Morning Coffee Looks At Probabilities

by PB @ BON on Oct 24, 2008 4:32 AM CDT in Quick Hits Comment 13 comments

Morning Coffee Mailbag. BON's favorite (also: only) intern checked in Thursday with an email about the tailback situation:

So are we just accepting this? Chris Ogbannaya is suddenly the answer at tailback? Do you have any insight on the meteoric rise of OG and whether we should expect old OG or new OG for the rest of the season? I just think I haven't seen anyone on the site acting sufficiently surprised by this.

--Andrew R.

First of all, don't be such a stranger, old friend. As to your question, I'll ask readers: Does anyone remember the 2007 spring game? Andrew's email prompted me to go digging through the BON archives and I found my notes from that game. On Ogbonnaya:

For one thing, Chris Ogbonnaya, who is officially "my boy," had a terrific scrimmage, picking up all the yards he was supposed to, finishing runs around the down markers and end zone, and even providing an impressive burst of speed around the corner. I truly think Texas has a Priest Holmes-type runner on their hands, and the sooner the coaches realize it, the better.

Now, to be fair, by the fall of last year, Ogbonnaya appeared to have used the summer of '07 to bulk up considerably (thanks Mad Dog!); as the season got underway in September, he more resembled a fullback than the player who had so tickled my fancy six months prior. Even so, to the extent your question suggests that Ogbonnaya's strong performance in Big 12 play this year is so far out of left field as to warrant more skepticism... well, I'm the wrong guy to ask: I was an OG fan waaaaay before it was cool.

My own biases aside, here's the rub: I think your question is a little off base insofar as its entirely tailback-oriented. I think what you're seeing right now is the Texas offensive line run blocking better than it has in a long time, to such a degree that even if I'm batsh*t crazy with my Ogbonnaya love, it's not such a huge leap of faith to believe he's capable of being effective when the O-Line is performing at a high level.

And right now, it is. So rather than wonder if everyone's insane for not questioning Ogbonnaya's recent explosion, you might just wonder whether I'm insane (you wouldn't be the first) and assume for everyone else that it's not so far fetched to feel comfortable liking the overall developments in the running game.

Setting the tone: Crowd noise. Mack Brown opened his mid-week remarks with a comment not on the Cowboys, but on the crowd:

One thing we'd like to do is encourage our fans to have the same attitude this weekend they had last week. It was just unbelievable the excitement, coming early, there were so many people around the stadium when we drove up two hours before. The fact that so many fans met the team at the north end two hours before, and then just the noise level that our fans had during the ballgame, we felt like made a really big difference.

Honestly, the entire transcript from Mack's Wednesday presser is an interesting read. Among the nuggets:

He considers this year's squad the most unselfish group since he arrived in Austin.
OSU's balance allows them to attack you on the outside with the running game or, if you commit DB support to the run, pick on you with Dez Bryant in one-on-one coverage.
The effort to bring the young receivers along is a concerted one.
Blocking Orakpo in practice prepares our O-Line to block anyone.
Mack's as concerned about the OSU kick return game as I am.

There! Right there! See that?! Just a quick programming note that it'll be Bob Griese, Brad Nessler, and Paul Maguire in the booth calling Saturday's game. Paul Maguire repeatedly will ask you if you can--Look! Right there!--see the TV that you're, you know, watching.

Checking the spreads. The Las Vegas Sports Consultants opened Saturday's line at Texas -10.5, but that hasn't lasted: Now that bettors have loaded up on Texas, the best you can find in Vegas is Texas -12. Jeff Sagarin's predictor ratings forecasts Texas by 13.2. Bruce Feldman is done picking against the Horns (Texas by 20).

Checking the spreads, Part 2. Stewie Mandel has caught a lot of flak over the years from the college blogosphere, but his dip into the SEC Defense vs Big 12 Offense debate is respectably measured:

There's no reason to believe that Big 12 defenses are any worse this year than in the past. If anything, several teams (Texas, Texas Tech, Oklahoma State) appear to have gotten better, albeit still not great. But clearly, the league has never had this many elite offenses at once. It's a perfect confluence of great quarterbacks and dangerous, no-huddle/spread schemes.

Well, almost. Okay, no total free pass for Stew today: He does toss this unsubstantiated bit into the column:

History shows that a team with a dominant defense is far more likely to win a national-title game when facing a team with a powerful offense. In fact, that's exactly how LSU beat Oklahoma in the leagues' last head-to-head title matchup in 2003. That said, Texas showed against USC in '05 that it's possible to win a championship in a shootout.

History actually shows a grand total of 10 BCS Title games which paint no such clear picture. It might well be that a team with a great defense is better positioned to win a national title game than one with a great offense, but we'll need a lot more history to establish whether one is more likely--or "far more likely"--than the other.

Morning Coffee Mailbag, Part 2. I've been looking for a good reason to talk long-term odds for a week now, and the following email gives me the perfect in:

Thought this might be of interest:

Two weeks into their stay at No. 1 in the Associated Press poll, the Texas Longhorns may not want to get too comfortable.

According to the odds makers at BetOnline.com, there's good odds (7-to-2) this will be their second - and final - week atop the AP poll with No. 7 Oklahoma State looming, and even better odds (8-to-5) they last just three weeks in the top spot with a showdown with No. 8 Texas Tech on the horizon.

As for the chance of them making it four straight (with wins over OK State and Tech)? 10-to-1.

--Kate P.

Unfortunately, Kate didn't include a source link, but I did some digging and found this near-identical blurb from Natalie England at the San Antonio Express-News. However, as I tried to work through the material, neither Kate's email nor Natalie's note are clear enough to say definitively what these odds really are. My best guess is as follows:

7:2 odds ($1 to win ~$1.30) Texas loses to Oklahoma State on Saturday
8:5 ($1 to win ~$1.60) Texas loses to OSU or Texas Tech
10:1 ($1 to win ~$4) Texas loses one of its next four (OSU, Tech, Baylor, Kansas)

The way the blurbs are worded makes little sense, so I'm honestly not sure if that's what's being said, but let's set aside the material presented and--using our own numbers--take a look at how Texas' future shakes out from a statistical perspective. Rather than deal with the confusing realm of betting odds, we'll break this down into two groups--probability that Texas beats OSU + Tech and probability that Texas wins its next four. Below I chart side-by-side the probabilities using both the most aggressive and conservative estimates to win each game.

Beat OSU + TECH

Game	TX Win% (High)	TX Win % (Low)
OSU	0.8	0.6
TECH	0.7	0.55
BOTH	56%	33%

Returning to the betting perspective: If you lean towards the aggressive (High) estimates, that 8:5 bet isn't so bad, but if you're less confident in the 'Horns and assign them the lower win percentages, it's a poor value.

Now let's add Baylor and Kansas to the mix:

Game	TX Win% (High)	TX Win % (Low)
OSU	0.8	0.6
TECH	0.7	0.55
BU	0.95	0.9
KU	0.8	0.6
ALL	43%	18%

Nothing changes from the betting perspective (only the confident Longhorn fan should wager on four straight wins), but I really lay all this out so that I can add in one last final step--the potential Big 12 Title Game which would follow winning our next four. (Yes, plus A&M. I refuse to consider the possibility that we lose that game in Austin. Refuse.)

Game	TX Win% (High)	TX Win % (Low)
OSU	0.8	0.6
TECH	0.7	0.55
BU	0.95	0.9
KU	0.8	0.6
B12 TG	0.65	0.55
ALL	28%	10%

Setting allllll that betting junk aside, Texas' odds of winning its next six games are between roughly 10 and 30 percent, depending on how much better you think Texas is than the competition. That is, even if you're Left Column Optimistic, the math ain't good for an undefeated run.

Which demands one and only one approach: For now, let's take things one game at a time. After all, the numbers improve nicely if we win on Saturday. And they rocket upward if Texas were to wake up on November 2nd at 9-0.

Whatever happens... Is this a great time to bleed burnt orange or what?

0 recs | Comment 13 comments |

Story-email Email Printer Print

Comments

History a guide?

In 2005….

Tech came in at #10, was expected to be the biggest conference test Texas would have that season, ended up on the wrong end of 52-17.

OSU wasn’t ranked – in fact, was in last place in the South – but had that annoying history of racing out to huge leads (35-7 in 2004). Down by 19 at one point, Texas won 47-28.

Baylor: 62-0, nuff said.

Kansas: After escaping a Rose-Bowl-killing loss in 2004, Texas had no trouble with KU in 2005, winning 66-14.

Now, that being said, OSU and KU are much, much better than they were in 2004. Even Baylor is. Tech, I would argue, is not. The Kansas game this weekend will prove that. So I will grant you that beating the Cowboys this weekend is going to be a struggle, and will likely be the shootout we expected with Missouri. Playing Tech in Lubbock? Scary ever time. But it was supposed to be a challenge in 2004, and VY put on a show in a 51-21 romp. And playing Kansas in Lawrence is the last obstacle and will no doubt be nerve-wracking if the weather is bad.

After starting the year as one of the BON’er Pessimists (8-4 was my prediction) I’ve now got my Burnt Orange Spectacles firmly in place. I’d say our odds of going 4-0 in this stretch are well north of 50%.

Life is an Occasion. Rise to it.

by patienthornsfan on Oct 24, 2008 5:41 AM CDT reply actions 0 recs

If I weren't broke, I'd take the bet on going all the way to the B12CG

by TXinDC on Oct 24, 2008 8:58 AM CDT reply actions 0 recs

the key is not looking ahead of any other games, and focusing on the current game that week. We cant afford any bloopers like losses against A&M or IMO, a kansas team thats better than last yr

by MJY6087 on Oct 24, 2008 10:33 AM CDT reply actions 0 recs

I can't help it, I have to do this...

The odds of winning the remaining 5 games plus the Big 12 championship calculated the way PB did it assumes that each contest is independent, that is, the outcome of any one game does not affect the outcome of any other. However, the probabilities of victory in any future game are likely not independent of outcomes in past games. Teams learn how to play and get better, and in college football, experience is such a big factor. Texas’ confidence and history are big factors, as well. On the downside, injuries in one game can affect future outcomes. Rather than try to dissect the ridiculous number of different factors affecting possible outcomes, history can provide some insight.

After seven games, Texas has been 7-0 seven times since 1960. In all cases, they finished the regular season unbeaten. In 3 of 7 cases (42 percent) they also finished the season unbeaten.

Texas has been 6-1 or 6-0-1 ten times since 1960. They won the rest of their regular season games 5 times (50 percent) and won the rest of all their games, including bowls 3 times (30 percent).

Texas has been 5-2 or 5-1-1 seventeen times and won the rest of their regular season games 3 times ( 18 percent) and the rest of all games once (6 percent).

This comparison shows the effect of prior history (record through 7 games) on future odds. Like it or not, statistically, the past often predicts the future.

Furthermore, since 2000, against ranked Big 12 teams other than Oklahoma, Texas is now 14-0 at home and 4-1 on the road.

These statistics obviously do not guarantee victories, but they do suggest reason for considerable optimism and better odds than those calculated by PB. Texas being 7-0 means something, i.e., that the likelihood of running the regular season table is close to 100 percent, and the likelihood of winning the national championship is close to 50 percent (let’s leave aside for now the possibility that Texas could go to the MNC with one loss).

This optimism is based on the history that when a Texas team and its inherent talent is good enough to get to 7-0, it is more likely than 30 percent to be good enough to win the rest of its regular season games and even all the rest of its games.

by burnt in ny on Oct 24, 2008 11:36 AM CDT reply actions 0 recs

Not quite

It’s an enticing argument, but simply untrue.

Both the individual games this season and the individual seasons are independent of one another. There’s no getting around that.

What you’re saying, though, isn’t without merit. And though what Texas did in 1963 has no bearing whatsoever on whether, for example, the Longhorns beat Oklahoma State on Saturday, you might well set the odds of Texas winning on Saturday at higher than 80%. If history colors your analysis, I won’t tell you it’s a mistake. But independent they are.

--PB--

by PB @ BON on Oct 24, 2008 2:44 PM CDT up reply actions 0 recs

agree with burnt in ny

Within a reasonable time frame (game to game at the very least, but also from one season to the next), I think it is more than fair to say that performances from one time period to the next are not independent of each other. How could they be? If a team has demonstrated greatness over half a season, there is a strong chance that it will continue to do so.

The probability calculation, which is based on the assumption of independence, requires you to ballpark what the chances are of Texas’ beating the given team in question, and I can’t see a reasonable estimate being made without taking previous games into account. For example, if OU had slaughtered us by 40 points during the RRS, this would lower (on average) peoples’ reasonable expectations of how Texas would perform against a good Mizzou team. Similarly, a lot of people (foolishly, to my mind) weren’t sweating the Mizzou game because Mack’s teams had won the week after the RRS 10 years running. The past was playing a role (as it has to) in thinking about the future.

Pundits discussing the Horns-Pokes game invariably have to lean on considerations of the past – most are predicting a Texas win because of the wins over Oklahoma and Mizzou – the team has played like a national title contender to this point and chances are that it will continue to do so, at least well enough to beat OSU in Austin. Others are predicting a shootout (or a Texas loss, even) based on an emotional letdown that Texas must presumably confront after two very intense games, and OSU’s historical ability to play hard against the Horns. There is simply no other way to make a reasonable prediction without grounding it in the past. Even a “pure” approach of looking only at coaching or position matchups, etc, can’t avoid this as any conclusions drawn will be based on past performance. The only way to do this would be to take each player’s purely exogenous characteristics (e.g. height, weight, speed, etc…) and build a projection like: our O-Line outweighs their D-line by x number of pounds, and is 3 inches taller on average, therefore we have an advantage. Or our corner is 5 inches shorter and slightly slower than their WR, therefore they have the advantage. This is of course part of the discussion, but ultimately not that useful – Dez Bryant is a threat not because he’s 6’3", but because he’s been busting big TD runs all year. Being tall helps, but there is no way I would take a similarly tall Dan Buckner over Dez Bryant if I had to go play a game on Saturday.

Setting boundaries on win probabilities (e.g. a “best case” and “worst case” scenario) addresses this concern to some extent, but it does not get around the fact that even saying UT has between a 60% and 80% chance of beating OSU has to be based on some sort of prior information about the two teams. Why not best case 60% worst case 20%?

I agree that Texas’ chances of winning out are probably less than even, but I still think that this calculation strongly underestimates it, for the reason that burnt in ny points out. By it’s logic, even if Team X has an estimated 80% chance at winning every individual game on its schedule, the odds before the season of it winning 13 games are .8^13, or about .05. Assume you boost the odds to 90% for each game, which is not something that I could reasonably imagine doing for ANY team given opponents they face, not to mention myriad other random factors that effect game outcomes. That’s only a .25 chance to win every game from the outset. Replace any one of those 90% games with a real matchup (e.g. 50% chance of winning) and the odds drop to below 15%. Imagine how badly the poorer teams would fare…

Agreed, though, that reaching back to 1963 is probably not valid for predicting current games in any causally rigorous way, except to show that it is historically possible for really good teams to consistently play at a high level and beat everybody on their schedule.

by longhornglory on Oct 24, 2008 4:42 PM CDT up reply actions 0 recs

independence

PB, I thought the post was thought provoking (clearly, since I spent a lot of time thinking about it) and absolutely love that readers can discuss these issues on this site. My argument does assume that seasons are independent, because I agree it doesn’t make sense that how the Horns did in 1963 predicts how they will do in 2008. That’s why I calculate the odds out of how many independent seasons that had certain within-season results. However, the games within a season cannot be independent because otherwise a team’s record would not predict the likelihood of it winning a future game.

by burnt in ny on Oct 25, 2008 6:21 AM CDT up reply actions 0 recs

I wouldn't make too much of this...

It’s not like our odds are SUPPOSED to be something like 75%. At this point, all teams have a low shot at winning it all. Some are just much lower than others.

I mean, similar to what longhornglory said, if we have, say, an 80% chance to win every game individually (minus A&M), which is pretty high for each game, that all collectively only means we have a 41% chance to win out. Add in a 95% chance to beat A&M, that falls to 39%.

I agree this means that we should realize how hard it is to go undefeated and we should take it one game at a time, but we shouldn’t be surprised by such low numbers.

by TheElusiveShadow on Oct 24, 2008 4:46 PM CDT reply actions 0 recs

Clarification

I’m calculating OSU, Kansas, Tech, and Baylor. And then A&M. That’s not even adding the Big12 title game, if we make it.

by TheElusiveShadow on Oct 24, 2008 4:47 PM CDT up reply actions 0 recs

Sorry for Being a Stranger

Obviously my email was a little too abrupt. Let me lay out some of the thinking behind my exclamation. Ogbannaya has had good spring practices the past two years, but we heard similar good reports about Rashad Bobino and Erick Jackson in the spring of ’07, so we can’t trust reports from spring ball without any corresponding improvement in the fall. You have definitely been lauding Ogbannaya for most of two years, but even you projected in May that OG would be the third string tailback despite his seniority, and before the season defined his role as "anchoring the third down back who can catch and block," not as our primary runner who will relentlessly pounds the holes until they open, which as far as I can tell is what he has become. In the run up to conference play, the solution to the running back problem was either a move to the I-formation with McGee as the featured back, "or as much Fozzy as his body can handle." No plans involved more rushes for Chris Ogbannaya, who has now produced 7.1 yards per rush in 37 rushes during conference play, while he only got 7 carries during non-conference play at 3.14 yards per carry. And last year he averaged only 2.5 yards per carry. Another one of your tenets that I have taken to heart is that career backups don’t usually break out as starters in their senior year.

“The offensive Defensive Player Of The Week was: Erick Jackson. Among the things I can get a little preachy about is the idea that players who haven’t been good enough to start through three or four years in the program are extremely unlikely to be good starters in their final year(s). In programs as deep with talent as Texas, the starters show enough to take the field no later than their junior seasons. If a player doesn’t, and he suddenly finds himself a starter as a true or fifth-year senior? It’s far more likely to indicate a weakness or lack of depth on the team, rather than a leap forward in ability from the player. There are certainly exceptions, but it works as a general rule.”

Obviously OG is an exception.

Improvement on the O-line definitely has a part in the improvement, but if it were the only factor, then Vondrell McGee would have improved his ypc comparably, but he has only improved .3 ypc (4.0 to 4.3) and if you discount the Missouri game (there’s no real reason to other than that it’s a statistical outlier at 9.7 ypc) his average is at 3.6 this season.

Obviously I salute you on your prescience, but you’ll have to forgive me if what I saw from Ogbannaya in 2007 and during the non-conference schedule this year did not prepare me for what he has been capable of against Colorado, Oklahoma, and Missouri, and I was hoping that some light might be shed on it. In a very non-technical and useless phrase I would say he’s really fitting into the holes, but I’m not even sure what that means.

-AR-

by aorist9 on Oct 24, 2008 5:33 PM CDT reply actions 0 recs

"Stats are for losers", but...

I like the RedZone figures:

Texas is scoring 95% of the time of which 83% of attempts are TD’s. That compares to OSU scoring 89% of the time of which only 69% of attempts are TD’s.

Muschamp’s black shirters are holding offenses to only 68% scoring in the zone of which only 48% of attempts are converted to touchdowns. OSU’s defense is giving up 88% and 75% respectively.

The actual number of attempts in the scoring zone is remarkably even on both sides of the ball.

Bend but don’t break appears to be a working formula navigating through this Big 12 beast of a schedule.

by TXStampede on Oct 25, 2008 5:27 AM CDT reply actions 0 recs

these stats reflect something real:

that Oklahoma State tends to score its touchdowns on big plays, many from outside the red zone. When they get in tight, they often end up with field goals. The Cody Johnson/Roy Miller jumbo package is a huge advantage in the game for Texas because it is what provides that 83% TD’s, 95% scoring figure.

by burnt in ny on Oct 25, 2008 6:26 AM CDT up reply actions 0 recs

National Championship Probabilities

You guys might be interested in this. Kind of ties into the Mailbag post from a couple of weeks ago. We put together a probabilities analysis to analyze top teams’ chances of making it to the National Championship.

http://www.atomicteeth.com/euclid-claude/would-you-like-to-play-a-national-championship-game/

The probabilities are based on anticipated lines for upcoming matchups. You can edit the spreadsheet online and see what it does to each team’s odds of a NC berth.

Texas is still in great shape.

by Euclid on Nov 7, 2008 10:32 PM CST reply actions 0 recs