dntn31 Posted December 21, 2021 Report Share Posted December 21, 2021 Just now, Atticus Finch said: If you view something in a particular way and tinker with the product to reach that outcome what other word is there other than bias? I think this is an insinuation that you have to prove or at least provide more evidence for in order for it to be constructive. I don't think anyone disagrees that Ned has "tinkered" with the algorithm (fwiw so has Massey). What's open to debate is what the goal of that "tinkering" was. Quote Link to comment Share on other sites More sharing options...
dntn31 Posted December 21, 2021 Report Share Posted December 21, 2021 1 minute ago, Atticus Finch said: We just know less about Massey's inputs. I can't remember where I read it but I remember reading two things: 1) He uses the past 2 years of data to seed the ratings for the current year 2) Eventually the starting ratings are filtered out as more data becomes available for the current year (same claim as CalPreps) Quote Link to comment Share on other sites More sharing options...
Atticus Finch Posted December 21, 2021 Report Share Posted December 21, 2021 Just now, dntn31 said: I think this is an insinuation that you have to prove or at least provide more evidence for in order for it to be constructive. I don't think anyone disagrees that Ned has "tinkered" with the algorithm (fwiw so has Massey). What's open to debate is what the goal of that "tinkering" was. He doesn't think that 1-point wins are almost equal to 1-point losses so he tinkers with the math to reach that outcome. He thinks that playoff games are a better indicator of strength than regular season games so he gives more weight to them to reach that outcome. He thinks that head-to-head between a 1-loss team and undefeated team should matter above the normal math and so he tinkers with it to reach that outcome. 1 Quote Link to comment Share on other sites More sharing options...
dntn31 Posted December 21, 2021 Report Share Posted December 21, 2021 Just now, Atticus Finch said: He doesn't think that 1-point wins are almost equal to 1-point losses so he tinkers with the math to reach that outcome. He thinks that playoff games are a better indicator of strength than regular season games so he gives more weight to them to reach that outcome. He thinks that head-to-head between a 1-loss team and undefeated team should matter above the normal math and so he tinkers with it to reach that outcome. These are all things he publishes on his site, it's not really new information. These are all things that are set in advance of the games being played for the current year. Is the insinuation that Ned knows that these yearly "tinkerings" will give CA teams an edge? I mean if he wanted to put his foot on the scales to give CA an edge, he could just do that without actually publishing anything on his site. Or do you think he's trying to be extra sneaky by thinking that people will deem his ratings more trustworthy (despite him actually putting his foot on the scaled behind the scenes) if he only gives the appearance of being more open about how his ratings are produced? Quote Link to comment Share on other sites More sharing options...
golfaddict1 Posted December 21, 2021 Report Share Posted December 21, 2021 13 minutes ago, dntn31 said: Without putting words in GoBigBlack's mouth, I think the point he's driving at here is that human pollsters have "reactionary bias" in that they 1) have bias (impossible not to) and 2) those biases infiltrate whatever logic they end up utilizing to come up with a ranking. Whereas computer algorithm's don't necessarily do that. I don't think "bias" the correct word to use when talking about computer algorithms. Yes, Freeman and Massey both have biases (as all humans do) and they ultimately have to make some decisions about the heuristics they end up using within their algorithms, but at the end of the day these decisions are made well in advance of the games that are played. The question that needs asking is: "are the sum total of these decisions in any way biased towards a specific subset of teams?". It's easy enough to be dismissive of CalPreps (as has been done to the Nth degree on this forum over the past decade), but very few ever seem to address Massey. I do agree that both Massey and Calpreps seem to undervalue Florida (compared to CA and TX), but I haven't really seen anyone provide a convincing argument as to why. If anyone could provide a well-reasoned argument without invoking the words "CalPreps", "bias" or simply trolling I think it would go a long way in figuring out a path forward instead of sitting here constantly chasing our tails. It's pretty clear to me that Freeman jacked up GA's scale recently... it is evidenced by the number of teams that not only rated higher , but the number of teams total in state rated higher. I've shown examples of this in the past where retread scaling in favor or not in favor of states affects their exposure in the top ratings list. It's simple logic of course and it's his show to make that choice. In years past, seemed to me FL was closer to TX and CA with Freeman, but now it's really the top 3 with GA at the head of the table... FL and OH well back in 4-5. He may have valid points for doing this. It's his show, his choice... but imho his recent ratings system as you indicted prior is out of wack. If you look at the single season top ratings by CP, you'll notice an easy trend and that is the more recent years are more prevalent in the top of the ratings. You'll have to scroll quite far down to see Bishop Gorman 2016 for example and to find a FL team you may need to reach 80th or more. 1 Quote Link to comment Share on other sites More sharing options...
Atticus Finch Posted December 21, 2021 Report Share Posted December 21, 2021 Just now, dntn31 said: These are all things he publishes on his site, it's not really new information. These are all things that are set in advance of the games being played for the current year. Is the insinuation that Ned knows that these yearly "tinkerings" will give CA teams an edge? No, those are just the known human biases that Ned has included in his ratings. The California teams benefit from structural advantages that are partly outside of calpreps and also party due to those biases. Quote Link to comment Share on other sites More sharing options...
Atticus Finch Posted December 21, 2021 Report Share Posted December 21, 2021 4 minutes ago, dntn31 said: I mean if he wanted to put his foot on the scales to give CA an edge, he could just do that without actually publishing anything on his site. Or do you think he's trying to be extra sneaky by thinking that people will deem his ratings more trustworthy (despite him actually putting his foot on the scaled behind the scenes) if he only gives the appearance of being more open about how his ratings are produced? To be clear, I don't think anybody is claiming that Ned Freeman is purposefully inflating ratings for certain teams. I know that's not what I've been saying. Quote Link to comment Share on other sites More sharing options...
golfaddict1 Posted December 21, 2021 Report Share Posted December 21, 2021 15 minutes ago, OldTerrapin said: I like both those shows.. Been watching Partridge Family reruns on Tubi from time to time.. Liked Benny Hill too but mostly for the boobies, not sure I got a lot of the humor back then. It's been a long time since I have seen a Benny Hill episode.. Did find some Man Show ones though..Man Show was probably a more modern Benny Hill show..lol The little old bald man on the Benny Hill show was priceless. He always got the quick smacks on the top of his head lol. Yea the Man Show was pretty funny I thought. I watch a lot of old programs. On Hulu, in the past year I've watched every episode of Taxi and the Odd Couple. Other favs I watch on regular TV often are Hogans Heroes, Twilight Zone and yes even Alice... so kiss my grits Quote Link to comment Share on other sites More sharing options...
dntn31 Posted December 21, 2021 Report Share Posted December 21, 2021 1 minute ago, Atticus Finch said: To be clear, I don't think anybody is claiming that Ned Freeman is purposefully inflating ratings for certain teams. I know that's now what I've been saying. It seems like this is just semantics then. Generally the word "bias" carries a connotation that the bias directly influences the output of some process. In the case of the computer algorithms, I think it's difficult to argue that the bias does more than indirectly influence the output - barring precognition. When you say "the computers are biased" it could be easily inferred that Freeman/Massey are purposely "inflating ratings for certain teams" or the converse. Quote Link to comment Share on other sites More sharing options...
badrouter Posted December 21, 2021 Report Share Posted December 21, 2021 1 hour ago, Atticus Finch said: In other words, shitty teams in California have higher ratings than shitty teams elsewhere. Or even above average teams in California (Los Alamitos) have higher ratings than good teams elsewhere (Cardinal Gibbons). Exactly. People fixate on the top 10-20, but there are major head-scratchers as you go through the top few hundred. And all of those ratings help with the self-fulfilling prophecies, for reasons you've explained, which remain unrefuted. Quote Link to comment Share on other sites More sharing options...
golfaddict1 Posted December 21, 2021 Report Share Posted December 21, 2021 30 minutes ago, dntn31 said: I can't remember where I read it but I remember reading two things: 1) He uses the past 2 years of data to seed the ratings for the current year 2) Eventually the starting ratings are filtered out as more data becomes available for the current year (same claim as CalPreps) 1) Yes, this is what he uses to attempt to gain more OOS tilts to make a better judgement call for state scale placement strength/limitations. Imo, it's futile to do this as it still is a tiny sampling and it also affects two different seasons into one, with different rosters. The interaction is so tiny with OOS, yet so vital to create what we would deem a quality looking list. 2) Eventually... if they reach what is it 3 minimum games to receive an actual rating for the season and I'm assuming Freeman's projection was for teams to have more games for the algorithm to "accurately" place correctly in the ratings based on the actual number of games required to rid the starting rating out completely (like the owl licking the tootsie roll pop, you just don't know how many ). I assume Massey has a similar model in place. Quote Link to comment Share on other sites More sharing options...
OldTerrapin Posted December 21, 2021 Report Share Posted December 21, 2021 21 minutes ago, golfaddict1 said: The little old bald man on the Benny Hill show was priceless. He always got the quick smacks on the top of his head lol. Yea the Man Show was pretty funny I thought. I watch a lot of old programs. On Hulu, in the past year I've watched every episode of Taxi and the Odd Couple. Other favs I watch on regular TV often are Hogans Heroes, Twilight Zone and yes even Alice... so kiss my grits A week ago I finished watching all the NYPD Blues.. Before that I watched all the Buck Rogers in the 25th Century.. very underrated and Erin Gray is a smoke show lol I watched Alice as a kid lol.. I love all the older shit like Petticoat Junction, Green Acres.. 1 Quote Link to comment Share on other sites More sharing options...
golfaddict1 Posted December 21, 2021 Report Share Posted December 21, 2021 2 minutes ago, OldTerrapin said: A week ago I finished watching all the NYPD Blues.. Before that I watched all the Buck Rogers in the 25th Century.. very underrated and Erin Gray is a smoke show lol Big fan of that show... I also liked the comedy side police work of Barney Miller. It may not be your thing, but give Brooklyn 99 a try for a police comedy... imo, great cast and witty writing. 1 Quote Link to comment Share on other sites More sharing options...
badrouter Posted December 21, 2021 Report Share Posted December 21, 2021 38 minutes ago, dntn31 said: Without putting words in GoBigBlack's mouth, I think the point he's driving at here is that human pollsters have "reactionary bias" in that they 1) have bias (impossible not to) and 2) those biases infiltrate whatever logic they end up utilizing to come up with a ranking after the games are played. Whereas computer algorithm's don't necessarily do that. I don't think "bias" the correct word to use when talking about computer algorithms. Yes, Freeman and Massey both have biases (as all humans do) and they ultimately have to make some decisions about the heuristics they end up using within their algorithms, but at the end of the day these decisions are made well in advance of the games that are played. The question that needs asking is: "are the sum total of these decisions in any way biased towards a specific subset of teams?". It's easy enough to be dismissive of CalPreps (as has been done to the Nth degree on this forum over the past decade), but very few ever seem to address Massey. I do agree that both Massey and Calpreps seem to undervalue Florida (compared to CA and TX), but I haven't really seen anyone provide a convincing argument as to why. If anyone could provide a well-reasoned argument without invoking the words "CalPreps", "bias" or simply trolling I think it would go a long way in figuring out a path forward instead of sitting here constantly chasing our tails. Yes. Perhaps not in a league like the NFL, where common opponents are abundant, and the rules/dynamics around team building are nearly identical for all teams. Perhaps not even within individual leagues at lower levels of football. But, on a national level in high school, clearly some subsets of teams have a significant advantage built-in. This may not even be a product of human bias on the part of the programmer, but just the nature of the beast at a level with such wide-ranging opponents. 1 Quote Link to comment Share on other sites More sharing options...
badrouter Posted December 21, 2021 Report Share Posted December 21, 2021 Just now, badrouter said: The question that needs asking is: "are the sum total of these decisions in any way biased towards a specific subset of teams?". Yes. Perhaps not in a league like the NFL, where common opponents are abundant, and the rules/dynamics around team building are nearly identical for all teams. Perhaps not even within individual leagues at lower levels of football. But, on a national level in high school, clearly some subsets of teams have a significant advantage built-in. This may not even be a product of human bias on the part of the programmer, but just the nature of the beast at a level with such wide-ranging opponents. Quote Link to comment Share on other sites More sharing options...
OldTerrapin Posted December 21, 2021 Report Share Posted December 21, 2021 Just now, golfaddict1 said: Big fan of that show... I also liked the comedy side police work of Barney Miller. It may not be your thing, but give Brooklyn 99 a try for a police comedy... imo, great cast and witty writing. I watched Barney Miller too, great show.. A show I never watched when it was out that I may binge on is Hill Street Blues..but right now I am all caught up in Yellowstone. Yellowstone is ligit 1 Quote Link to comment Share on other sites More sharing options...
badrouter Posted December 21, 2021 Report Share Posted December 21, 2021 43 minutes ago, dntn31 said: If anyone could provide a well-reasoned argument without invoking the words "CalPreps", "bias" or simply trolling I think it would go a long way in figuring out a path forward instead of sitting here constantly chasing our tails. My hunch is the relative lack of connectivity among the top teams in Florida compared to what's seen in CA, Ohio and, at least by week 16, Texas. The more the best teams are concentrated in Dade and Broward, the less this may be an issue. But, even as it is now, STA didn't face Central or Chaminade. So, there's no linkage there between STA and SJB (or Mater Dei and Servite to a lesser extent). Quote Link to comment Share on other sites More sharing options...
dntn31 Posted December 21, 2021 Report Share Posted December 21, 2021 39 minutes ago, Atticus Finch said: He thinks that playoff games are a better indicator of strength than regular season games so he gives more weight to them to reach that outcome. Not sure what Massey uses in his current algorithm, but this is what his thesis has to say on this: Quote Link to comment Share on other sites More sharing options...
golfaddict1 Posted December 21, 2021 Report Share Posted December 21, 2021 4 minutes ago, OldTerrapin said: I watched Barney Miller too, great show.. A show I never watched when it was out that I may binge on is Hill Street Blues..but right now I am all caught up in Yellowstone. Yellowstone is ligit Never heard of it... thanks for the tip. I'll check it out. Quote Link to comment Share on other sites More sharing options...
badrouter Posted December 21, 2021 Report Share Posted December 21, 2021 40 minutes ago, dntn31 said: I can't remember where I read it but I remember reading two things: 1) He uses the past 2 years of data to seed the ratings for the current year 2) Eventually the starting ratings are filtered out as more data becomes available for the current year (same claim as CalPreps) It seems to me that, in practice, this actually ends up happening less than stated (or intended.) IMO, Venice's higher rating than Jesuit and STA has something to do with all of its games with both STA and IMG- both of which have "connectivity with the preferred teams in the Trinity league- over the years. Quote Link to comment Share on other sites More sharing options...
Atticus Finch Posted December 21, 2021 Report Share Posted December 21, 2021 30 minutes ago, dntn31 said: Generally the word "bias" carries a connotation that the bias directly influences the output of some process. I'd call that corruption. Quote Link to comment Share on other sites More sharing options...
dntn31 Posted December 21, 2021 Report Share Posted December 21, 2021 40 minutes ago, Atticus Finch said: He doesn't think that 1-point wins are almost equal to 1-point losses so he tinkers with the math to reach that outcome. Massey: Quote Game Outcome Measure Because of the intricacies of sports, including motivation and coaching philosophy, it could be argued that a diminishing returns principle should be implemented in any rating model. This strategy attempts to avoid the situation in which a team is rewarded for blowing out weak opponents, an unfortunate characteristic of the general least squares method. This occurs because least squares ratings are actually averages, and the mathematical properties of means demand sensitivity to outlying or extreme values. Since margin of victory is the only statistic upon which the ratings are based, a team can improve its rating by running up the score. In reality this is usually done to impress fans, the media, or, in the case of college sports, the voters for the top 25. Conversely, it is possible for a team’s rating to decline unfairly after a solid victory over a pathetic team because it didn’t win by enough points to compensate for the opponent’s weakness. This is sometimes caused by the additive nature of least squares ratings. If A is predicted to beat B by 30 points and B is likewise 30 points better than C, then we get an unrealistic expectation that A would defeat C by 60 points. Although it results in a loss of mathematical significance, more reasonable results may be obtained by applying a diminishing return function to the least squares rating method. Such a function stipulates that as the margin of victory increases, the benefit to the winner increases at a slower rate. The function result replaces margin of victory as the dependent variable y of the regression. An example would be the signed square root of the margin of victory. In this case the difference between an 81 point win and a 36 point win is equivalent to the difference between a 4 point win and a 1 point loss. The ultimate case of diminishing returns completely throws out any information about the score of a game. Each win is treated as if it were by one point, so y = 1 no matter what. Hence, there is no distinction between an 80 point massacre and a 1 point nail-biter. Teams that subscribe to the “Just Win Baby” philosophy are rewarded because winning, especially against opponents who win, yields a higher rating (Zenor 1995). The term game outcome measure (GOM) has been coined to describe any function, including those that exhibit diminishing returns, used to calculate the dependent variable y in a least squares rating model (Leake 1976). It is based on the idea that it is impossible to accurately establish how superior one team is to another from margin of victory alone. In theory, any concoction of game statistics can be taken as variables from which to formulate the GOM, but usually only the score is necessary. The GOM becomes a single numerical representation of a team’s performance in a game relative to its opponent. The general least squares method is modified by replacing margin of victory with the GOM result. Usually the implementation of a function to discount blowout scores produces a set of ratings that possesses a certain degree of “fairness” absent from the general ratings. Although an inverse function can be derived to translate the new ratings back into a predicted outcome, the mathematical legitimacy of these expected results is sacrificed. Despite this fact, these versions of least squares regression may model reality more effectively than the original because they incorporate a knowledge that the nature of sports is not entirely consistent with ideals assumed by regression methodology. Quote Link to comment Share on other sites More sharing options...
Atticus Finch Posted December 21, 2021 Report Share Posted December 21, 2021 31 minutes ago, dntn31 said: It seems like this is just semantics then The overvaluing of teams is certainly not semantics. Quote Link to comment Share on other sites More sharing options...
Atticus Finch Posted December 21, 2021 Report Share Posted December 21, 2021 7 minutes ago, badrouter said: It seems to me that, in practice, this actually ends up happening less than stated (or intended.) If calpreps filters them out then I'd say that the scaling renders it meaningless. Quote Link to comment Share on other sites More sharing options...
golfaddict1 Posted December 21, 2021 Report Share Posted December 21, 2021 3 minutes ago, dntn31 said: Not sure what Massey uses in his current algorithm, but this is what his thesis has to say on this: With so little connect the dots, imho every game is meaningful and will and should be scrutinized. The better opponents and performances will take care of itself, but an opening game between two OOS powers matters for national ratings and impacts much of the ratings between states. Much of the ratings system is based on scaling states. So when he mentions playoff emphasis, I read state emphasis. Granted, he didn't mention early OOS tilts and I'm sure that weight plays a main factor in the rating system... but it should not be downplayed based on date. Yes, some state playoffs are not the same and should be magnified. But if STA vs SFA is downgraded vs. STA vs playoff opponents not in the same tier or tiers apart... that's wrong. SFA didn't travel to lose nor do they even participate in the playoffs. Games like that should be accentuated without an expiration date weakness nor a missing playoff boost. It is also a frustrating period of time in November, waiting for each playoff power state to take turns having multiple teams in the top 25 before the next power state reaches a higher round... non playoff OOS tilts between national powers shouldn't require an expiration date in the same season. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.