Woo Hoo

[dbminn]

The projections have fulfilled their mission: all of us are talking about it!

 

As Brock has pointed out, the projections are far better than just eyeballing the standings. The math works if you think the assumptions are reasonable. To me, it's a quick and dirty statistical analysis but a defensible starting point for debate.

 

 

 

 

[Thegrin]

All I know is that when each game starts, I feel the starting pitcher has a reasonable chance of throwing a good game.No Kevin Correa or Sam Deduno in this rotation... Thank God.  

[Brock Beauchamp]

Here is a graph of the Twins' postseason chances through the season.

 

Ugh. Look at that nosedive since the ASB, where they reached a 48% chance of making the postseason.

 

[big dog]

 

As Brock has pointed out, the projections are far better than just eyeballing the standings. The math works if you think the assumptions are reasonable. To me, it's a quick and dirty statistical analysis but a defensible starting point for debate.

Which assumptions are those?  I can't make a decision about whether they are reasonable if they don't tell me what they are.

[big dog]

 

.

 

Ugh. Look at that nosedive since the ASB, where they reached a 48% chance of making the postseason.

 

Wow, looks like that was a pretty terrible probability projection...

[Brock Beauchamp]

 

Wow, looks like that was a pretty terrible probability projection...

I think some people are letting the tail wag the dog on this one. The games still need to be played and these probabilities are a snapshot in time. What applied yesterday isn't necessarily applicable to today because scenarios change as games are won or lost.

 

If the Twins go 62-60 through 120 games, they have a relatively low chance to make the postseason after 122 games. That's common sense, right?

 

If the Twins somehow manage to go 30-10 through their final 40 games - a statistical improbability but certainly possible - their postseason chances after 162 games are nearly 100%. Was the projection model wrong because the Twins played well outside their statistical norm? I don't think so.

 

The Twins doing the statistically improbable after 162 games does not invalidate the 122 game number. Teams get lucky/unlucky, succeed/fail against improbable odds, and change their dynamic on a regular basis. After all, that's why the Twins were given that 8% chance to make the postseason in late August... The projection model gave them a small chance of bucking the odds and playing well beyond their previous ability, thereby making the postseason.

 

Are these metrics perfect? Of course not and I haven't seen anyone claim that to be the case... But they're a decent snapshot of the Twins' postseason chances as of August 24, 2015.

[Otto von Ballpark]

 

These playoff probabilities sound really cool, but I have no idea if they are any good.  I think the biggest problem with this is that they claim to be using statistical methods, and they give us broad descriptions of the variables they use, such as "current roster".

You're right, the projected performance of the current roster appears to be the biggest questionable variable, although it's not a complete black box.  Here are Baseball Prospectus rest of season projected standings:

 

http://www.baseballprospectus.com/fantasy/dc/

 

The Twins have the worst projected record in the league, although only by a game (as with any projection, the teams are pretty bunched up).  But that's probably holding the Twins playoff odds back more, as compared to Tampa, and it's probably the biggest source of disagreement on this board.  And unfortunately, unless you pay for a Prospectus subscription, you can't see any more detail on this projection I don't think.

 

That's why I prefer the Fangraphs playoff odds.  They have a "Fangraphs Projections Mode" which is like the Baseball Prospectus version except you can actually click on the team and see how they project each player for free (and the Twins again have the worst rest-of-season projected record in the league).  But in the headers at the top, you can also select "season-to-date stats mode" (which helps the Twins a lot) or coin-flip mode if you just want to know the rough mathematical likelihoods:

 

http://www.fangraphs.com/coolstandings.aspx

 

I find that set-up to be far more informative than the Baseball Prospectus one.  Not sure why MLB partners with Prospectus rather than Fangraphs in this endeavor.

 

And if that's the objection, I agree.  But projected standings by themselves are not worthless, nor do they necessarily require complete documentation and error bars etc. if they are simply extrapolating current season performance or doing coin flips.

[Mike Sixel]

so if a prediction turns out wrong.....it was the process that was wrong, not the random outcome? I don't think people understand what a projection is.......

 

If a plane crashes, one of them, should someone have predicted that, or should we all realize that it is very likely that any one plane will not crash, and therefore fly?

[Squirrel]

 

so if a prediction turns out wrong.....it was the process that was wrong, not the random outcome? I don't think people understand what a projection is.......

 

If a plane crashes, one of them, should someone have predicted that, or should we all realize that it is very likely that any one plane will not crash, and therefore fly?

In that scenario, I think there is always a possibility for a crash while not probable. And no statistic can show that that one particular plane will crash even though the possibilities are there despite the probabilities.

 

It's the same in this. While the possibilities and probabilities might predict a particlar outcome, I go into thinking, well, this could happen, it might even be likely, but until it does, for me it's just an interesting tidbit to discuss/debate but nothing to hang my hat on. There's no way any stat will be truly predictive; but they are definitely something to consider when tempering expectations.

 

 

[big dog]

I don't see why you are casting this as a simplistic process vs. uncertain outcome question and equating it to plane crashes.  There are good methods for predicting plane crashes and bad methods, along with the basic uncertainty of outcome.

 

Of course I understand what a projection is.  I also understand what standard errors are.  Without knowing standard errors, there is no way to distinguish between the process being wrong, and typical variation within outcomes.  They are, by virtue of using statistics, making assumptions about the underlying distribution of the observations, which is always a leap of faith unless you are looking at games of chance or other special cases.  

 

Just because they are wrong doesn't make their process bad...but if your prediction goes from 48% to 7%, it might be fair to question the process a little bit, too.  Perhaps they were assuming that the Twins would continue their insanely high BA w/ RISP, and that Perkins would save 100% of the rest of his opportunities for the remainder of the year?

 

My main point is that these are interesting to discuss, but we don't have enough data to determine whether they are actually very useful.  They might be, but how would we know?  We could play the same season over again 50 times, that would be a start.  It's just fodder for discussion. 

 

[Squirrel]

 

I don't see why you are casting this as a simplistic process vs. uncertain outcome question and equating it to plane crashes.  There are good methods for predicting plane crashes and bad methods, along with the basic uncertainty of outcome.

I was just responding to the post above mine. Or was that the post you were commenting on?

[Mike Sixel]

That's not the argument some are making, though, is it? They are stating that since they aren't "right" or they change, that the whole idea is bunk......

 

I'm merely trying to point out that just because a projection turns out wrong does not make the process or math or science bunk, which seems to be most of the argument (not yours, yours is actually about the process).

[Platoon]

Yeah, this. The simulation takes into account past play and future schedule.

These simulations aren't about pure mathematic possibilities, they're more grounded in reality than that. Hence the 100% chance to make the postseason.

If it's "grounded in reality" what's it doing in a TD post? :)

[Squirrel]

 

That's not the argument some are making, though, is it? They are stating that since they aren't "right" or they change, that the whole idea is bunk......

 

I'm merely trying to point out that just because a projection turns out wrong does not make the process or math or science bunk, which seems to be most of the argument (not yours, yours is actually about the process).

I don't think, outside of one, who has called it 'bunk.' But there are those, such as myself, who see it as I said above and don't take it seriously but as a point of interest. 

[spinowner]

 

so if a prediction turns out wrong.....it was the process that was wrong, not the random outcome? I don't think people understand what a projection is.......

 

If a plane crashes, one of them, should someone have predicted that, or should we all realize that it is very likely that any one plane will not crash, and therefore fly?

This is not a prediction, it's a probability. (Although if the Royals are projected at 100% I'd say that does constitute a prediction.) An 8% probability means that there's about a 1 in 12 chance of the Twins qualifying for the postseason. It doesn't mean the projection is wrong if the Twins make it. 

[Mike Sixel]

A fair distinction on probability vs prediction.......

 

Either way, I think we've beat the topic to death.

[Platoon]

But is all seriousness they likely take a mathematical analytical view of it, which does not account for "winning ugly". And they take a long view, not the view of the last 4 games played against a team in a slumber, or the previous against a team against which all hope is lost. (At least for the Twins) When presented with the scenario that the Twins would win a game in the bottom of the 12th on two 2 base ground ball errors, Mr. Spock said to Kirk, "not logical, Captain"! (Little known fact, Spock was a huge Twins fan!)

[spinowner]

I haven't reviewed these projections. But I'm curious about this: the percentages listed for each team should  add to 500%. If that is not the case their methodology is flawed.

[Otto von Ballpark]

 

My main point is that these are interesting to discuss, but we don't have enough data to determine whether they are actually very useful.  They might be, but how would we know?

Did you see my reply to you?  When you take out the roster/performance component (which is easily done at Fangraphs), these are absolutely very useful.

 

It's like at the start of the season in 1993, the Twins had a basic 1-in-7 chance of making the playoffs.  Then at the beginning of 1994, they had a basic 1-in-5 chance (plus wild card) simply due to realignment.  That's good to know.  Same concept here, with the Fangraphs season-to-date or coin-flip playoff odds, except instead of season-to-season it's in-season and some teams simply have a head start over others.

[Mike Sixel]

 

I haven't reviewed these projections. But I'm curious about this: the percentages listed for each team should  add to 500%. If that is not the case their methodology is flawed.

 

why is that, I'm just curious......for example, the Royals have a 100% chance because they are either going to win the division, or make the WC.....so if you think this because there are 5 teams that make it, the math won't do what you think it should do....

[stringer bell]

Now at 10.7% per the same source. The weekend did increase the playoff probability FWIW.

[Otto von Ballpark]

Someone at Twins Daily should tweet at the Twins and MLB and ask why they don't they use the Fangraphs season-to-date or even coin flip playoff odds.  I can't imagine the Twins would be too happy that MLB is advertising playoff odds on their site that severely downgrade the Twins chances based on PECOTA or BaseRuns or whatever.  I find those things interesting to discuss, but I don't think they should be a major factor in computing playoff odds at this late point in the season.

 

The Twins should be able to advertise a 1-in-5 or even 1-in-4 chance at postseason play right now, based on their overall 2015 performance to date, and instead MLB is effectively saying it is only 1 in 10 because of PECOTA presumably.

[Otto von Ballpark]

 

I haven't reviewed these projections. But I'm curious about this: the percentages listed for each team should  add to 500%. If that is not the case their methodology is flawed.

So rather than take two minutes to add them up yourself and see if this very simple but potentially very damning allegation is true, you just state the allegation instead?  Nice.

 

For the record, the division (100%), wild card (200%), and playoff (500%) percentages all add up like they should.

[Otto von Ballpark]

 

why is that, I'm just curious......for example, the Royals have a 100% chance because they are either going to win the division, or make the WC.....so if you think this because there are 5 teams that make it, the math won't do what you think it should do....

He's right, though, the playoff percentages should add up to 500%, the wild card to 200%, and each division to 100%.  And they do.  So it was a pretty pointless post.

[spinowner]

 

why is that, I'm just curious......for example, the Royals have a 100% chance because they are either going to win the division, or make the WC.....so if you think this because there are 5 teams that make it, the math won't do what you think it should do....

There are 5 postseason berths, so the probabilities add to 500%. Think of it this way: when it gets to the point where 10 teams have been mathematically eliminated each remaining team has a 100% probability. Adding the probabilities gives a sum of 500%.

[spinowner]

 

So rather than take two minutes to add them up yourself and see if this very simple but potentially very damning allegation is true, you just state the allegation instead?  Nice.

 

For the record, the division (100%), wild card (200%), and playoff (500%) percentages all add up like they should.

Thanks for checking. And it's not an allegation it's a mathematical truth.

[Otto von Ballpark]

 

Thanks for checking. And it's not an allegation it's a mathematical truth.

It's an allegation when you suggest they may not be following that simple mathematical truth.

[Squirrel]

 

It's an allegation when you suggest they may not be following that simple mathematical truth.

Let's stop this back and forth. I believe he said that if they didn't there would be a flaw, but he was going on the premise that it was correct. Let's not nitpick back and forth here.

[Mike Sixel]

Huh, thanks, I sit corrected.

[Otto von Ballpark]

 

Let's stop this back and forth. I believe he said that if they didn't there would be a flaw, but he was going on the premise that it was correct. Let's not nitpick back and forth here.

The post was: "I haven't reviewed these projections. But I'm curious about this:the percentages listed for each team should  add to 500%. If that is not the case their methodology is flawed."

 

To me it was the equivalent of:

 

"I haven't checked the roster.  But I'm curious about this: a healthy Aaron Hicks should be in the starting lineup for this team right now. If that's not the case, they are making a serious mistake."

 

When the allegation can be checked so easily, I'm not sure there is any point to such posts other than to be inflammatory and unfairly cast doubt on the target.

 

I'll drop it now, though.  Unless warning points can be exchanged for prizes.