PDA

View Full Version : Do people really believe in Kyle Orton? Jay Cutler


Que
11-04-2009, 12:38 PM
Folks,

Thought I'd share this. My company mines about 400K articles, forums, blogs a day and tracks expressions of sentiment and the context in which the sentiment was expressed for about 33 million entities. Think of entity as a proper noun - in this case two NFL QB's.

One of the things I like to do in demos is find a good timely story to tell and what better story for football followers is the ongoing Jay Cutler vs. Kyle Orton saga. It makes for good theater so I thought I would share with this group.

Kyle Orton
The graph below shows the 3 day moving average of the sentiment expressed for Kyle Orton. NOTE: we're using a median measurement so velocity or quantity of buzz isn't taken into consideration. Both QB's have sufficient quantity of sentiment observations that this isn't too relevant to the measurement.

http://barkingatthemoon.wordpress.com/files/2009/11/orton.jpg

Couple of things to note here. Note the first huge dip. That is reflective of the preseason loss to Chicago (see below). You'll notice that the dip actually occurs a couple of days after the game and that's because we're dealing with a three day moving average. The second dip is pretty alarming too - that's the win against Cincy. So despite a win, the consensus was overwhelming negative against Orton.

The next thing I want to draw your attention to how his sentiment stays marginally positive throughout the next few weeks with occasional small spikes coming from our ensuing victories against the Browns, Raiders and Cowboys. I call this the "silencing of the critics" portion. Basically, people aren't exactly singing his praises but they're not bashing him left and right either.

The next portion of the graph is where people start believing in him. The first spike we see occurs on 10/12 - the day after the victory over NE. It doesn't leap off the graph but you see his sentiment start to stay in the +1 to +2 range where previously in the silencing phase it was in the 0 to +1 range. You see successive spikes of course coming from the NE and SD games.

Then you see a dramatic spike a few days after the SD game. That's the hype leading into the bye week. This sort of confirms something that I've suspected is that when a team or player goes into a bye week on a roll, the hype magnifies (Indy anyone?) briefly as people reflect on the team/player without an upcoming opponent in mind. Think of it as running unopposed in an election. That momentum is fleeting as you become yesterday's story in the off week - note the precipitous drop. That's actually caused by lack of chatter by the way. You start to fall back to earth.

You see it come back up in preparation for the Ravens game and the resulting drop after the loss. But the drop isn't that BIG. It doesn't even break the 0 threshold. This is caused by two things in my opinion: 1) the momentum/belief support he had going into the loss and 2) the reluctance of people to pin the loss on him. What I think is really interesting is whether that loss had come earlier. It is my supposition that if that loss had come in Week 1 or 2, his sentiment would have fallen off the chart down to Kanye West after the VMA awards levels.

Kyle Orton vs Jay Cutler
I won't go into as much detail on this one but a few things to note. Cutler is in light green by the way and Orton is yellow. Look at the divergence early on - that's the result of the preseason match up. Interesting how a head to head match up against a former team can be a lightning rod for sentiment.

http://barkingatthemoon.wordpress.com/files/2009/11/orton-v-cutler.jpg

You can see the impact of the early season loss to GB. There is a sharp decline but nothing compared to Orton's "miracle" win against "lowly" Cincy. That's due to belief support levels.

Moving on, though the loss to ATL takes its toll, Orton and Cutler are occupying the same sentiment band. Downward pressure on Cutler and upwards belief pressure on Orton. Basically, doubters are starting to crop up for Cutler.

Now here it gets really interesting. Look at the last part of the graph. Look at the impact that Cincy loss had. Cutler's sentiment falls to a point that it matches Orton's after the preseason loss. The believers are certainly silenced and the doubters are talking loud and clear. Finally, there is the victory against Cleveland but the sentiment bounce is rather muted don't you think? The doubters aren't silenced and fan boys aren't exactly rushing back in. Remember, this is a 3 day moving average so the numbers have a short memory. Even with a big victory, the reaction is muted and we saw articles and mentions of dumbing down the playbook.

Overall, what does this mean?

Orton - It is true. People were/are starting to truly believe in him and that underlying belief colors their comments. You get natural support levels.

Cutler - This is the opposite. Support levels are eroding quickly and even a big win against CLE doesn't do much to change things. The worm is turning so to speak.

In General - I think this also shows the impact expectations plays on collective sentiment. Orton easily outpaced low expectations while Cutler is failing to meet lofty ones.

RMT
11-04-2009, 12:41 PM
part of the reason for Orton's low expectations and Cutler's lofty ones has to do with each QB's personality

Orton is humble
Cutler is not

gyldenlove
11-04-2009, 12:45 PM
Dude, you should do a moving average on those curves, because your conclusions are pure BS based on those graphs with all the noise in there.

jhns
11-04-2009, 12:52 PM
What kind of company tracks support for Cutler and Orton? What purpose does this serve? Why would anyone pay employees for this?

That is all I could think of when reading this stuff. Also, the only thing this should show you is people like players when they are playing well and hate them when they don't. This isn't rocket science which again brings me back to the questions I asked above.

Que
11-04-2009, 12:53 PM
You mean something like this?

http://barkingatthemoon.wordpress.com/files/2009/11/moving-avg.jpg

I hear you on the noise levels - you should see if when we don't use a 3 day moving avg. Talk about skitzo. You can't see the underlying data in terms of volume but the drop in the moving average at the end for Orton is about 50% due to the lack of post game chatter on the bye week. Assuming a decent showing and victory instead of our bye week you'd see more of plateau and then the resulting drop stemming from the Baltimore loss most likely.

But talking moving averages, look at the Cutler line and the lack of meaningful impact of the CLE win.

Que
11-04-2009, 12:56 PM
What kind of company tracks support for Cutler and Orton? What purpose does this serve? Why would anyone pay employees for this?

That is all I could think of when reading this stuff. Also, the only thing this should show you is people like players when they are playing well and hate them when they don't. This isn't rocket science which again brings me back to the questions I asked above.

We track 33 million different entities. Doesn't make a difference if it is the Verizon Droid, Jon Corzine or Kyle Orton. As for who pays for this sort of information on players & teams - i'll remain silent on that point if you're not smart enough to figure that out.

Jason in LA
11-04-2009, 12:57 PM
Oh no. Not another Cutler vs. Orton thread. Yikes!

jhns
11-04-2009, 12:58 PM
part of the reason for Orton's low expectations and Cutler's lofty ones has to do with each QB's personality

Orton is humble
Cutler is not

I would say it is Cutlers pro bowl, 4550 yard, and 25 TD season that got him high expectations. That and being traded for all that we got for him.

People have low expectations for Orton because he couldn't beat out a horrible QB group in Chicago and has never thrown for even 3000 yards, which is kinda bad.

If you have different expectations for players because of personality, you are what is wrong with football fans today.

gyldenlove
11-04-2009, 12:59 PM
You mean something like this?

http://barkingatthemoon.wordpress.com/files/2009/11/moving-avg.jpg

I hear you on the noise levels - you should see if when we don't use a 3 day moving avg. Talk about skitzo. You can't see the underlying data in terms of volume but the drop in the moving average at the end for Orton is about 50% due to the lack of post game chatter on the bye week. Assuming a decent showing and victory instead of our bye week you'd see more of plateau and then the resulting drop stemming from the Baltimore loss most likely.

But talking moving averages, look at the Cutler line and the lack of meaningful impact of the CLE win.

That is exactly what I mean. Based on those lines your previous conclusions are correct.

jhns
11-04-2009, 01:00 PM
We track 33 million different entities. Doesn't make a difference if it is the Verizon Droid, Jon Corzine or Kyle Orton. As for who pays for this sort of information on players & teams - i'll remain silent on that point if you're not smart enough to figure that out.

So if I figure it out you will talk about it?

Que
11-04-2009, 01:02 PM
Oh no. Not another Cutler vs. Orton thread. Yikes!

LOL. Ya, but this one is with pretty charts. Actually, the one I want to put up is Larry Johnson. Heh! Talk about reaching Kayne West after the VMA or Air France after losing an airliner levels....

Don't have him loaded right now unfortunately.

broncocalijohn
11-04-2009, 01:03 PM
Sounds like we should have bought stock in Orton after the Bears game and sold on high on the Cutler stock. Let me know when Cutler stock is on the bulletin board selling $1 a share. Better make it $.50. Too much risk at $1 to buy.

Que
11-04-2009, 01:08 PM
So if I figure it out you will talk about it?

In generalities sure. Actually, I'll give some color. There are all sorts of markets in this world with all sorts of buyers and sellers. One of those markets happens to be in Vegas. Betting is an emotional thing for the non-fundamental betters out there just like it is for non-fundamental traders in the stock market. Betting pressure on either side of the card drives the line fluctuations to some extent.

DomCasual
11-04-2009, 01:10 PM
I saw the thread title on this, and I was all ready to rip on you for starting another thread, like all the other threads.

Yet, your post turned out to be pretty dang cool. Very nice.

oubronco
11-04-2009, 01:17 PM
Cutler is gone "MOVE ON" DAMN!!!!!!!

Endy
11-04-2009, 01:20 PM
I found that to be very interesting. Of course, I love data.

I wonder how importance to the team factors in. I'd like to see the QB's sentiment line compared with the team's line (if that exists).

The reason I say that is because Jay Cutler is obviously far more important to his team than Kyle Orton is. Orton is a system guy and Cutler is the playmaker. When Orton doesn't play spectacularly well, the team can still win despite his lack of great plays. When Cutler has one of his 3 or 4 INT days, it is tough to come back from that.

Cutler is what Plummer was in 2006 (sorta). Bears fans are figuring out that Cutler will kick ass for 2 or 3 games and then put out a stinker of epic proportions. That might lead you to 11-5 or 12-4, but in the playoffs it only takes one gigantic flop on his part to end the season (see Plummer in the AFC championship game against the Steelers).

Orton's sentiment fell after that preseason game, but so did the rest of the team's sentiment (I'd bet). I wonder if Cutler's does the same? Cutler's been winning games for the Bears, but he's been losing them too. Orton doesn't win or lose games for the Broncos.

Again, I found that to be very interesting. How does the methodology for culling "sentiment" from so many sources work?

Taco John
11-04-2009, 01:27 PM
This kind of stuff is fascinating to me. Thanks for the post Que!

Taco John
11-04-2009, 01:28 PM
I'd love to see the line on Josh McDaniels.

Que
11-04-2009, 01:35 PM
I found that to be very interesting. Of course, I love data.

I wonder how importance to the team factors in. I'd like to see the QB's sentiment line compared with the team's line (if that exists).

The reason I say that is because Jay Cutler is obviously far more important to his team than Kyle Orton is. Orton is a system guy and Cutler is the playmaker. When Orton doesn't play spectacularly well, the team can still win despite his lack of great plays. When Cutler has one of his 3 or 4 INT days, it is tough to come back from that.

Cutler is what Plummer was in 2006 (sorta). Bears fans are figuring out that Cutler will kick ass for 2 or 3 games and then put out a stinker of epic proportions. That might lead you to 11-5 or 12-4, but in the playoffs it only takes one gigantic flop on his part to end the season (see Plummer in the AFC championship game against the Steelers).

Orton's sentiment fell after that preseason game, but so did the rest of the team's sentiment (I'd bet). I wonder if Cutler's does the same? Cutler's been winning games for the Bears, but he's been losing them too. Orton doesn't win or lose games for the Broncos.

Again, I found that to be very interesting. How does the methodology for culling "sentiment" from so many sources work?

Ya, that's a good question and of course to a large extent the two are tied together. Especially so for the QB-Team relationship. I don't have the team data loaded in right now but I'll grab it. A separation between the Bears and Cutler after the CLE win would be the most interesting. You also see a larger separation in general between non QB's and teams - such as a RB and a team.

On data culling, that is an interesting challenge and it really depends on what you're going for. Do negative opinions on the OrangeMane "matter" as much as negatives voiced on ESPN? All depends on what you are trying to get at you know? Right now for this data a observation is an observation regardless of where it comes from. There is some culling inherent to the system as we only process about 400K worth of articles, posts, tweets (don't get me started on that) a day. Very small sites don't make it in but from a representational sampling perspective you are usually OK if you can get enough observations. Super niche entities that don't garner much chatter are frustrating. In football terms, this would be a DJ Williams or a Tony Scheffler. You find some stuff but not really useful for trending purposes.

Pick Six
11-04-2009, 01:38 PM
The conclusions fall right into line with the small sample size (3 Bears fans)with whom I've discussed Cutler's impact with the Bears. Even though the Bears beat Cleveland, there is a strong sense among those people that the Broncos won the trade (at this point).

jhns
11-04-2009, 01:50 PM
In generalities sure. Actually, I'll give some color. There are all sorts of markets in this world with all sorts of buyers and sellers. One of those markets happens to be in Vegas. Betting is an emotional thing for the non-fundamental betters out there just like it is for non-fundamental traders in the stock market. Betting pressure on either side of the card drives the line fluctuations to some extent.

I guess it just sounds like information that should be easy to figure out without research. Oh well, I think the majority of stuff sold is useless. I'm not talking bad about it though. If it gives you work and people buy it, It is at least good for you.

If you use message boards and such, how do you tell if people like them or not? Message boards have always been about the vocal minority. Analysts are all about selling stuff and they do that with controversy. How do you get accurate information from that? I guess it sounds harder than I thought when I think about it.

BroncoMan4ever
11-04-2009, 02:00 PM
part of the reason for Orton's low expectations and Cutler's lofty ones has to do with each QB's personality

Orton is humble
Cutler is not

very good point. Orton got to Denver, happy to be here, in fact didn't he get the news he was being traded to Denver, and he grabbed his **** and caught the 1st flight to Denver. he got here and didn't make himself out to be great, just simply said he was happy to be here, couldn't wait to get to work and that he was going to do his best to be a winner for the franchise, and help the team win.

Jay always had the attitude of being better than everyone on the team or bigger than the franchise, which wasn't true, and just made him seem arrogant and disrespectful

I can honestly say, i am more comfortable with Orton leading the team than i was with Jay. he doesn't have all the abilities Jay does, but he has the mental and leadership abilities that Jay does NOT.

BroncoMan4ever
11-04-2009, 02:02 PM
The conclusions fall right into line with the small sample size (3 Bears fans)with whom I've discussed Cutler's impact with the Bears. Even though the Bears beat Cleveland, there is a strong sense among those people that the Broncos won the trade (at this point).

with each passing week, i am begining to think more and more we raped them in that trade.

this trade has the markings of a Shanahan style prison raping of the Redskins that Mike seemed to pull almost every year for awhile.

cmhargrove
11-04-2009, 02:04 PM
How about a chart on how many guys like this girl's boobs?

Que
11-04-2009, 02:17 PM
I guess it just sounds like information that should be easy to figure out without research. Oh well, I think the majority of stuff sold is useless. I'm not talking bad about it though. If it gives you work and people buy it, It is at least good for you.

If you use message boards and such, how do you tell if people like them or not? Message boards have always been about the vocal minority. Analysts are all about selling stuff and they do that with controversy. How do you get accurate information from that? I guess it sounds harder than I thought when I think about it.

It completely is. That's why starting with a broad foundation of sources, message forums being just one, is critical. We parse the actual text being used to determine sentiment. So from a high level we have to desconstruct sentences, look for sentiment, measure it, and then apply it to an object (in this case Kyle Orton) for which it is being expressed. We do that for 400,000 articles/posts a day. The challenges are numerous - just getting clean data is huge vs. scraping it off of a website much less trying to figure how what in the hell someone is trying to say with "This Is It is the baddest assed movie i've seen in a long time. Woot!". Don't get me started on other slang like FTW or FTL or FKNA. Heh! That's why just parsing Twitter is like tossing darts with a blindfold. Press Releases also piss me off to no end.

The results can be pretty interesting and in some cases alarming. But hey, if it was easy everyone would be doing it... heh!

By the way, I called the NJ Gov. race on Friday. I also called NY-23 correct - but hey I had a 50/50 chance of being right. ;)

oubronco
11-04-2009, 02:17 PM
Oh oh oh mr. Carteeer

Que
11-04-2009, 02:19 PM
How about a chart on how many guys like this girl's boobs?

99% overall.

1% have bigger ones waiting for them at home.

TonyR
11-04-2009, 03:08 PM
Que, are you a fan of Nate Silver and fivethirtyeight.com?

BroncoMan4ever
11-04-2009, 03:11 PM
99% overall.

1% have bigger ones waiting for them at home.

i am a part of that 1% that has bigger ones available to me, however i still like those boobs

NYBronco
11-04-2009, 03:37 PM
Wow! The flailing multi-million dollar QB is making a poor showing versus Orton. Not surprised, the franchise has all kinds of excuses for all his problems and money to match. And Orton just quietly played through his hand injury.

errand
11-04-2009, 03:39 PM
How about a chart on how many guys like this girl's boobs?

That's easy....

xxxxx
x x
x x
x x
x x
xxxxxxxxxxxxx xxxxxxxxxx

Ambiguous
11-04-2009, 04:20 PM
Great thread, interesting stuff. Do you have one on McDaniels?

lazarus4444
11-04-2009, 04:40 PM
Interesting. What company do you work for?

Rock Chalk
11-04-2009, 04:48 PM
It completely is. That's why starting with a broad foundation of sources, message forums being just one, is critical. We parse the actual text being used to determine sentiment. So from a high level we have to desconstruct sentences, look for sentiment, measure it, and then apply it to an object (in this case Kyle Orton) for which it is being expressed. We do that for 400,000 articles/posts a day. The challenges are numerous - just getting clean data is huge vs. scraping it off of a website much less trying to figure how what in the hell someone is trying to say with "This Is It is the baddest assed movie i've seen in a long time. Woot!". Don't get me started on other slang like FTW or FTL or FKNA. Heh! That's why just parsing Twitter is like tossing darts with a blindfold. Press Releases also piss me off to no end.

The results can be pretty interesting and in some cases alarming. But hey, if it was easy everyone would be doing it... heh!

By the way, I called the NJ Gov. race on Friday. I also called NY-23 correct - but hey I had a 50/50 chance of being right. ;)

That's a regular expression nightmare from the programmer's end.

Scraping one website and pulling out the useful data and stripping the bull**** is hard enough, writing regular expressions for 400,000+ websites is a logistical nightmare.

Rock Chalk
11-04-2009, 04:53 PM
For instance, to validate an email address (this isn't even scraping, but this is an example of something simple to do) I use 4 regular expressions
^[^@]{1,64}@[^@]{1,255}$

This just checks to see if there is a single @ sign in the email address.
^(([A-Za-z0-9!#$%&'*+/=?^_`{|}~-][A-Za-z0-9!#$%&'*+/=?^_`{|}~\.-]{0,63})|(\"[^(\\|\")]{0,62}\"))$

This checks the prefix of the email address (looped through each character) to make sure there are no illegal characters.
^\[?[0-9\.]+\]?$

This checks the second part of the email address to make sure its not an IP address
^(([A-Za-z0-9][A-Za-z0-9-]{0,61}[A-Za-z0-9])|([A-Za-z0-9]+))$

This checks the last part of the address, the domain part (i.e., gmail.com) to make sure that it falls within the email rules parameters.

And this is the easy stuff. Based off set rules for a specific piece of data. WHen there are articles from websites using different styles of coding practices and different layouts and such, it becomes a masterful headache for each website.

Unreal that your company scrapes that data and compiles it.

Rock Chalk
11-04-2009, 04:56 PM
This is actually the PHP function I regularly use. This is how much code you need just to validate whether or not what you are reading is an email address.

function validEmail($email)
{
// First, we check that there's one @ symbol, and that the lengths are right
if (!ereg("^[^@]{1,64}@[^@]{1,255}$", $email))
{
return false;
}

// Split it into sections to make life easier
$email_array = explode("@", $email);
$local_array = explode(".", $email_array[0]);
for ($i = 0; $i < sizeof($local_array); $i++)
{
if (!ereg("^(([A-Za-z0-9!#$%&'*+/=?^_`{|}~-][A-Za-z0-9!#$%&'*+/=?^_`{|}~\.-]{0,63})|(\"[^(\\|\")]{0,62}\"))$", $local_array[$i]))
{
return false;
}
}

if (!ereg("^\[?[0-9\.]+\]?$", $email_array[1]))
{
// Check if domain is IP. If not, it should be valid domain name
$domain_array = explode(".", $email_array[1]);

if (sizeof($domain_array) < 2)
{
return false; // Not enough parts to domain
}

for ($i = 0; $i < sizeof($domain_array); $i++)
{
if (!ereg("^(([A-Za-z0-9][A-Za-z0-9-]{0,61}[A-Za-z0-9])|([A-Za-z0-9]+))$", $domain_array[$i]))
{
return false;
}
}
}
return true;
}

baja
11-04-2009, 05:24 PM
So Que with all your Internet savvy can you size you chart to fit the OM format. Thanks. ;D

orange 4 life
11-04-2009, 06:02 PM
Interesting data, and I'm in no way trying to diminish it. Thanks for taking the time to put it together.

That said, this situation is VERY simple IMHO.

Jay Cutler is extremely talented, but he has no heart and no leadership ability. It doesn't take long to realize that's just the way he is and it's not going to change.

Kyle Orton on the other hand doesn't have as much talent in his whole body as crybaby has in that golden right arm (much like Jake Plummer before him), BUT he absolutely DOES have alot more ability than most gave/give him credit for, and more importantly, he has an ABUNDANCE of heart, grit, toughness (remember the ankle last year? The finger this season?), and leadership.

After two and a half years of watching a gutless turd with a wealth of talent, I'll take Orton STRAIGHT UP for Cutler.
Bottom line is if our draft picks make an impact, we'll definately have gotten the better of the trade.
Ask Ryan Leaf or Jeff George what talent without leadership will get you.

Nada.

Bronco Warrior
11-04-2009, 06:20 PM
I just bought a Jay Cutler orange jersey at a Thrift store! Does that make me a bad person? I found a Eddie Mac and a beat up Elway too..that make it better?

Circle Orange
11-04-2009, 06:53 PM
What kind of company tracks support for Cutler and Orton? What purpose does this serve? Why would anyone pay employees for this?

That is all I could think of when reading this stuff. Also, the only thing this should show you is people like players when they are playing well and hate them when they don't. This isn't rocket science which again brings me back to the questions I asked above.

Stalkers. 8')

And don't get me started...that boob pic has to go. Or I swear I'll go on a tear of revenge! :~ohyah!:

Bronco Warrior
11-04-2009, 07:01 PM
We don't post pics like this here it's a family site.

Yeah So you repost it so I can't correct the mistake, kinda hypocritical don't you think? And claim it's a family site when half you clowns don't act like the Manson Family :thumbsup:

Bronco Warrior
11-04-2009, 07:03 PM
Stalkers. 8')

And don't get me started...that boob pic has to go. Or I swear I'll go on a tear of revenge! :~ohyah!:

I removed the original but Baja is tugging to the copy he posted :rofl:

BroncoMan4ever
11-04-2009, 11:13 PM
Me too but those are sweet boobs!
Two Exs that still cfome around. the smaller pic's boobs I paid for so I still get visitation rights! LMAO!

i approve of the purchase :thumbs:

broncocalijohn
11-05-2009, 12:33 AM
BW, your ex kinda looks like Valarie Bertanilli from One Day at a Time or for you to remember THE WIFE OF EDDIE VAN HALEN.....ROCK ON!

BroncoMan4ever
11-05-2009, 02:00 AM
BW, your ex kinda looks like Valarie Bertanilli from One Day at a Time or for you to remember THE WIFE OF EDDIE VAN HALEN.....ROCK ON!

yeah, but with way better t its

Que
11-05-2009, 11:55 AM
TonyR - Oh ya, totally. Nate Silver is the man

McAmbiguous - I do. Let me dig Josh up.

RockChalk - You totally hit it on the head. It is a supreme challenge just to do it but then being able to do for a huge amount of volume is quite the challenge. Heh! The other side is how to mine it without insane query times. Our primary DB usually sits around 330 million rows. Ugh.

s0phr0syne
11-13-2009, 02:43 PM
BUMP for update after last night's fiasco, Plllllllllllleeeeeease!

::strong::