THE BOOK cover
The Unwritten Book is Finally Written!
An in-depth analysis of: The sacrifice bunt, batter/pitcher matchups, the intentional base on balls, optimizing a batting lineup, hot and cold streaks, clutch performance, platooning strategies, and much more.
Read Excerpts & Customer Reviews

Buy The Book from Amazon


SABR101 required reading if you enter this site. Check out the Sabermetric Wiki. And interesting baseball books.
MOST RECENT ARTICLES
MAIL : You ask | We say

Advanced


THE BOOK--Playing The Percentages In Baseball

<< Back to main

Friday, October 14, 2011

Poz on playoff structures of various leagues

By Tangotiger, 11:04 AM

Joe,

Great article on the playoff structures of the various leagues. 

http://joeposnanski.si.com/2011/10/08/theres-only-one-october/

Let me point out a couple of things to consider:

1. Playing 162 games in MLB is equivalent to playing 82 games in NHL, in terms of establishing which team is better than another.  Basically, the ordinal rankings are as meaningful.  It takes 2 MLB games to tell you what it would take 1 NHL game, basically.

2. Soooo… in the NHL, you allow 16 teams in the playoffs, but they have to go through 4 playoff series of 7 games.  The equivalent in baseball would be to allow 16 teams in the playoffs, and they’d have to play 4 playoff series of 13 or 15 games.

3. Given that you don’t want to have 13 and 15 game playoff series, nor 4 rounds, in MLB, you cut that number of games and series down.  This is the tricky part:

4. How many rounds and games would you need so that the resultant World Series winner is actually the best team in MLB, as much as the resultant Stanley Cup winner is actually the best team in the NHL?

It basically becomes a probability question, one that I’ve not yet done but have always wanted to do. 

Given the limited empirical results, it could very well be that having only 4 teams in the playoffs in a 7-game series is what gives the MLB regular season legitimacy, as the NHL gives it legitimacy with 16 teams.

It seems weird, but 16 teams in the NHL might be equivalent to 4 teams in MLB.

Tom


#1          (see all posts) 2011/10/14 (Fri) @ 11:28

Nice analysis!

Maybe I’ll try simulating this if I have some time this weekend.

Tango, can I assume log5 works the same way for hockey as for baseball?  The “true talent” is what changes with game length, right?  The Log5 formula remains the same.


#2    Tangotiger      (see all posts) 2011/10/14 (Fri) @ 11:29

Continuing also with NBA: as we know, the NBA regular season (or each game) is far too long, in comparison to NHL and MLB.

The NBA season should either be cut in half, or each game should be cut down in two.  There’s simply no drama.

Given that the NBA wants to have 16 teams, then what kind of playoff structure could you have so that you’ve have results more in-line with what NHL gives us?  My guess is that you’d have to have each series a best 2-of-3, with the possibility of the first round being a single-game elimination.  Just a guess.

Just speculating, but NFL is probably as balanced as NHL.  Maybe a bit less.  But, there’s not much more you can do with the NFL.


#3    Tangotiger      (see all posts) 2011/10/14 (Fri) @ 11:30

Phil: right, I would presume so.  But, it’s not a given!  Treat it as an assumption for your purposes.


#4    Tangotiger      (see all posts) 2011/10/14 (Fri) @ 11:38

Phil, you might also be interested in this:

http://www.insidethebook.com/ee/index.php/site/comments/how_many_teams_in_mlb_playoffs_part_2/

In that case, using limited samples, I don’t come to the same conclusion as my expectations in this thread.


#5          (see all posts) 2011/10/14 (Fri) @ 11:42

Tango/3,4: Thanks!

Just realized that for hockey, I may have to do Poisson or something for the stupid extra point on regulation ties.

Drat that NHL!  Not only is the third point bad for the game, but it also makes life worse for sabermetricians!


#6          (see all posts) 2011/10/14 (Fri) @ 12:23

Tango, if you happen to know the SD of NHL team talent, could you post it?  It would save me some time.


#7    Tangotiger      (see all posts) 2011/10/14 (Fri) @ 12:35

Based on my post here:

http://www.insidethebook.com/ee/index.php/site/comments/true_talent_levels_for_sports_leagues/#1

I have the following, as one SD in team true talent level win%:

.060: MLB
.083: NHL
.134: NBA
.143: NFL


#8          (see all posts) 2011/10/14 (Fri) @ 12:38

Thanks, Tango!


#9          (see all posts) 2011/10/14 (Fri) @ 12:38

Does the NHL have a balanced or unbalanced schedule?


#10    Tangotiger      (see all posts) 2011/10/14 (Fri) @ 13:00

The NHL is unbalanced, and has been for most of its history.  I think there was a period of 3 or 4 years where it was balanced back in the 1980s.


#11          (see all posts) 2011/10/14 (Fri) @ 13:10

What would be wild is if the NBA played two 42-game regular seasons (home and home with each team in your conference, and one game against each out-of-conf team) and crowned two champions per year. 

They have the luxury of not being as tethered to weather considerations (baseball) nor tradition (football historically being a cold, fall/winter sport).  Thanks to this, they could strategically start their playoff seasons during sports “down time” to maximize revenue and TV ratings.  So, in this two-season format, the NBA playoffs start in February after the Super Bowl, and conclude in early March. There would be some nice synergy with NCAA Tourney, and then an NBA draft in April, while many players are still fresh in fans’ minds after the Final Four. Then there’s a month-long layoff, and the “2nd season” starts with the recently-drafted players are immediately integrated, while their “Q” rating is still high with the public. “2nd season” runs from May-August, with playoffs in September. The NBA free agency period would be in the month of October, when fans are obsessed with MLB playoff baseball each evening and college/NFL football on Saturdays and Sundays.

This proposal would basically create non-stop basketball action for junkies, since the two off-months - October and April - would feature the NBA draft and the NBA free agency period. 

And it wouldn’t really be extra work for the NBA labor force, who would get two month-long breaks a year, and play the same amount of games. 

Would their revenue go up significantly?


#12    Tangotiger      (see all posts) 2011/10/14 (Fri) @ 13:19

We’ve talked about near year-round play, and I’m all for it.  Basketball is perfectly positioned for it.

In soccer, we have indoor soccer as well, and, having players overlap in the winter indoor league and the summer outdoor league seems ok with me.

So, yeah, I’m all for it in basketball.


#13    Tony B      (see all posts) 2011/10/17 (Mon) @ 11:16

I’m not sure that I see an answer to your question given that, based on a binomial distribution, a best-of-7 series gives a team that is 10% worse a 40% chance of winning the series. Baseball would need a championship series of almost 300 games for us to be 95% certain that the series winner was the better team, again assuming that the better team had a true-p of 55%.

If the final two teams’ winning probability was separated by only 2%, 20 seasons of baseball wouldn’t get us to 90%.

Throwing more teams into the mix in shorter series (merely 5 or 7 games) serves only to make the results more noisy.


#14    VZelepukin      (see all posts) 2011/10/17 (Mon) @ 16:36

How about a stepladder approach for the MLB playoffs.  I don’t remember seeing this idea floated much before, but after doing a little research it turns out Korea and Japan are already using it for their playoffs. 

Round 1 - Wild Card plays #3 division winner (best of 3)

Round 2 - Round 1 winner plays #2 division winner (best of 5)

Round 3 - Round 2 winner plays #1 division winner (best of 7).

This way you still have all the drama and “excitement of the current system, yet the #1 seed gets to the WS 50%+ of the time.  You could even easily expand to a 2nd wild card by adding a 1-game round at the beginning between the 2 wild cards.

There would also be tremendous value in finishing #2 versus #3, so there would be multiple races to follow down the stretch of the regular season.


#15          (see all posts) 2011/10/28 (Fri) @ 19:36

I’m simulating the NFL.  Occasionally, I am getting a winning percentage > 1.000, because that’s only about 3.5 SDs away. 

Should I use odds ratio instead?  An SD of .143 means that 1 SD is .643.  That’s an odds ratio of 1.82.

If I were to simulate a normal odds ratio with a mean of 1 and an SD of 0.82, would that work?  I imagine I might have to use more than 0.82, actually, since the second odds SD contributes less to winning percentage than the first one, and we still want an SD of .143. 

I’ll experiment, but has anyone done this before?


#16          (see all posts) 2011/10/28 (Fri) @ 19:48

OK, here’s what I did.  I got a random normal variable.  If it’s positive, I add it to 1.  If it’s negative, I add it to 1 and take the reciprocal.

So, if my normal variable is 1 (SD), I get +2.  If it’s -1 (SD), I get +0.5.

That becomes the odds ratio.  So 1 SD is .667, and -1 SD is .333.

2 SD is an odds ratio of 3, which is .750.  3 SD is an odds ratio of 4, which is .800.

The SD of winning percentage seems to come out around .140, but the distribution is too skewed.  I’m going to stick to normal for now, but truncate at .900/.100.


Page 1 of 1 pages


Name (required)
E-Mail (optional; WILL be published)
Website (optional)

<< Back to main


Latest...

COMMENTS

Feb 23 01:15
How much should minor leaguers make?

Feb 22 22:31
Not everything you learn in college is true (duh)…

Feb 22 17:27
Would you cut to a regularly scheduled show, if the main event ran long?

Feb 22 17:02
This week in chart failure

Feb 22 16:26
Who’s evaluating the 2011 forecasts this year?

Feb 22 12:21
MLB 2012 Odds: BetOnline

Feb 22 07:11
K minus BB differential or ratio?

Feb 22 01:18
Two players have the same stats: one is much younger.  Which one will be better next year?

Feb 21 14:49
Knuckleball pitchers: all of them

Feb 21 13:57
Proper compensation for Epstein?