»
Topic: "It's hard to beat the same team twice"
Replies: 14   Last Post: Dec 7, 2019 4:29 PM by: CleverTiger®
This topic has been archived - replies are not allowed.


[ Archives - Tiger Boards Archive ]
Start New Topic
Replies: 14  

"It's hard to beat the same team twice"

[1]
Posted: Dec 7, 2019 2:21 PM
 

It's something that's repeated over and over by the talking heads, so I was curious if the stats bear that out.

I took the data from https://collegefootballdata.com/
and since 1950 there have been 78 times the same 2 teams played each other twice in the same season.

One team won both 44 times (56.4%)
The teams split 34 times (43.6%)

So more often than not one team wins both games, which doesn't much place credence in how hard it is to beat the same team twice.

2020 white level memberbadge-donor-15yr.jpg


But splitting a third of the time is still significant.***

[1]
Posted: Dec 7, 2019 2:24 PM
 




That is true.


Posted: Dec 7, 2019 2:25 PM
 

But it doesn't rise up to the "Oh no, we have to play them again" mantra that the talking heads seem to imply.

2020 white level memberbadge-donor-15yr.jpg


Re: "It's hard to beat the same team twice"

[1]
Posted: Dec 7, 2019 2:27 PM
 

Assuming both teams are close to equal, it is hard to beat the same team twice.


Interesting, and +1 evidence based approach!

[1]
Posted: Dec 7, 2019 2:27 PM
 

What would be even more interesting is knowing the expected results based on differential strengths of the teams - if it was expected to be a 60-40 split on average, this would lend support to the notion. But, assume teams meeting twice are usually doing so in championship games, where a 50/50 split is probably often expected if they’re both good


True.

[1]
Posted: Dec 7, 2019 2:35 PM
 

Not gonna spend the time today, but I suspect that could be determined from the same API using historic lines/rankings data.

2020 white level memberbadge-donor-15yr.jpg


There is a reason why the team won the first meeting

[1]
Posted: Dec 7, 2019 2:29 PM
 

Typically because they are there better team. Therefore, likely to also win again.


Re: "It's hard to beat the same team twice"


Posted: Dec 7, 2019 2:35 PM
 

Coots lose both every Time! Except for big puppie uga

2020 white level member

“Let’s make the others want to be like us” Dabo Swinney!!

The best is yet to come!! I’m so grateful for Clemson Football going into the 2020’s


That would mean they play the same team twice


Posted: Dec 7, 2019 4:25 PM
 

Which would mean they’re in a championship.

Silly Humble... the Coots haven’t done that in a while.

2020 student level member

Re: "It's hard to beat the same team twice"


Posted: Dec 7, 2019 2:43 PM
 

Considering the expected outcome (which assumes one team is superior) would be 100% same team wins, I would say 44% is a huge number for split games. Original premise confirmed.


Well, I don't think thats how that works


Posted: Dec 7, 2019 2:48 PM
 

I think the post earlier mention expected outcomes (percentage win likelihood based on strength) would be an interesting metric. But the odds of the same team winning twice certainly shouldn't be expected to be 100%.

For instance if TEAM A has a(somehow calculated) 80% likelihood of beating TEAM B, then the likelihood of doing it twice (assuming independent trials) would be .8 * .8 = 64%.

2020 white level memberbadge-donor-15yr.jpg


Re: Well, I don't think thats how that works

[1]
Posted: Dec 7, 2019 3:33 PM
 

It is a statistical distribution. The 6% advantage over a completely random outcome (50%) works out to a p value of like 0.13 for this sample size which does not reach the level of statistical significance by research standards (i.e. not likely to be based on something other than chance). If beating a team twice was easy the expected outcome would be very far from 50% with a low p value < 0.05. Since it is not statistically significantly more likely than a coin flip that you win both games, I would say that yes, beating a team twice is "hard."


Re: Well, I don't think thats how that works


Posted: Dec 7, 2019 4:29 PM
 

You can expect normalized data and statistical significance with a sample of at least 40. I don’t see 78 as a bad sample size... but you’re right, that’s a high enough p-value to fail to reject but not on the merit of small sample size.

2020 student level member

Sounds about right. Any teams playing twice


Posted: Dec 7, 2019 2:58 PM
 

should be of a fairly similar caliber since the 3nd meeting would be in the postseason. It also stands to reason that the team that wins the first time is probably a little better than the other team so them winning around 56% makes sense.

This cliche is, of course, ridiculous though. If the one team is much better than the other, beating them twice would be quite easy.

2020 white level member

The problem with stats is that they are dumb.


Posted: Dec 7, 2019 2:59 PM
 

Filter these stats by teams who are both highly ranked.

2020 white level memberbadge-donor-10yr.jpg

Replies: 14  

TIGER TICKETS

FB GAME: Season Tickets
FOR SALE: Looking to sell your tickets? Contact me, will pay IPTAY + ticket cost or more dependent if you have...

Buy or Sell CU Tickets and More in Tiger Tickets!

[ Archives - Tiger Boards Archive ]
Start New Topic
884 people have read this post