How are these rankings calculated?
They are purely results based. The only inputs are game scores and home field advantage. Your rating depends on these results relative to the strength of the opponent in a particular game, and what result you were expected to deliver. These probabilities are determined based on the assumption the teams' quality levels form a normal distribution.
Winning is, of course the most important factor, followed by the score, with diminishing returns for increasing goal differential. The returns also diminish quickly when playing teams far worse than you. In fact, it is quite likely when playing vastly inferior teams, to lose points even with a win. Conversely, it is possible to gain points when losing to a superior team.
Let me know if you have any questions.
These are crazy.
You have a team that is ranked 57th and they have been beaten by the 68th, 71st, 93rd and the 98th ranked teams?
Not puting that team down. Just looking at the rankings of teams I know.
It just doesn't make sense.
Boys or girls? I'd be happy to look into it, but I assure you they aren't crazy. They are statistically sound. They clearly aren't perfect though. There are always a few teams incorrectly placed, but we are running between 85-90% prediction rates by this point in the season on the whole.
^Boys....and yes. They are crazy. You can look and you will see what I'm talking about.
OffTheHook Wrote:These are crazy.
You have a team that is ranked 57th and they have been beaten by the 68th, 71st, 93rd and the 98th ranked teams?
Not puting that team down. Just looking at the rankings of teams I know.
It just doesn't make sense.
Does it really matter when you are talking about teams in the 60's and 90's?
I guess if your on one of those teams it does...lol:popcorn: but seriously, no poll is perfect and I prefer this poll to the coaches poll, which you must pay dues to in order to vote...leaving a very narrow opinion of team rankings.
sstack Wrote:Does it really matter when you are talking about teams in the 60's and 90's?
Well, yes it does. No matter were they are ranked. I remember when this system had a team at #10 and they barely beat the #98 team in the All A. 1-0. In NO system should a #98 EVER compete with a supposed #98. The Rankings just don't make any sense.
OffTheHook Wrote:Well, yes it does. No matter were they are ranked. I remember when this system had a team at #10 and they barely beat the #98 team in the All A. 1-0. In NO system should a #98 EVER compete with a supposed #98. The Rankings just don't make any sense.
Soccer is one of those sports where a big mismatch on paper does not always go that way on the field. Games are won a lot with 1-0 scores or in a shoot out and not uncommon for a mental break down by higher ranked team to allow lower ranked team to pull the upset. I think ranking soccer is harder than most sports.
OffTheHook Wrote:These are crazy.
You have a team that is ranked 57th and they have been beaten by the 68th, 71st, 93rd and the 98th ranked teams?
Not puting that team down. Just looking at the rankings of teams I know.
It just doesn't make sense.
Not only that, Estill County Boys stand at 13-4-3, has one of the top scorers in the state, took#56 GRC to the limit with GRC winning 1-0 on goal with/around 15-20 minutes left, they did lose 7-1 to #37 Madison Southern, #98 Rowan County 5-3 and they aren't even ranked.....I maybe biased saying this but Estill deserves to be on that list more than some of the teams on there that has losing records....
sstack Wrote:Soccer is one of those sports where a big mismatch on paper does not always go that way on the field. Games are won a lot with 1-0 scores or in a shoot out and not uncommon for a mental break down by higher ranked team to allow lower ranked team to pull the upset. I think ranking soccer is harder than most sports.
Close games in pros or International play. Cause all these teams have the best players in the world. BUT, High School Soccer? In KY, the difference is VERY obvious. But, no matter the sport. Like I posted earlier. There is no way a team should be ranked higher than four teams that beat them. Especially when those team are pretty much from the same region and beat up on each other. So yes their records will be different.
I took a detailed look at the Johnson Central boys case study as suggested by OffTheHook, as I often find these types of examples informative for improving the statistical model. This is a very interesting example.
This team started the year ranked 111, and appeared probably to not quite play to their potential the first couple weeks, starting the year 3-3 on August 31, with the aforementioned (by OffTheHook) losses to Rowan County and Ashland Blazer. Their ranking at this point dropped to 115. Rowan County and Ashland Blazer were ranked 82 and 84 at the time, respectively. The other two opponents in question, Russell and Pikeville were ranked 77 and 89 at the time â all higher than Johnson.
Johnson Central then went on a tear, a ten game winning streak over teams currently ranked 84, 87, 89, 91, 105, 110, 110, 136, 144 and an unrankable team from West Virginia. This included a 5-1 win on the road over Pikeville â one of the key opponents in question. This run ended on October 2 with as OffTheHook notes, the following ranks: 57 Johnson Central, 68 Russell, 71 Ashland Blazer, 93 Pikeville, and 98 Rowan County. The big win over Pikeville was a large help to Johnsonâs jump in the rankings, but other big wins over teams in the 80s and 90s would be very similar to wins over the 4 opponents mentioned. (Yes Strikeout King, this does include a win over #89 Estill County ,whom I believe to be more properly rated now.)
OffTheHook noted the rankings at this snapshot in time (10/6) in the post above. Only later, on 10/8 and 10/9, did the Johnson boys promptly lose 2-1 to Pikeville and 3-1 to Russell, dropping their ranking to 69. It was at this point that OffTheHook sited the 4 key losses. Had he cited the 10/13 rankings to go along with the games played through 10/13, then he would have noted the following: 47 Russell, 69 Johnson Central, 73 Ashland Blazer, 84 Rowan County, and 87 Pikeville. Then, Johnson Central would have been a team ranked above only 3 three teams that is lost too, one of which they had also beaten badly on the road. At this point, we are talking about 18 games played with 2 unexplained events. 16 of 18 is better than I claim the system to be, so I will take it. However, as noted above, Johnson Central clearly had a rocky start, and were not favored to win the early games against Rowan and Ashland at the time, so I am not sure I should count them against. At this point most would say, slightly overrated perhaps, but crazy?
Pikeville provides an interesting comparison since the teams played 3 games. We all know that the better team doesnât always win. Johnson won in Pikeville 5-1 and 3-2, Pikeville won in Johnson 2-1. Given no other data, one would have to rank Johnson over Pikeville, and we do.
Please note, a proper analysis here would involve a similar look at Russell, Ashland, Pikeville and Rowan. Then you would have to look at what their opponents did throughout the season, and their opponentsâ opponents. Ainât nobody got time fa datâ. That is exactly what the math does, however. It keeps track of all that stuff. It is not perfect! Groups of sequestered teams that do nothing but play each other â if they produce one dominant team â can produce a team a team ranking that is out of whack. I am not seeing that here. I see a team played poorly early on, and improved. Perhaps they are slightly overrated, next to Russell and Ashland, but without looking into their full schedule and results, I hesitate to even say that.
Again, thanks for pointing out this case to me, I enjoy looking into these kinds of examples from time to time with which to measure the quality of the system against.
Math, I like your rankings. There is no way it will ever be perfect, but I think you have a good working model. I wonder Johnson Centrals ten game tear has any direct correlation with their handful of foreign exchange students...lol..there is no way to be able to predict every nuance.
SoccerMath Wrote:I took a detailed look at the Johnson Central boys case study as suggested by OffTheHook, as I often find these types of examples informative for improving the statistical model. This is a very interesting example.
This team started the year ranked 111, and appeared probably to not quite play to their potential the first couple weeks, starting the year 3-3 on August 31, with the aforementioned (by OffTheHook) losses to Rowan County and Ashland Blazer. Their ranking at this point dropped to 115. Rowan County and Ashland Blazer were ranked 82 and 84 at the time, respectively. The other two opponents in question, Russell and Pikeville were ranked 77 and 89 at the time â all higher than Johnson.
Johnson Central then went on a tear, a ten game winning streak over teams currently ranked 84, 87, 89, 91, 105, 110, 110, 136, 144 and an unrankable team from West Virginia. This included a 5-1 win on the road over Pikeville â one of the key opponents in question. This run ended on October 2 with as OffTheHook notes, the following ranks: 57 Johnson Central, 68 Russell, 71 Ashland Blazer, 93 Pikeville, and 98 Rowan County. The big win over Pikeville was a large help to Johnsonâs jump in the rankings, but other big wins over teams in the 80s and 90s would be very similar to wins over the 4 opponents mentioned. (Yes Strikeout King, this does include a win over #89 Estill County ,whom I believe to be more properly rated now.)
OffTheHook noted the rankings at this snapshot in time (10/6) in the post above. Only later, on 10/8 and 10/9, did the Johnson boys promptly lose 2-1 to Pikeville and 3-1 to Russell, dropping their ranking to 69. It was at this point that OffTheHook sited the 4 key losses. Had he cited the 10/13 rankings to go along with the games played through 10/13, then he would have noted the following: 47 Russell, 69 Johnson Central, 73 Ashland Blazer, 84 Rowan County, and 87 Pikeville. Then, Johnson Central would have been a team ranked above only 3 three teams that is lost too, one of which they had also beaten badly on the road. At this point, we are talking about 18 games played with 2 unexplained events. 16 of 18 is better than I claim the system to be, so I will take it. However, as noted above, Johnson Central clearly had a rocky start, and were not favored to win the early games against Rowan and Ashland at the time, so I am not sure I should count them against. At this point most would say, slightly overrated perhaps, but crazy?
Pikeville provides an interesting comparison since the teams played 3 games. We all know that the better team doesnât always win. Johnson won in Pikeville 5-1 and 3-2, Pikeville won in Johnson 2-1. Given no other data, one would have to rank Johnson over Pikeville, and we do.
Please note, a proper analysis here would involve a similar look at Russell, Ashland, Pikeville and Rowan. Then you would have to look at what their opponents did throughout the season, and their opponentsâ opponents. Ainât nobody got time fa datâ. That is exactly what the math does, however. It keeps track of all that stuff. It is not perfect! Groups of sequestered teams that do nothing but play each other â if they produce one dominant team â can produce a team a team ranking that is out of whack. I am not seeing that here. I see a team played poorly early on, and improved. Perhaps they are slightly overrated, next to Russell and Ashland, but without looking into their full schedule and results, I hesitate to even say that.
Again, thanks for pointing out this case to me, I enjoy looking into these kinds of examples from time to time with which to measure the quality of the system against.
Nice. I just never was a big fan of computer rankings.
You do a good job. At least someone is making an attempt for these kids on the Pitch around here.:Thumbs:
First, in the interest of full disclosure, let me say that I am a good friend of Soccermath, the creator of the Maher Rankings, and am a fan of girls high school soccer only. ( I only have soccer playing daughters).
That being said,there has been some periodic gnashing of teeth about some specific games or team rankings included in each week's addition of the Maher Rankings (hereinafter referred to as MR). There are glitches in the system. Soccermath has never suggested otherwise.
However, his success rate in predicting winners is borderline astounding. He picked the winners of 15 of the 16 regions. He predicted the winner of all 8 semi state game. If the next round goes as predicted, the final four teams will be the top four teams in MR.
There are those of you who might pooh pooh this, saying that many of the games were easy to call. I submit to you that they were easy to call because you had access to MR. There is noone in the state...noone.. who would have a reasonable frame of reference by which to predict the outcome of a game between teams from different areas of the state. Noone has seen enough teams to have that reasonable frame of reference. 15 of 16regional winners!!
Soccer math has given us that frame of reference and all of us high school soccer fans are richer for it.
There will always be situations where the MR math will result in one team or another not being ranked where you or I might want it to be. I have personally raised a couple of specific cases with Soccermath, as have others. In each case he has reviewed the specific case and has responded. It ain't perfect people!!!...but please recognize how amazingly accurate MR is. If you arenot impressed with his accuracy, show me your system that is formulated without using MR that can successfully predict the winner more than 80% of the time IN EVERY SINGLE GAME IN THE STATE.