Maxwell Ratings playoff simulation: Collins Hill best bet in tight AAAAAA field

BY LOREN MAXWELL

The Maxwell ratings were used to simulate this year’s playoffs 1 million times for each of the classifications.  Each team is shown with it rating,  class ranking, strength of its remaining draw, average depth advanced to, and number of times winning the state championship out of the 1 million simulations.  Home advantage is factored into the next round of the playoffs with future rounds simulated at neutral sites.
Each region is also presented as well as the most common semifinal and championship matchups.
AAAAAA
- Despite the loss of Peachtree Ridge, Region 7 continues to beat the remainder of the field in the AAAAAA playoff simulation with the top three teams in terms of championship odds and an 81.1% of boasting the eventual winner.  Collins Hill displaces North Gwinnett at the top of the AAAAAA ratings and pushes past Norcross to the top of the playoff simulation in AAAAAA, winning the title 29.6% of the time.  However, despite the shift, there is still little difference between the three Region 7 teams.
- Four of the remaining five non-Region 7 teams have a greater than 1% chance of winning the title, led by Colquitt County, with an 11.9% chance.  When Colquitt County’s chances are combined with Lowndes, Region 1 has a 13.2% chance of winning the title.
Most Common Matchups:
Semifinals
Norcross – Colquitt County, 629,851 times
North Gwinnett – Collins Hill, 582,630 times
Final
Collins Hill –Norcross, 268,865 times
Team Rank Rating Draw Depth Wins
Collins Hill 1 95.08 87.34 3.6 296,437
Norcross 3 93.17 85.7 3.67 278,786
North Gwinnett 2 94.74 87.77 3.33 236,591
Colquitt County 4 87.65 86.2 3.21 118,850
Hillgrove 5 84.95 91.86 2.43 37,307
Lowndes 12 79.03 91.04 2.26 13,221
McEachern 9 80.53 92.19 2.23 12,864
Dacula 15 74.63 90.54 2.28 5,944
Region Wins
7 – AAAAAA 811,814
1 – AAAAAA 132,071
4 – AAAAAA 50,171
8 – AAAAAA 5,944
AAAAA
- Gainesville’s strong outing against South Paulding put them back on top of the AAAAA ratings and playoff simulation, now with a 39.1% chance of capturing the title.
- Creekside trails Gainesville by only 1 point in the ratings, but has a tougher remaining draw by 4.51 points.  Still, they enjoy a healthy 24.1% chance of winning the title.
- Seven of the eight remaining teams have better than 25 to 1 odds.
- Although a Gainesville-Tucker semifinal is most likely, there is a 40.0% chance Gainesville and Ware County will play a rematch of last year’s championship game in the semifinals.
Most Common Matchups:
Semifinals
Gainesville – Tucker, 496,359 times
Creekside – Kell, 385,295 time
Final
Gainesville – Creekside, 272,634 times
Team Rank Rating Draw Depth Wins
Gainesville 1 82.92 71.81 3.92 391,212
Creekside 2 81.91 76.32 3.31 240,560
Kell 3 75.45 77 2.95 98,833
Tucker 5 74.8 77.45 2.83 84,294
North Paulding 4 75.33 78.51 2.64 77,801
Ware County 6 73.76 77.8 2.66 60,255
Warner Robins 7 72.16 77.81 2.56 42,631
Mundy’s Mill 16 62.42 78.95 2.13 4,414
Region Wins
8 – AAAAA 391,212
4 – AAAAA 244,974
7 – AAAAA 98,833
6 – AAAAA 84,294
5 – AAAAA 77,801
3 – AAAAA 60,255
2 – AAAAA 42,631
AAAA
- No change at the top of the AAAA playoff simulation, where Sandy Creek wins the championship 50.1% of the time.  However, Sandy Creek’s Region 5 rival, Carrollton, and Sandy Creek’s next opponent, Griffin, each have a greater than 10% chance to win as well.
- Sandy Creek, Carrollton, and Alexander combine to give Region 5 an 83.0% chance of boasting the eventual winner.
- With Monroe Area behind them, Carrollton now has the easiest remaining draw.
Most Common Matchups:
Semifinals
Sandy Creek – Wayne County, 463,226 times
Carrollton – Marist, 443,502 times
Final
Sandy Creek – Carrollton, 532,386 times
Team Rank Rating Draw Depth Wins
Sandy Creek 1 93.64 77.05 3.92 501,193
Carrollton 2 86.77 73.99 3.98 326,225
Griffin 3 85.02 79.92 2.65 133,709
Marist 5 68.1 80.36 2.6 11,309
Burke County 7 66.92 80.75 2.59 9,416
Wayne County 9 65.41 78.91 2.69 8,738
Stockbridge 6 67.58 83.4 2.18 7,099
Alexander 10 60.84 79.84 2.38 2,311
Region Wins
5 – AAAA 829,729
4 – AAAA 140,808
3 – AAAA 18,154
6 – AAAA 11,309
AAA
- Buford’s rating received a boost from Gainesville’s strong playoff performance, in turn pushing their chances of winning the AAA title to 71.2%.  Washington County, with a 27.2% chance of being crowned champion, still remains the only other team with more than a token chance.  There is an 80.1% chance they will face each other in the title game.
- Having dispatched Woodward Academy, Buford now has the easiest remaining draw, which is five points lighter than Washington County’s.
- If Buford were to replace Dacula in the AAAAAA bracket, they would have a 34.1% chance of winning the championship.  If they were to replace Mundy’s Mill in the AAAAA bracket, they would have a 67.3% chance of winning the championship.  If they were to replace Alexander in the AAAA bracket, they would have a 53.7% chance to win the championship.
- For the third week in a row, AAA has the team with the lowest playoff odds of any GHSA team in any classification.  In 1,000,000 simulations, Central (Carrollton) won the simulation championship 353 times.  Their most likely path to the title would be through Ringgold, Washington County, and Buford.
- There is a 54.4% chance Buford and St. Pius X will play a rematch of last year’s championship game in the semifinals.
Most Common Matchups:
Semifinals
Washington County – Ringgold, 592,334 times
Buford – St. Pius X, 544,173 time
Final
Buford – Washington County, 809,119 times
Team Rank Rating Draw Depth Wins
Buford 1 99.19 69.52 4.65 712,324
Washington County 2 90.04 74.56 4.02 271,846
Blessed Trinity 5 68.05 81.89 2.2 5,483
St. Pius X 4 68.25 83.28 2.59 5,145
Callaway 6 65.05 84.35 2.47 2,609
Ringgold 9 60.07 78.97 2.72 1,878
Carver (Columbus) 10 57.96 87.28 2.02 362
Central (Carrollton) 14 54.29 79.83 2.35 353
Region Wins
7 – AAA 712,324
3 – AAA 271,846
6 – AAA 10,628
4 – AAA 2,962
5 – AAA 1,878
2 – AAA 362
AA
- Due to their superior rating and an easy draw, Lovett continues to march along as the heavy favorite in AA, winning the simulation 62.0% of the time.
- Amazingly, all eight regions are still represented.
- There is an 8.6% chance Calhoun and Jefferson will play a rematch of last year’s championship game in the semifinals.
Most Common Matchups:
Semifinals
Lovett – Brooks County, 595,847 times
Benedictine – Lamar County, 441,669 times
Final
Lovett – Lamar County, 323,523 times
Team Rank Rating Draw Depth Wins
Lovett 1 81.89 61.7 4.4 620,231
Lamar County 4 71.4 69.59 3.32 130,230
Benedictine 2 72.8 73.24 2.98 117,688
Calhoun 3 72.09 73.57 2.81 91,760
Brooks County 6 60.9 66.91 2.76 24,043
Bowdon 9 56.34 69.35 2.41 6,856
Jefferson 7 58.02 74.05 2.23 5,299
Laney 10 54.31 70.89 2.1 3,893
Region Wins
6 – AA 620,231
4 – AA 130,230
2 – AA 117,688
7 – AA 91,760
1 – AA 24,043
5 – AA 6,856
8 – AA 5,299
3 – AA 3,893
A-Public
- Although Seminole County still leads the simulation with a 30.9% chance of winning the championship, the Class A Public bracket has the most parity of all the GHSA playoff brackets.  All of the remaining teams have a greater than 1% chance and four teams have better than 6 to 1 odds.
- Region 2 and Region 4 each have three teams left and respectively have a 44.0% and 30.9% chance of boasting the eventual winner.
Most Common Matchups:
Semifinals
Seminole County – Irwin County, 560,498 times
Clinch County – Marion County, 465,011 times
Final
Seminole County – Clinch County, 207,522 times
Team Rank Rating Draw Depth Wins
Seminole County 3 46.65 39.3 2.55 309,442
Clinch County 4 43.53 39.45 2.36 214,020
Irwin County 6 42.53 39.31 2.28 184,157
Marion County 7 41.54 40.3 2.12 149,349
Dooly County 14 36.63 41.81 1.54 47,441
Charlton County 13 36.83 42.58 1.4 42,175
Hawkinsville 16 35.21 42.22 1.42 34,115
Johnson County 18 32.74 41.77 1.33 19,301
Region Wins
2 – A 440,352
1 – A 309,442
4 – A 230,905
3 – A 19,301
A-Private
- Still towering over the Class A Private bracket is Eagle’s Landing Christian, winning the simulation 73.0% of the time.
- Mount Pisgah Christian tops the remaining teams with a 12.5% chance of winning the title, but seven of the eight remaining team have a greater than 1% chance of winning.
- There is a 51.9% chance Eagle’s Landing Christian and Prince Avenue Christian will play a rematch of last year’s championship game in the semifinals.
Most Common Matchups:
Semifinals
Eagle’s Landing Christian – Prince Avenue Christian, 518,950 times
Mount Pisgah Christian – Mount Paran Christian, 396,506 times
Final
Eagle’s Landing Christian – Mount Pisgah Christian, 437,290 times
Team Rank Rating Draw Depth Wins
ELCA 1 64.22 42.18 3.48 730,163
Mount Pisgah 2 47.28 44.78 2.44 125,071
Prince Avenue 5 43.47 46.43 1.69 41,744
Mount Paran 9 40.73 48.18 1.74 31,207
Aquinas 11 39.42 48.62 1.71 25,826
Calvary Day 10 39.73 46.99 1.5 19,786
Darlington 8 41.19 49.86 1.14 17,642
Holy Innocents 15 35.43 50.26 1.31 8,561
Region Wins
5 – A 738,724
6 – A 173,920
8 – A 41,744
7 – A 25,826
3 – A 19,786

24 comments Add your comment

Robert

November 26th, 2013
11:14 am

Where are the links to the previous weeks projections? It would be fun to compare. Simulations are always fun to talk about. These are fun talking points for the play-offs.

RobFromNorcross

November 26th, 2013
11:27 am

Suwanee 0wns

November 26th, 2013
12:06 pm

Interesting that Collins Hill’s dominating performances so far in the playoffs have moved them into the Number 1 spot in the Maxwell Rankings. Hard to tidagree with that.

The only thing I can see that might be wrong with the projections is the 29% chance that Burfrod will not be the AAA State Champion. Don’t take the odds. Don’t take the points. Buford will win in AAA.

My guess is that Hillgrove has a better chance against NG than the rankings indicate. The rankings are based upon the entire NG season during which NG’s QB was healthy for most games. NG is doing a pretty good job of QB by committee during the playoffs, but it must weaken the chances for an NG victory against Hillgrove. Especially given that Hillgrove is a good team.

WolfDawg

November 26th, 2013
12:07 pm

Glad the games are decided on the field of play… I am an IT guy, so I do NOT place a lotta time into the “numbers”.. They are fun to look at, but a number cruncher does not have the human element.

ANYWAY – The AAAAAA & AAA Championships Will come from Gwinnett Co… & it does not take a box to tell us that!!

7AAAAAA

November 26th, 2013
12:15 pm

Gonna be a dogfight down the stretch. The CH vs McEachern matchup is very interesting and I’m shocked they have never played before. RobFromNorcross: what is your analysis of this game? McEachern has a lot of really good players and CH just seems to keep rolling with “non superstar” players that very few people have heard about. CH is very well coached and they just seem very efficient. I have not seen McEachern play but do know their RB is a stud and they have some very talented players on Defense.

WolfDawg

November 26th, 2013
2:39 pm

CH is led by a very good QB & plodding O… Much like the Brookwood of old O was, 3, 4, 6 yards a pop on the ground, wearing down & then mauling the opponents D. Then dropping passes to their talented WR’s…

The Eagles D is very good & takes the air outta the sails outta a lotta teams. Very disciplined & fast. Think Buford..

This team has been under the radar until the win over North. Now, they are predicted to take it all & very well could..

I saw them dismantle a very good North team… But other than that, just read a lot about them.

Just my .02

itsonlywrite

November 26th, 2013
3:49 pm

Hopefully we can keep the Championships in AAAAA, AAA, and AA within 985 and HWY 129 again this year!!! :) Northeast Ga football baby!!! Gotta luv it!!!

#ELEPHANTNATION

[...] Here are the latest Maxwell ratings used to project this year’s Georgia high school football state champions via the Atlanta Journal-Constitution: [...]

RobFromNorcross

November 26th, 2013
8:26 pm

Thanks Loren, always interesting info.

Loren Maxwell

November 26th, 2013
9:47 pm

@RobFromNorcross: “Thanks Loren, always interesting info”

It’s an interesting year in AAAAAA for sure. Can’t wait to see how it resolves itself.

RobFromNorcross

November 26th, 2013
10:08 pm

@Loren == Just from the 7-AAAAAA outliers North Gwinnett>>Norcross>>Collins Hill>>North Gwinnett, it is clearly not possible to predict the final outcome this year. It is all about matchups.

With that said, it is highly probable that if teams play as they have been playing then Collins Hill and Norcross are both on a roll. (Norcross has another gear they could shift into- will they do it though?)

North Gwinnett and Colquitt maybe a Field Goal back. Anything can happen at this point. Thats what makes it fun!

Swamp Lizard

November 27th, 2013
8:03 am

@ Loren Maxwell ….. greatly appreciated !! Thanks for all that you & Todd do to keep all of us HS “junkies” in the know.

nourstuff

November 27th, 2013
9:55 am

@Loren Maxwell, several class A fans opine that the strength of class A overall is down this year. Can your Rating #’s be used to compare across years? If so, is there an archive site available? If not, can you tell me if ELCA’s Rating is unusually high this year, or is the rest of the field unusually low-why are they such a prohibitive favorite in your system?

jimb

November 27th, 2013
10:59 am

If Mill Creek would have been in a different region we would have had 5 teams in the quarters, thats the SEC of Ga football

yo

November 27th, 2013
11:52 am

@jimb if Mill Creek would have been in a different region this year can’t figure out how that would put 5 teams from Region 7 in the playoffs much less the quarter. Is that what they are teaching in Region 7 schools these days? :-)

Loren Maxwell

November 27th, 2013
2:12 pm

@Swamp Lizard: “@ Loren Maxwell ….. greatly appreciated !! Thanks for all that you & Todd do to keep all of us HS ‘junkies’ in the know.”

No problem. Thanks for the interest and the support!

@nourstuff: “several class A fans opine that the strength of class A overall is down this year. Can your Rating #’s be used to compare across years? If so, is there an archive site available? If not, can you tell me if ELCA’s Rating is unusually high this year, or is the rest of the field unusually low-why are they such a prohibitive favorite in your system?”

The common reference point between every season is 100 points, which is the hypothetical rating of a team that would win 91.5% of its games in an infinite round robin tournament against every team in the highest classification (this season, AAAAAA). From there, all the ratings are relative to each other.

Anyway, that’s the long answer to yes, the ratings can be used to compare across seasons. Unfortunately, I don’t have an archive up at the moment so you can compare 

However, here are the ratings of a typical Class A team each season from 2000 to 2013, with “typical” being the hypothetical rating of a team that would win 50% of its games in an infinite round robin tournament against every team in Class A:


2000      36.28
2001      27.84
2002      31.51
2003      35.08
2004      30.96
2005      38.66
2006      35.75
2007      31.01
2008      18.86
2009      20.13
2010      27.67
2011      28.93
2012      22.95
2013      16.79

This obviously shows a general downward trend in Class A, especially after 2007.

Another way to look at how the teams are spread out is to look at the same number except what would be required to win 80% of the games:


2000      67.04
2001      51.3
2002      57.68
2003      64.51
2004      58.02
2005      61.7
2006      62.34
2007      54.31
2008      45.35
2009      46.14
2010      57.11
2011      56.94
2012      47.45
2013      41.45

The drop isn’t quite as drastic, but it’s still there, indicating it’s much easier for a team with a 65ish rating in Class A to dominate.

Finally, here’s a chart showing the number of Class A teams that pass a certain rating for each season:


      40  50  60  70  80
2000  20  14   9   5   0
2001  17   4   1   1   0
2002  21  11   2   1   1
2003  21  13   6   3   1
2004  20  12   3   2   0
2005  31  17   7   1   0
2006  28  17   5   2   0
2007  21  10   2   1   0
2008  10   5   0   0   0
2009  14   4   0   0   0
2010  25  14   7   2   0
2011  25  15   6   3   1
2012  14   3   2   1   0
2013   9   1   1   0   0

This is the only year Class A has less than 10 teams rated above 40 and only one team rated above 50.

So, everything in the ratings suggests Class A is not as strong as what it previously had been.

The only other note I would make is that from 2000 to 2011 there were only five classifications. With the advent of Class AAAAAA, the ripple effect would have been to pull some of the larger (and most likely strong) Class A teams up to AA. There is a drop from 2011 to 2012, but I can’t explain why 2008 and 2009 were relatively weak.

Also, you asked a question last week that I responded to, but because I included a link to last year’s results the comment shows it is still “awaiting moderation”.

I’ve included it below in its entirety, but after this more thorough analysis I’ll obviously have the retract my previous conclusion!

@nourstuff: “I’m curious if ELCA’s calculated chance this year is an all-time high in your experience-is this a all-time high across all classes? I don’t know how long you have been running these simulations? Any comment as to whether ELCA’s prohibitive favorite status this year is more due to the strength of the team or the weakness of the field?”

I wrote the program in either 2009 or 2010, but never did the simulation on previous seasons.

Although I could reconstruct the odds for any season, the only ones I have documented are from last season, at:

[Search AJC blog for Maxwell Ratings Odds of Each Playoff Team of Winning State Title, and look for the post on Nov 14, 2012]
and:
[Search AJC blog for: Maxwell Ratings Odds of Each Playoff Team of Winning State Title, and look for the post on Nov 20, 2012]

However, while 76% is certainly a powerful favorite, I’m hesitant to say they are the strongest favorite in GHSA history. The playoffs go back to 1948 and there have been many dominant teams over the years, especially in the lower classifications. Plus the brackets have not always been four or five games, so fewer games can make a significant difference as well since it equates to fewer chances to lose.

My guess would be ECLA’s status is due to their team strength. I don’t show the Class A ratings appreciably better or worse than last year’s.

bucfan

November 27th, 2013
3:27 pm

I’m calling it right now-Colquitt County is gonna win AAAAAA. That D Line is just too good and those RB’s are bad to the bone. I’ll take the upset in AAAAA, North Paulding comes out of nowhere to take it all….

nourstuff

November 27th, 2013
4:56 pm

@Loren Maxwell, thanks for the comprehensive response Re class A-very interesting! The move from 5 to 6 classes would certainly have had a significant impact on the top end-assuming a positive relationship overall between enrollment #s and team strength. Thanks again for weighing in with some data, its always fun to mix that in with the opinions-much to the dismay of many posters!

CoffeeC100

November 28th, 2013
8:44 pm

@ – Loren. Don’t know if you are still checking these. But. If you are could you please publish the latest comparison of all of Georgia’s high school teams by class. I very much enjoy watching how the play-offs continue to effect the relative rankings of non-playoff teams. Thank you for your efforts. They are entertaining, stimulating and though provoking.

Loren Maxwell

November 29th, 2013
9:43 am

@CoffeeC100: Thanks for inquiring. Todd just posted them.

Fred

November 29th, 2013
12:57 pm

@Loren – In your discussion above on the relative strength of class A, does that combine the public and private into an overall class A ranking or could the split have left the public class A that much weaker?

And for the geek in me, how long does it take to run 1 million simulations? I realize the hardware has a lot to do with it but I was wondering is it a minute or two or and hour or two or what? Sorry, I can’t help myself! :-D

Loren Maxwell

November 29th, 2013
2:13 pm

@Fred: “In your discussion above on the relative strength of class A, does that combine the public and private into an overall class A ranking or could the split have left the public class A that much weaker?”

That’s a combined Class A with all public and private schools together.

“And for the geek in me, how long does it take to run 1 million simulations? I realize the hardware has a lot to do with it but I was wondering is it a minute or two or and hour or two or what? Sorry, I can’t help myself! :-D

Actually, it only takes about a second. However, the playoff simulator is not a Monte Carlo simulator where each single game is played over and over. Rather, the program sets up the brackets for 1 million simulations and then advances each team according to the probability of them winning each game.

For example, if Team A and Team B play and Team A should win 80% of the time, then the next round shows Team A in it 800,000 times and team B in it 200,000 times. If the bracket pair is Team C defeating Team D 60% of the time, then Team A should meet Team C 480,000 times, Team A should meet Team D 320,000 times, Team B should meet Team C 120,000 times, and Team B should meet Team D 80,000 times.

The simulator continues through the remainder of the playoff brackets in the same way.

For those interested, this is referred to as a Markov Chain.

However, I am hoping to unveil a full season simulator for next year that is a Monte Carlo simulator where each game of each simulation is in fact played over and over. So far it takes about 3 1/2 hours to complete the regular season, but I haven’t programmed in calculating region standings or the playoffs yet.

The Markov Chain approach works well when the structure is fairly straightforward, such as a playoff bracket. However, the Monte Carlo simulator makes more sense where there can be overwhelming paths a season could follow. Imagine trying to calculate all the different possible region standings between just 8 teams playing 7 region games. Now extrapolate that to 8 regions (with varying numbers of members) in six classifications (plus divide into Class A Public and Class A Private). It’s more practical to just simulate it all 1 million times and the most likely scenarios will surface.

Loren Maxwell

November 29th, 2013
2:14 pm

@Fred: “I realize the hardware has a lot to do with it”

I forgot to add, I use a Samsung Series 7 slate to run the ratings.

Fred

November 29th, 2013
8:04 pm

@Loren – Cool! Good info and I see your point about the probabilities working out even if you don’t run the individual games. Law of Large Numbers is amazing for sure. A Samsung Slate huh? Even though it is an I5 I’m still amazed. I was thinking your man cave had racks of servers and extra air conditioning and now you tell me you run this from your recliner with the computer in your lap! :-D