minimum number of games

Scipio Africanus · 06-05-2020, 06:12 PM

Does anyone have any thoughts on what the minimum number of games in a season could be in order to be significant. Let's just say in a division with 5 teams, how many times does each team have to play each other in order to have an idea of which of the 5 teams is the best?

Does anyone have any links to statistical analysis of this? I've seen articles in the past regarding two teams playing each other in 5- and 7-game series. They purport to show that such "short" series are nearly meaningless statistically and two teams would need to play 30+game series in order to confidently state one team is better than the other.

CBeisbol · 06-05-2020, 06:27 PM

Interesting question. It really all depends on how much randomness (or how sure you want to be).

This article about possible shortened seasons gives some numbers÷
https://blogs.fangraphs.com/how-much...horter-season/

Given a full 162 game season the teams with the highest playoff expectation are the Dodgers 99%, Yankees 90%, and Astros 84%. With an 81 game season those numbers fall to 71%, 63% and 59%

On the other end, the teams with the lowest probability of reaching the playoffs are the Tigers, Mariners and Orioles at less than 0.1%. With the 81 game season that increases to 6%, 5% and 2%

Scipio Africanus · 06-05-2020, 06:56 PM

I mostly play fictional and am trying to figure out how many games my new league should have per season. I find 100+game seasons to be a chore because I play out my team's games and, with 162 games, most divisions seem to be well decided by the 2/3 mark, often much earlier.

With 5-team divisions, is each team playing the others 6 times enough to be at all meaningful? Or would that be almost the same as a coin flip? How about 8, or 12? That's the sort of thing I'm trying to find out but I didn't turn up anything online.

Thinking about it does make me wonder how significant baseball tournaments are. If two teams only play each other 1 to 3 times, isn't the victor decided almost randomly?

seven · 06-05-2020, 09:12 PM

The answer to your question is that it depends on how widely spaced are the teams with respect to overall ability. If they are closely bunched, it will take more games to accurately determine their relative positions. If there are big differences, fewer games will be needed. This is a basic question in the field of statistics and relates to what is called statistical power (the probability of detecting true differences). With smaller true differences, a bigger sample size is needed to detect them. This is a very strong relationship.

The interesting question would be, in baseball what is the specific relation between team differences and number of games needed to detect them? It would be somewhat complicated to answer that one.

CBeisbol · 06-11-2020, 05:30 PM

I think Scipio has me on ignore, but here's a blog article that answers their question

https://www.coverthosebases.com/blog...ate-mlb-season

Quote:

There are some pretty powerful insights that can be gained here. First, we see that after only 40 games, half of all team win percentage variability can already be explained. After 60, we’re nearing 72%, and after 80 games, we’re over 78%. The breakdown in tabular form is here:

60 games is the answer that I would give - if I had to just give a number of games

andyhdz · 06-11-2020, 07:24 PM

Interesting since high school, college, and euro leagues arec way less than 60 games.

CBeisbol · 06-11-2020, 07:44 PM

Quote:

Originally Posted by andyhdz

Interesting since high school, college, and euro leagues arec way less than 60 games.

There could be more (or less) variance in those leagues than MLB.
As seven said above, it depends on the talent distributions of the teams.

Also, the linked study isn't great, it's ok.

The author compares the sample seasons to 162 game seasons. There's nothing magic about 162 game seasons. Seasons of 162 games aren't always sufficient to decide which teams are better.

lightgrenade07 · 06-12-2020, 12:26 AM

Quote:

Originally Posted by Scipio Africanus

I mostly play fictional and am trying to figure out how many games my new league should have per season. I find 100+game seasons to be a chore because I play out my team's games and, with 162 games, most divisions seem to be well decided by the 2/3 mark, often much earlier.

It's hard for me to say in terms of matchups between two specific teams. However, in terms of the shortest schedule that's still statistically "long enough," I think you answered your own question right here: 2/3 of a 162-game season, or 108 games.

Reed · 06-12-2020, 07:33 AM

The 162 game schedule was not developed to determine the best team. It was made to maximize profits given the constraints of weather, playoffs, etc.. Depending on the size and talent distribution in the league, you might know who the best team is after just a handful of games. When you have a dominant team like the Yankees in the 20s or the big red machine, you don't need very many games to confirm their dominance. Some seasons where 2 or 3 teams are fighting it out going into the last weekend, then maybe 162 games is not enough to determine the "best" team.

06-05-2020, 06:12 PM	#1
Scipio Africanus Major Leagues Join Date: Jun 2019 Posts: 374	minimum number of games Does anyone have any thoughts on what the minimum number of games in a season could be in order to be significant. Let's just say in a division with 5 teams, how many times does each team have to play each other in order to have an idea of which of the 5 teams is the best? Does anyone have any links to statistical analysis of this? I've seen articles in the past regarding two teams playing each other in 5- and 7-game series. They purport to show that such "short" series are nearly meaningless statistically and two teams would need to play 30+game series in order to confidently state one team is better than the other.

06-05-2020, 06:56 PM	#3
Scipio Africanus Major Leagues Join Date: Jun 2019 Posts: 374	I mostly play fictional and am trying to figure out how many games my new league should have per season. I find 100+game seasons to be a chore because I play out my team's games and, with 162 games, most divisions seem to be well decided by the 2/3 mark, often much earlier. With 5-team divisions, is each team playing the others 6 times enough to be at all meaningful? Or would that be almost the same as a coin flip? How about 8, or 12? That's the sort of thing I'm trying to find out but I didn't turn up anything online. Thinking about it does make me wonder how significant baseball tournaments are. If two teams only play each other 1 to 3 times, isn't the victor decided almost randomly? Last edited by Scipio Africanus; 06-05-2020 at 06:57 PM.

06-05-2020, 09:12 PM	#4
seven All Star Reserve Join Date: Jul 2013 Location: Victoria, Canada Posts: 609	"significant difference" The answer to your question is that it depends on how widely spaced are the teams with respect to overall ability. If they are closely bunched, it will take more games to accurately determine their relative positions. If there are big differences, fewer games will be needed. This is a basic question in the field of statistics and relates to what is called statistical power (the probability of detecting true differences). With smaller true differences, a bigger sample size is needed to detect them. This is a very strong relationship. The interesting question would be, in baseball what is the specific relation between team differences and number of games needed to detect them? It would be somewhat complicated to answer that one.

06-11-2020, 07:24 PM	#6
andyhdz All Star Starter Join Date: Jul 2009 Location: Fresno, CA by way of Texas Posts: 1,754	Interesting since high school, college, and euro leagues arec way less than 60 games. __________________ *************************************** It's your game. Play it how you like it. ***************************************

06-05-2020, 06:27 PM	#2
CBeisbol Banned Join Date: Aug 2019 Location: Ban land in 3...2... Posts: 2,943	Interesting question. It really all depends on how much randomness (or how sure you want to be). This article about possible shortened seasons gives some numbers÷ https://blogs.fangraphs.com/how-much...horter-season/ Given a full 162 game season the teams with the highest playoff expectation are the Dodgers 99%, Yankees 90%, and Astros 84%. With an 81 game season those numbers fall to 71%, 63% and 59% On the other end, the teams with the lowest probability of reaching the playoffs are the Tigers, Mariners and Orioles at less than 0.1%. With the 81 game season that increases to 6%, 5% and 2%

06-12-2020, 07:33 AM	#9
Reed Hall Of Famer Join Date: Oct 2013 Posts: 2,339	The 162 game schedule was not developed to determine the best team. It was made to maximize profits given the constraints of weather, playoffs, etc.. Depending on the size and talent distribution in the league, you might know who the best team is after just a handful of games. When you have a dominant team like the Yankees in the 20s or the big red machine, you don't need very many games to confirm their dominance. Some seasons where 2 or 3 teams are fighting it out going into the last weekend, then maybe 162 games is not enough to determine the "best" team.