting in 8 of 10 doesn't really dilute the regular season as much as it dilutes the first round of the playoffs. Basically, if you're a 3 or 4, all you get to do is say "we made the playoffs." Look at it carefully, and it's a hollow reward.

Since you support SLC, I should bring up another point. All of this changes a bit when you go to 12 teams, but we have had 12 teams before, and (notwithstanding the small sample size) the results weren't any different. 3 #1s, and 1 #2. (And in that case, it was the SJ Quakes, who played in the same division as the other finalist. Under the conference alignment, they could never have met in the Final.)

 

It would take a huge data set that we just don't have in order to prove it, but I suspect it doesn't matter. A three seed that was also the third best team in the league should still lose to the second seed (and bydefault second best team in the league) at a pretty high rate, if the general tendency to home field advantage holds. (And yes, with divisions, the outcome could easily theoretically be different, but it didn't happen to be the case when we had them).

Also, that difference is *usually* going to be pretty narrow. Last season, for instance, the second seeds (NE Revs and KC Wizards) had winning percentages of .550 and .517, respectively. The three seeds (Metro and CO Rapids) were .517 and .483.

The differences were both .033, or only 3 and change wins every hundred games. Pretty even matchup, if you ask me. And yet, who won? The two higher seeds. Both times. Just like they usually do. (That only has a 28% chance of happening if it was random, but I don't think it was, at least not entirely).*

I would have to suspect home field (and therefore the value of seeding) is reduced with the home and away. But we have only one year of data for it, and the higer seeds went 4 for 4 last year.
----------------------------------------------------------------------------------------

scaryice,

Good data. Notice the huge difference between the playoff rounds where there *is* a home field advantage, and the Cup Final where there *isn't* (except on accident). Last year, the only game that wasn't won by the higher seed in the playoffs was the neutral-site Final. When you throw the Final out, the percentage goes to 75%.

At 25%, the odds of winning two series in a row as a lower seed are supposed to be 6.25%. And how often has it actually happened? 6% of the time. Could be pure coincidence, but it's exactly what the model would predict.
----------------------------------------------------------------------------------------

Bottom Line: Seeding Matters. A lot more than most BS people (and most soccer fans, I think) believe it does.



* - That's a crude calculation. If the difference in winnning percentages is .033, then it was assumed the better record team has a .533 percent chance of winning a given game, assuming neutral turf and that somebody had to win. That would mean a .284 chance of it happening twice in a row.

 

Numerista, That's some good stuff. I will take modest debate, going from the small points to the big point.

Small points:

1. Maybe we're doing our math a different way. I've got San Jose's 2003 record at .616. But we'll go with yours in case I'm wrong.

2. I think you're partially right that in a mythical 'every game at full strength, every game counts for seeding' world, the quakes season winning percentage might have been higher.

But the argument can be made that you're removing the Quakes' worst games retroactively, and that will always flatter the argument a little. Some of that could have been taking the foot off th gas, and some of it could have been that they were were just cold that night.

3. You noted that other teams face callups as well. And though it's probably likely that on average, the higher ranked teams face more callups, it's still quite possible that the Quakes faced a team or two that, on a given night, had *their* difference-maker unavailable.

So while I think an adjustment up from .63 is likely, .75 should be overdoing it. By how much, unfortunately, is just about anyone's guess.

The bigger point:

Given just about any revision upward, though, I think I should concede your argument that the 'better team' factor might outweigh the 'home field' factor in some playoff series. (After all, a 4 seed in a 5 team conference is usually going to have a sub-.500 record).

The Earthquakes were a 1 seed, perhaps even a more dominant 1 seed than usual. Certainly, their record was better than either of the 1s were this year ('at press time'). They faced a 4 seed with a pretty solid 'better team' advantage and a some kind of home field advantage (how much the home field means in the head to ehad is still subject to speculation, I think, but SJ did get the OT at home, and they did score in it).

In the second round, they faced KC, with a (by my math) .516 record. That's a .11 difference (or .10, by my accounting), or slightly less than the home field.

What I'm getting at here is that if the 2003 SJ Quakes had a series where 'better team' was more of a factor than 'home field', that's certainly not representative of all playoff series. And if on certain occasions better team is a bigger deal than home field, it would perhaps weaken, but certainly not destroy my point. If one says better team might have even slightly more important, it's far, far away from saying home field didn't help.

It shouldn't, to my mind, explain the size of the dropoff between the chances of one seeds and two seeds, and really shouldn't explain the dropoff between those of 2 seeds and 3 seeds, who will often have close, and sometimes even identical (except for the tiebreakers) records.

And further, the two factors, whatever their relative importance, are in our playoff system acting in concert. If you think another team is better than you, you probably want to avoid that team in the playoffs as long as you can, but you *definitely* don't want to be playing said better team on the road.

 

crusio,

A lot of what's presented as "common sense" is really just groupthink.

 

My personal preference was always using 6 teams in the playoffs, forcing the 2 and 3 to play each other and giving 1 a bye to rest up for the winner.

In writing this thread, I was going to say something like, "You could invite all ten teams to the playoffs and still have a meaningful regular season, as long as you stacked the deck enough against the lower seeds."

And I think it's true. If you went with a playoff structure of 5 vs 4, winner vs 3, winner versus 2, winner versus 1, single elimination, at the home of the higher seed, with two non gamedays in between each gameday, I think you'd have it. Not that I'm suggesting that (don't think you could sell the tickets), but I haven't yet met the 5 seed I thought had any chance in hell of winning 4 in a row on the road in 10 days (and any team that did might have earned it in my eyes). That's the "Diggstown" scenario applied to soccer.