MLS reveals new set of standings tiebreakers for '12 season

Of course not.. The playoffs were a complete joke at that point that most teams vieing for the Cup were undeserving.

 

The only teams on a center coast are Salt Lake and Chicago, right?

 

So, I got into a discussion on yesterday's "News-Tuesday, August 14" post about whether wins would be a viable tiebreaker (since theoretically a wins tiebreaker would probably promote attacking soccer better than goals scored). Being contrarian, I argued that wins would not make a viable tiebreaker, as it would never be implemented. Someone else did the legwork, and based on the MLS history from 2002-2011, in 7/10 ties, wins would have been a viable tiebreaker.

The small sample size kind of bothered me, so I decided to run a computer simulation. I did an 18 team league, with a single table, and double round robin. I ran a simulation of 10,000 seasons, in which home win, draw, and away win had roughly equal chances of occurring. I gathered information on 1) Number of seasons with ties, 2) Number of total ties in standings, 3) Number of total ties in standings that were also tied in wins. For the initial simulation, I returned the following: 9998 out of 10000 seasons contained at least one tie in the standings; most contained more than that, as there were a total of about 59,000 ties; and 18,000 of those ties were also tied in wins. So, it would seem that the there would usually be at least one tie, and that in 2/3 of cases, wins would suffice to make the tiebreaker. Discuss.

(Since the program worked like a charm, I set up the simulation to run for 1,000,000 seasons, and ran it overnight. I woke up this morning to find that my computer had auto-restarted in the night, erasing my program. Sad day. However, I am undeterred. Going forward, MLS will likely be 20 teams in two 10 team conference. Ties between teams in different conferences won't matter, so we will have fewer cases where ties exist. I plan to rewrite the simulation program to model this scenario - 37 game season (3xRR in conference, 1xRR against other conference).)

 

Good point. Does anyone have an idea for a hypothetical 34 game schedule with 20 teams and 2 conferences?

 

2x in conference, 1x out of conference, 2x 3 "rivals" = 34 games?

Supposed groupings of regional rivals
NY2-NYRB-DC-PHI
MTL-TOR-NE-CLB
LA-CHV-RSL-COL
HOU-FCD-CHI-SKC
SJ-SEA-POR-VAN

I don't now how the teams would be grouped exactly, but that's my best guess as to how MLS would do it. Though the thought of playing a team 4 times in a season is very bleh (I don't like it at all), it's happened before.

Or they could go the NFL route and make the groups based on the previous years' standings.

 

I doubt SJ would enjoy being left out of the LA group.

More likely the split would be something along the lines of:

1 - Inter-Conference - 10 games
2 - 3 Intra-Conference - 6 games
3 - 6 Intra-Conference - 18 games

So, using Seattle as an example, they'd have 1 game against all the Eastern Conference teams, 2 games against HOU, FCD, and RSL, and 3 games against VAN, POR, LAG, SJE, COR, CHV. That is, of course, assuming Houston goes back to the Western Conference once MLS20 is added and the specific teams selected are purely random. I could also see a situation where the Cascadia teams convince MLS to only schedule two games against each other to make Cascadia Cup easier and then Seattle might do 2 games against SJE to make the Heritage Cup easier.

 

So, while developing a hypothetical 34-game schedule, I went ahead and ran the 37-game schedule.

In 10,000 simulated seasons, 9706 ended in ties (so slightly more rare occurrence), for a total of 32197 ties (many fewer ties), of which 9812 could not be resolved by wins.

I highly doubt we'll get results from the hypothetical unbalanced situation that would lead us to draw a different conclusion about wins as a tiebreaker.

Edit: For those of you who might be interested, I ran the simulation in R, using the attached code.

 

Sorry to get into the conversation so late, but I wanted to say "nice job" with the simulation.

So whaddya know - most wins would solve almost 70% of all tiebreakers. Simple and easy...

 

Chicago and Kansas City would have won the SS in 2000 and 2004, respectively...

 

On the bright side, it is out there.

On the down side, they got it wrong. There are two basic options for tie-breakers, as I see it.

1) Head to head. I don't like it b/c one result can be an aberration. Plus, in MLS, with an unbalanced schedule, it wouldn't be fair.

2) Goal Difference. This is the better reflection of the better team, over the course of the season. Next to points, of course.

Goals scored should only come into it if GD is even.

I'd take most wins as an acceptable TB as well.

 

Totally agree, it's a grave mistake to penalize teams with balanced defense / offsense.

In soccer, an attacking team will not be successful if the goals keep leaking at the back.

A prime example was LA in 2008, who ended up in the bottom 2 despite scoring the most goals.

 

I've never been a fan of most wins as a tiebreaker. You guys seem to forget that wins are already magnified due to the use of 3 points for a win, which heavily discourages ties. If anything, fewest wins (or least losses as I'd like to call it) should be the first tiebreaker. Also, keep in mind a team with less losses would on average have a better goal difference as well which would make it closer to a consensus choice.

 

#7 should be If LA finish in the top 5 they automatically advance to MLS CUP(will also host). That would make everyone happy right.

 

Did you just say you would reward teams for having fewer wins? And fewer losses? So you would have the tiebreaker be most ties? Putting aside the mild irony in that thought, if more points for a win is truly "heavily discouraging" for teams to play for a draw - and I agree that's a good thing - then why not really pile it on and add playoff-berth significance to the wins-total also?

 

Bingo, always better to see teams playing to win rather than settling on ties, it's the purpose of the 3-points-per-win system anway...

 

Why is it ironic? It's just saying that "let's not pile it on" since our system already has a bias. Or to put it in a more simple argument, "If one team is a .500 (or .500 plus) team and another one isn't, then we should break the tie in favour of the 500 team". Seems quite reasonable to me. The way you dismiss it seems a little shortsighted. If instead of calling it fewest losses, if I packaged it as best win-loss difference, would that packaging make you somehow less incredulous?

 

But why would you want to reward a team for more ties at all? Especially if the system is set up to encorage teams to actually go for more wins during the season, what's the point of contradicting that? It's unecessarily complicated. The team with the higher win count will automatically have the higher win percentage anyway.

As for the irony - it only has to do with the words you used in your idea - breaking a tie based on who has more ties...

 

I guess you don't get the issue that the "3 points for a win" does, by some means, overly reward teams that win a lot and lose a lot, so you call it overly complicated as a defense mechanism.

It's not "complicated" to say team with fewest losses. It's actually really easy. I added the comment "or team with most ties" since it's effectively the same thing. You get that right.

So again my question, should a sub 500 team make it ahead of a 500 team, when they are tied on points? Yes or no. It's an intuitive thing. If you feel yes, that's cool. But own that belief. Don't say, "We need to encourage wins"

 

What in the world are you talking about?

You've described your preference as "fewest wins", "least losses", "best win-loss difference", and "team with the most ties". It IS overly complicated.

The system I prefer is "most wins." Pretty simple.

I totally own it. Yes, to me wins are more important than winning percentage. There.

 

Winning isn't everything; it's the only thing.