BCMoore Rankings: Final

[ghost80]

This thread should be pulled... period.

It grew dingleberries when Superbee and CP 84 starting comparing Moore's toy to actual gambling applications and posted the Week 12 NFL betting line on a high school sports site. Nice work fellas.

When the plug didn't get pulled I pushed the thread a little further south figuring that would do the trick. It didn't.

If people look at Moore's work and see gambling applications maybe they are all being posted on the wrong boards.

 

[Cardinal2012]

The main purpose of their reference to gambling was to provide the doubter, you, with a real life example of how statistical analysis can and is being used. Perhaps there was another alternative that could have sufficed to portray the same message, but nothing would have been widely known as gambling. It was about making it as easy to understand as possible.

Perhaps you do understand statistical analysis. I personally can not tell if you do or do not. If you did, you would understand that each game and only be predicted with a certain sense of accuracy, the wider the range obviously the more confident the computer will be. Some games that are predicted to be close may fall within the given range of the computer, so as even if the computer picks the wrong winner it is possible that both teams performed within the predicted values given to them. Maybe this is where your biggest concern lies, please tell me if its not, but you believe that everything is absolute. If a computer says Team A over Team B by 3 and the actual outcome is Team B over Team A by 3 you view it as the computer being 100% wrong. It is in the sense that it predicted the wrong winner, but it also may have predicted the value of each team to the 98% of it's actual result. This seems to be the biggest rift between the "believers and the nonbelievers". You view it as wrong, while others are impressed that a machine with no bias what-so-ever can predict the outcome of a game it has never seen, nor even knows the rules to, to within a few points. If you don't understand statistical analysis, then perhaps this is why you don't appreciate the applications of BCMoore as dearly as some. I'll admit there is a lot of speculation here, feel free to politely correct any false assumptions I have made.

 

[ghost80]

Cardinal,

A simple question.

Is sports gambling or even mention of sport gambling appropriate on a HIGH SCHOOL message board? You said the whole reason why statistical analysis' relationship to gambling was being brought up was to provide a real life example for doubters like me? Guess what..if it's for my benefit then I have some say if it is appropriate or not and posting an NFL betting line isn't appropriate PERIOD.

As far as your comparison to considering a 98% completion of an actual outcome still being "nearly" right and not 100% wrong is utterly foolish when applied to sports outcomes (which is afterall what we are talking about). In real life (since you want to go there) sporting outcomes of "98% of a victory" means nothing. Has their ever been a champion crowned or even a game considered a win for a team because they led the first 98% of the match and fell behind and lost during the last 2% of the contest? NO. Has their ever been a champion crowned or even a game considered a win because a team score 98 points while their opponent scored 100? Of course not. What would your bookie say if you thought you should collect 98% of the bet since your team accomplished "98% of a win"? Your 98% concept is 100% laughable.

Those 2 instances might be considered a moral victory as is a common theme while addressing losers....but noone in the real world will ever see it as anything but a loss. At the end of the day you win or you lose.

Maybe Moore should start predictions on horseshoes and hand grenades....close does count for something there.

 

[Cardinal2012]

I'm not the person to decide what is appropriate and what is not. I'd be willing to wager that most of the people on this HIGH SCHOOL forum, are out of high school and are responsible enough to not be offended by gambling.

A sticker on an object says it is 12 inches, your tape measure reads 11.8 inches. This is 98%. Which do you assume to be right? Do you say close enough? Maybe you want to get a more precise measuring tool, this can be viewed as introducing more variables into a computer equation which is why some computer models are different than others. I'm a person that views the model like this tape measure, for the applications I am using it for it is certainly close enough.

 

[ghost80]

Computers are capable of precision that humans are not capable of doing YET you want to give the computer a "pass" when it's close to being right but the humans who didn't get it right don't get one?

I'm glad Cardinal didn't build my house. If he is ready to "call good" any measurement that is within 98% he is going to build a very poorly constructed home. I hope he's also not a pharmacist and filling my prescriptions. "Calling good" at 98% will kill people.



This post was edited on 11/29 1:47 PM by ghost80

 

[L N D]

Ghost - You do realize this is high school football! Not life or death.
This post was edited on 11/29 2:14 PM by L N D

 

[ghost80]

LND, I'd like to have you try an experiment.

Walk up to your high school football coach (or if you are an adult have your kid do it) and in front of a public group of people announce to him , " This is just high school football so I'm only going to try 98%...that is good enough".

Let us know how well that works for you or your kid. I'm guessing if you don't quit the team you will be wishing you were dead by the time the coach gets done with you.

 

[Cardinal2012]

Originally posted by ghost80:
Computers are capable of precision that humans are not capable of doing YET you want to give the computer a "pass" when it's close to being right but the humans who didn't get it right don't get one?

I'm glad Cardinal didn't build my house. If he is ready to "call good" any measurement that is within 98% he is going to build a very poorly constructed home. I hope he's also not a pharmacist and filling my prescriptions. "Calling good" at 98% will kill people.



This post was edited on 11/29 1:47 PM by ghost80

There is not a single thing on this Earth that is built with no tolerance. Some applications 98% is well within reason, others you would like more precision. I'd guarantee your house is not perfectly level, especially in Iowa where the temperatures cause the ground and material to change size as the seasons progress. I also feel extremely confident in assuming that any drugs that have a significantly high chance of killing you are subject to stricter tolerances as well. Not everything in this would has to be plum perfect. I'm sure you've driven around with one tire lower than the other, not enough to notice, but if you took the time to check you would see. Notice how you are still alive to this day when your tires were not perfectly filled exactly to 100%. I tried to politely provide you with examples and asked you to correct me where I made false assumptions, you responded with personal attacks that outlined my explanation of the given ranges and not the actual concepts. If you believe that the computer projections of a high school football game are on the same magnitude as life or death situations, then please just say so and I will quit trying to convince you otherwise.

 

[DarkThunder#61]

ghost is going to say whatever he can to push his point about BCMoore's rankings. What both parties have to understand though (both being ghost and the people who read/reply to ghost) is that: he will not change his opinion and nothing we say will change it either. This has been proven because there is nothing left to say. BCMoore's rankings have been perfectly explained multiple times before and many people understand what they are. ghost knows this. ghost doesn't care.

The other thing is that ghost has to understand that nobody cares how much of a problem he has with BCMoore's rankings, and that he, himself, will not influence anyone (or at least he has done a poor job of convincing people) to take BCMoore's rankings for anything less than what they are.

It is wholly unnecessary to spend time discrediting them, ghost...but what we must accept is that it is ghost's right to waste time on here doing that any ways.

So carry on.....

 

[CP84]

I don't see why Ghost would complain about mere mention of game spreads yet alone show faux disgust with it. Gambling was never advocated and the spreads were used in context of demonstrating prediction modelling alongside other examples (i.e. BCS, Sagarin). Seems like a very phony complaint. As DT said, Ghost doesn't apparently see the value in prediction modelling so this isn't productive.

This post was edited on 11/29 8:46 PM by CP84

 

[ghost80]

You guys make me laugh.

Gonna fill you in on a little secret. I probably know more about statistical analysis and predictive modeling (there is only one L in the word modeling CP84) than 95% of the people who come on Moore's threads and congratulate him for his efforts.

Cardinal. You seem impressed if Moore can get you 98% of the way there and seem excited with just a 2% margin of error. I spent over a decade of my life dealing with thresholds in which your 2% margin of error would be considered as 200.000 basis points. More times than not my thresholds were right of the decimal. A product left of the decimal was considered slop and needed to be fully reviewed whether to consider it acceptable or a failure.

Alex, you are right in that you guys won't make me change my mind on Moore's toy. Something tells me though that you guys don't fight to try and change my mind. You are fighting to try and influence others back to believing that there is something more than minimal value to these rankings. I am going to disagree with you that people don't care about the errors and limitation of BCs system that I continually expose. Even if people don't comment they are forming opinions and based on a higher percentage of people questioning the flawed results than people jumping to defend them; it tells me that others are reading and thinking. There seems to be a shift to people saying, "They appreciate Brent's efforts or work" (and I do believe they appreciate his work) but fewer are any longer jumping up with enthusiasm and blindly accepting his results. There is a huge difference between to two. People saying, "Brent I appreciate your work" is the equivalency of giving Brent a courteous participation acknowledgement. People might think I'm crass, an a$$, and not like me but it is clear people are being influenced that I am largely right and the ranking are a cheap toy. Even within this thread Cardinal was willing to negotiate away the accuracy of the computer. While I realize that for most of us (including myself) getting a 98% on a test would normally be considered great, for a computer application 98% accuracy is terrible. In the real world very few applications would survive if that was the best they could do.

Sorry fellas...I'd love to chat more but it I'm not slaved to a computer and it is way too beautiful of a day to spend any more of it in here.

 

[CP84]

Comparing failure rates between sports prediction analysis to finance makes about as much sense as comparing evidence based medicine outcomes to meteorology yet both provide value.
This post was edited on 11/30 8:30 PM by CP84

 

[CP84]

Your "exposing of BCMoore" was baseless. You cited 5 games out of ~27 in which 4 of those games were decided by 1 score. If Moore's system was exposed I'd like to hear your critique of the March Madness seeding committee which rarely has four #1's in the final 4 yet alone getting 98% (appears to be your minimum acceptable standard) of the remaining bracket correct. Or would you finally admit that toss up games and upsets can influence tools. Both BCS computer models and AP favored Alabama over Auburn tonight so yes sports prediction analysis has wider variance ranges than sectors like manufacturing but you don't throw the baby out with the bath water. If Auburn played Alabama 5 times would BCS and Sagarin get it wrong the majority of the time? Doubtful, so it should be clear to all that a few toss up games don't determine a tool's value. You must compare to comparable alternatives.



This post was edited on 11/30 8:37 PM by CP84

 

[Cardinal2012]

His response will be that it doesn't require much skill to get the easy picks right, the real challenge comes from the toss-up games only.

 

[CP84]

The tool is designed for predicting score variances. Ghost cited 5 games BCMoore predicted the wrong team winning to dispute his tool. BCMoore projected all 5 games as toss up games (I.e. within 1 score). Only 1 of the 5 was decided by >8 points with one going to OT and one being a reversal in outcome from a regular season game (Harlan vs. Carroll).

My point is if BCMoore predicts Team A favored by 1.5 , essentially a toss up, and Team B ends up winning by a FG, is that proving his system wrong?

This post was edited on 11/30 11:11 PM by CP84

 

[ghost80]

Now I'm really laughing at you CP!

You made an ASSumption and sent yourself barking up the wrong tree? Finance? I never made a comparison between high school sports prediction and finance but I know why you made that ASSumption.

Just because you claim to have attended the London School of Economics you think that the financial world is the only ones who works in bips? Even though a bip has it's origins in the financial world (and I'm sure is still regularly used there since my financial advisor and I use it often in our conversations) as the rest of the world has evolved and required more precision they too have commonly reduced their unit of measurement for percentages from whole percentage points to basis points (bips). Matter of fact we don't even have to get authorisation (I threw that one in for you old chap) from the financial world to use it! A bip is 1% of 1%... nothing more or less and in today's world...anything dealing in whole percentages is a sign of imprecision.

Maybe we have stumbled onto the problem. The world has become too precise for an antiquated sports prediction model. That could explain why it failed to finish in the top half of the 20 participants in the 3A "pickum" contest as well as being bested in the 1A.

Cardinal..haven't forgotten about you. How come the computer was inferior and couldn't guess the "toss ups" as correctly as the humans did in the 3A pickum contest mentioned above? At the end of the day you win or lose. It doesn't matter if they are easy picks or hard picks...ALL the contestants, human and computer alike, had the same games to pick so it really doesn't matter and is silly to try to differentiate anything special about the close picks or hards ones.



This post was edited on 12/1 9:34 AM by ghost80

 

[ghost80]

Sorry chaps but I'm on holiday and ready to head out to bag a few critters. If you'd like to discuss BC's shortcomings I'll be more than happy but I am warning your that this will long holiday so it could be a spell before we next chin wag this tochy subject.

Cheers!

Almost forgot CP. When I look at your logic I have to ask myself. When you were in London, just how many weekends did you spend in Amsterdam?

 

[CP84]

Once again, comparing sports prediction models to other sectors with less variability is useless.

But again having higher variability in sports predictions hardly makes it irrelevant. The Computer Group made millions through sports betting using similar model formats (hope you're not aghast). Baseball analysts like Bill James and Jeff Sagarin revolutionized baseball and now most teams have adopted Sabermetrics while Sagarin's model has been adopted by the BCS. Dean Oliver and KenPom have done the same for basketball.

Hold the surprise, none of these have anywhere close to a 90% accuracy mark and quite frankly they shouldn't be expected to.

Maybe you should take a moment to read this article.

http://www.advancednflstats.com/2011/10/prediction-accuracy.html?m=1

This post was edited on 12/2 12:46 AM by CP84

 

[CP84]

"How come the computer was inferior to the humans."

Once again you fail on the issue of sample size. We are speaking of 31 games in which a small handful could be considered toss ups. The difference between middle of the pack and 1st is three games. Hardly significant to make any judgment. Of the 5 BC missed, only 1 was >1 score. Again BC was predicting several of those within 3-4 points to barely over a point. In other words he's stating the "losing prediction" to win > 40% of the time. So again, how does Moore predicting Team A favored by 1.5 and ends up losing by 1 score or less to Team B debunk his tool? To anyone applying logic it doesn't.

As the article I posted infers, you need to look at the long-term average of a prediction model to consider its value. If you look throughout the past several years BC is in the top half the majority of the time. A tool that can consistently predict ~80% right as BCMoore did in the example of the 3A pick'em, it would be at or above the average of other professional services like Sagarin in predicting NCAA football outcomes. With humans you are going to see larger variances. Occasionally they may pick more accurately but other times their predictions will be well off. The purpose of tools like this is to decrease the variance range. If a tool can consistently predict ~80% correct it's a very viable tool.
This post was edited on 12/9 10:02 AM by CP84

 

[bcmoore87]

NCAA football prediction data:

http://www.thepredictiontracker.com/ncaaresults.php

Best computer model is correct 79.4% of the time.
Vegas line is correct 78.2%-78.8% of the time.


Prediction is hard -- and thus, games are typically quite exciting.

 

[ghost80]

Wow!! I come back from my holiday looking for a good laugh before it's back to work and CP 84 comes through and doesn't let me down.

What could make me laugh about CP84? Well a lot of his rambling are so far off kilter and contradict themselves I've come to the conclusion that he has to be doing this for laughs. Here are some examples.

1. CP84 bashes the NCAA selection committee for rarely having the top seeds in the tournament right yet several posts later piles on compliments to other predictive modelers such as Sagarin. Here is the laugh.....the 2014 NCAA tournament will be the 30th year that the selection committee have used Jeff Sagarin as a consultant and his rankings as a guide.

2. Blaming Moore's failure to be competitive with the humans in the 3A playoff pickum contest on too small of a statistical sampling. Isn't the computer supposed to be more accurate with each week as the volume of data grows? Isn't picking the playoffs correctly what everyone is talking about when year after year as the computer stumbles the common excuse is, "wait, by the end of the season it will all work out". Know why the majority of the humans bested the computer in the 3A playoffs? Because the number of games being played became manageable for people to look at and perform meaningful analysis and that analysis topped a simple meaningless compilation of W/L and scores.

3. Comparing Moore's system alongside of a system like Sabermetrics. Moore's system does nothing but compile the team wins and losses and scores for 9 football games. (I left off the playoffs since CP 84 seems to have a fit if I include them and says they are statistically insignificant). What does Sabermetrics do? Briefly, Sabermetrics analyzes individual level statistics at the player level so teams can make decisions at the player level. At the team level they can then roll up players to play "what ifs" and see how a team of rolled up players matches against other teams. This is highly useful in accumulating the right mix of players to be successful at the team level.

CP 84, want to talk about scope? Let's talk scope. Sabermetrics is able to work and be effective because of the volume of individual level data available. The average MLB players plays 5.6 years. At 162 games per season that equates to over 900 games. (For reference Pete Rose holds the record for 3,562 games and over 14,000 at bats) Think of how many at bats the average player has during that time. To further expand the data available Sabermetrics even tracks minor league stats and assigns a factor for AAA, AA, ect. Once again....Moore merely tracks team W/L and scores. To put this into perspective..if you have a high school football player who played varsity for all 4 years and made it to the state championship every year, how many games would he have played in?? While my example will likely never happen the answer is only 64, Even if Moore were to somehow develope a system for individual stats and roll them up to team level like Sabermetrics (and Ken Pomeroy is doing for basketball which has 82 regular games per season) he wouldn't have nearly enough data available to make it meaningful (even if he were to somehow roll in SASF's middleschool stats).

CP...even mentioning Sabermetrics on a Moore thread is a side splitter!!!

3. BCS?? Why even mention them on a thread about Moore?? Are their systems alike??? One plays a regular season 1/3 longer than the other (which means a much greater volume of data to compile). I will mention this though which I have always found funny. Moore starts publishing his rankings on week 0 with nothing but the prior year rankings. The BCS delays their first published ranking until mid October after most teams have played 7 games. They realize that the raw compilation of W/L and scores needs some time to accumulate and gain credibility.


CP, you do realize this is about Moore's rankings? As I have pointed out above.....very irrelevant to mix watermellon and grapes and pull in some sporting predictive tools that do have some value.






This post was edited on 12/7 8:11 AM by ghost80

 

[stickman80]

I think Ghost's mind must still be on vacation. The maximum number of varsity high school games a kid could play would be 56 instead of 64. A more reasonable number would be even less. Limited play as a sophmore, 2 years starting and a playoff game or two a year might get you up to 32.

 

[ghost80]

Nah.....I'm liking 64 games as the max. Thought about 65 but decided those stats wouldn't matter.

Get back to work Stickman.

 

[DarkThunder#61]

You're probably doing this for a reason, but I'm pretty certain the most games a team can play in a year are 15.

10 regular season games if you play week zero, too. (unless there's some way to play 11, please share...)
5 rounds of playoffs (1st, 2nd, qtrs, semies, finals)

Multiply by 4 =..........?

 

[SparkItUp]

I like your rankings but not sure how a 10-2 team can be ranked behind a bunch of 5-5 teams, makes no sense sorry.

 

[ghost80]

64 games?

Let's see. 9 regular season games, 5 playoff games, PLUS an intersquad scrimmage AND a non intersquad scrimmage which is now allowed. That equates to 16 games per year or 64 over 4 years. Since this is at the individual level the Shrine Bowl could make number 65 but since it is played in the summer after graduation, the stats really don't matter from a predictive tool aspect.

Thunder, you are correct that there are a few 10 game schedules mostly to accomidate odd numbers of teams. If a kid were on one of these teams he could actually have a higher total.

Am I stretching it including an intersquad scrimmage and an "exhibition" game with another team? I don't think so. Both are competitive events and we have seen posters treating the exhibition games like thy were real. Moore could weight the stats from these "games" similar to Sabermetrics treating minor league baseball. The stats count but with a lower weightings. These "scrimmage games" are in fact more real than current early year ratings based on players who are no longer on a team.

 

[cidhawkeye]

Might as well throw in the summer time 7 on 7 games into the mix. Also you can simulate games on the Massey web site so if you do that for.every game that doubles the number of games that could be factored in. You might be on to something here....but then again maybe not.

 

[ghost80]

Cid,

I don't think throwing stats from 7 on 7 into the database is out of the realm. Those stats are being compiled by players who will be strapping it on that fall and playing high school football. As I've said before....the present starting point for the current ranking system is to take the prior year rankings which includes many players whose high school career is over. 7 on 7 stats are far more relevant than what is currently being done.

Massey? IDK. Keep in mind that what we are trying to accumulate is individual level stats...not another meaningless compilation of W/L and points. Does Massey's simulated games also project individual stats?

I do like your thinking outside of the box Cid!

 

[CP84]

Originally posted by ghost80:
Wow!! I come back from my holiday looking for a good laugh before it's back to work and CP 84 comes through and doesn't let me down.

What could make me laugh about CP84? Well a lot of his rambling are so far off kilter and contradict themselves I've come to the conclusion that he has to be doing this for laughs. Here are some examples.

1. CP84 bashes the NCAA selection committee for rarely having the top seeds in the tournament right yet several posts later piles on compliments to other predictive modelers such as Sagarin. Here is the laugh.....the 2014 NCAA tournament will be the 30th year that the selection committee have used Jeff Sagarin as a consultant and his rankings as a guide.

2. Blaming Moore's failure to be competitive with the humans in the 3A playoff pickum contest on too small of a statistical sampling. Isn't the computer supposed to be more accurate with each week as the volume of data grows? Isn't picking the playoffs correctly what everyone is talking about when year after year as the computer stumbles the common excuse is, "wait, by the end of the season it will all work out". Know why the majority of the humans bested the computer in the 3A playoffs? Because the number of games being played became manageable for people to look at and perform meaningful analysis and that analysis topped a simple meaningless compilation of W/L and scores.

3. Comparing Moore's system alongside of a system like Sabermetrics. Moore's system does nothing but compile the team wins and losses and scores for 9 football games. (I left off the playoffs since CP 84 seems to have a fit if I include them and says they are statistically insignificant). What does Sabermetrics do? Briefly, Sabermetrics analyzes individual level statistics at the player level so teams can make decisions at the player level. At the team level they can then roll up players to play "what ifs" and see how a team of rolled up players matches against other teams. This is highly useful in accumulating the right mix of players to be successful at the team level.

CP 84, want to talk about scope? Let's talk scope. Sabermetrics is able to work and be effective because of the volume of individual level data available. The average MLB players plays 5.6 years. At 162 games per season that equates to over 900 games. (For reference Pete Rose holds the record for 3,562 games and over 14,000 at bats) Think of how many at bats the average player has during that time. To further expand the data available Sabermetrics even tracks minor league stats and assigns a factor for AAA, AA, ect. Once again....Moore merely tracks team W/L and scores. To put this into perspective..if you have a high school football player who played varsity for all 4 years and made it to the state championship every year, how many games would he have played in?? While my example will likely never happen the answer is only 64, Even if Moore were to somehow develope a system for individual stats and roll them up to team level like Sabermetrics (and Ken Pomeroy is doing for basketball which has 82 regular games per season) he wouldn't have nearly enough data available to make it meaningful (even if he were to somehow roll in SASF's middleschool stats).

CP...even mentioning Sabermetrics on a Moore thread is a side splitter!!!

3. BCS?? Why even mention them on a thread about Moore?? Are their systems alike??? One plays a regular season 1/3 longer than the other (which means a much greater volume of data to compile). I will mention this though which I have always found funny. Moore starts publishing his rankings on week 0 with nothing but the prior year rankings. The BCS delays their first published ranking until mid October after most teams have played 7 games. They realize that the raw compilation of W/L and scores needs some time to accumulate and gain credibility.


CP, you do realize this is about Moore's rankings? As I have pointed out above.....very irrelevant to mix watermellon and grapes and pull in some sporting predictive tools that do have some value.







This post was edited on 12/7 8:11 AM by ghost80

1. Wow, I'm not bashing the selection committee or arguing that they shouldn't maintain their process. The example was given to illustrate how frequently upsets happen when multiple games are predicted within a few points. The point was to show how this is apples and oranges compared to prediction in other sectors. That's just a failure in reading comprehension.

2. "Isn't the computer supposed to be more accurate with each week as the volume of data grows?" As stated multiple times, only 1 out of the 5 BCMoore "misses" was greater than 1 score and he predicted all of those games to be tossups. BCMoore's data isn't factoring in subjective information (i.e. home field advantage, injuries, etc..), it's looking at historical data and using those results to predict future outcomes. For example he predicted Carroll a very slight favorite over Harlan (0.41 points). Carroll won their regular season game 14-7. Harlan squeaked out a 21-18 win in the playoffs. So in other words you are trying to attack a model's credibility based on a sample size of three in which Moore predicted the winners as less than 60% favorites to win. I suppose if you flip a coin twice and it happens to land tails twice the 50/50 odds are incorrect too?

3. Hey I'm not arguing that sample size isn't valid. Of course statistical models with more data should have a higher degree of accuracy but you could say that based on sample size high school football is inferior to college, college to the NFL, the NFL to the NBA and the NBA to MLB. That certainly doesn't mean prediction modeling is only valid for MLB and shouldn't be used in football due to fewer games. At least you finally appear to acknowledge that increasing sample sizes of data is typically better than subjective arguments.


The other #3) The systems use a similar methodology. So you find it acceptable to start using BCS rankings following week 7 but we can't trust BCMoore even after 9 games and into the playoffs? Makes about as much sense as everything else you've said. Who cares if BCMoore starts using rankings in week one. Most computer models do as well but Moore or Sagarin would both be the first to tell you it doesn't mean much until week a handful of games have been entered.




This post was edited on 12/9 9:58 AM by CP84

 

[CP84]

3A Pick'em results:

BCMoore: 26 of 31 (83.87%)
Total Humans: 650 of 776 (83.76%)


1A Pick'em results

BCMoore: 23 of 30 (76.67%)
Total Humans: 232 of 310 (74.84)


Combined 3A/1A:

BCMoore: 49 of 61 (80.33%)
Total Humans: 1312 of 1742 (75.32%)


So you claim the humans were superior to Moore's model based on cherry picking stats that benefit your argument. Also not surprising you jumped to the 3A board because it doesn't look like the 1A pick'em gave your argument much to work with. Looking at the total compilation of human to computer results it certainly doesn't give validity to the argument Moore is inferior. A computer model shouldn't be held to a litmus test that it outguesses each individual human all the time. It's value should be determined on how it fairs over a duration of time against the sum of human predictions. The question is if you compile the human scores and take that against Moore would the humans outscore the computer routinely? The total human score neither beat Moore in the 3A or 1A polls.




PS: If you are going to call people out on spelling then don't show such a lack of it on your own. watermellon [sic] - what was that about double consonants again? middleschool [sic] develope [sic] accomidate [sic] intersquad [sic]

That's just from your posts yesterday. Frankly, I could care less about spelling on a message board but if you're going to be anal enough to hold others to a higher standard then don't be oblivious to that standard in your own post. Cheers
This post was edited on 12/9 10:05 AM by CP84

 

[cidhawkeye]

Don't forget about team camps in the summer, gotta be getting close to 100 games by now.

 

[bcmoore87]

FWIW:

BCMoore Rankings

regular: 1283/1550 82.77%
playoff: 154/186 82.80%
total: 1437/1736 82.78%

usually close ratios between regular season and playoffs

 

[ghost80]

CP,

Unless you can pick up the level of your arguments, I'm losing interest with you.

Sorry but much of your repeat of how the computer was CLOSE on a bunch of toss up games I didn't even read mostly because it doesn't matter. You win, you lose....nothing in between. A large number of humans seems to do a better job of correctly picking these close games. It is insulting to them to whine about how close the computer was on these "toss ups". Here is something a very successful coach once told me. "I don't believe in moral victories in close loses. If you take the word CLOSE and remove the C you are left with the word LOSE." Maybe CP should think about my coaches philosophy.

Your accusation of cherry picking is hypocritical and makes little sense. I picked 3A for a couple reasons. They had 20 full time and 10 part time contestants. A larger statistical number than 1A and 2A combined. Moore also got 83.8710% right in the 3A picks. According to you Moore got 80.33% right of all 1A to 3A (why did you cherry pick 1A to 3A). Why would I hand pick 3A when Moore got a higher percentage right in 3A. In 1A Moore only got 77% right and finished in a 3 way tie 5th out of a total of 16 full time and part time people? 2A only had a grand total of 12 people play. See a trend?

CP, you contradicted yourself again by using all the part time people in the analysis you used for 3A. I had use only the full time people that the poster had used in declaring the winners. I'm guessing he didn't include the part timers because then he would have had to declare his winner as a person who went 1-1 with 100% accuracy. You have been critical of me for using data from too small of statistically significant pools yet you suddenly want to include participants who guessed on 1, 8, 16, 20, ect games instead of the full timers like the poster (not me) did. Why would you do that? Folks the answer is easy....he is bending the stats to meet his agenda. In 3A Moore correctly guessed 83.87% and finished tied for 11th of 20 Full Timers or 15th out of 30 for both FT and PT. The other 19 full time humans averaged 85.25%. That is significantly higher than Moore. The 10 part time players who offered incomplete skewed data (remember 1 for 1) only managed 79.27% If you combine both and include the people with incomplete data then the computer wins 83.87% to 83.76% but one thing doesn't change...in rankings among the other player, Moore doesn't finish near the top.

How twisted is CP84s logic of including the part timers? What if next year I pick one game from each class. I select the most lopsided 1 seed -4 seed matchups and likely finish a perfect 6-0 100% for the playoffs. CP seems to think it would be fine to proclaim myself the victor over Moore's ending percentage. I doubt many (including myself) though would crown me the winner.

I did get a good laugh over CP pointing out my terrible spelling and grammar. You are a little late to the party on that one CP!! I've justly criticized for that many times before...even here on Iowa Preps. Too many years of having underlings proof everything I do because I'm told that my time is to valuable to worry about spelling and grammar If you only found those few problems in my post then you must be in about the same shape as me. Looks like at least we have one thing in common....we both suck at spelling and grammer!!



This post was edited on 12/11 6:44 AM by ghost80

 

[CP84]

Let's have a look summary of your arguments:

1) BCMoore model averages ~83% correct cumulative for the season and playoffs which is higher than the average of Sagarin and other models utilized by the BCS but discredited by you since 83% is is lower than entirely unrelated sectors. So this is what you define as a high level argument?

2) "You win, you lose....nothing in between." You really don't understand prediction analysis in sports do you? The purpose is to predict score variances. What's a better prediction in this scenario:

A) Team A will beat Team B by 30
B) Team B will beat Team A by 0.41
Actual Outcome: Team A wins by 3

If you think prediction modeling is limited simply to an up and down vote based on which team will win then you need to take a remedial course in probability.

Also, what does your "coaches" [sic] philosophy regarding his team competing in a game have to do with prediction analysis? Once again you completely miss the boat by comparing kiwi to tangerines.

3) "CP, you contradicted yourself again by using all the part time people in the analysis you used for 3A."
Why wouldn't I use part-time people in the sampling? It would be skewing the data even more to throw out their guesses in the total. If a participant has fallen behind in the competition they are less likely to continue participating in future rounds. In other words you want to throw out the poor picks to boost the human totals. A remedial course in statistics would also do you wonders. Throwing out a small sample size when comparing individuals in a game is apples and oranges to adding the sum of all human picks comparing to the sum total of computer picks.

Quite frankly, using the pick'ems as your reason to attempt to discredit a model is flawed in itself. The control group isn't really a controlled group. You have many participants that are relying on BCMoore's system as a benchmark to enter some of their own predictions. If you look at the 3A pick'em there are two individuals who predicted score outcomes to go along with their win predictions. When you compare their guesses to Moore's, the majority are very comparable to Moore's. So in other words, a human that isn't familiar with some of the teams is likely deferring to Moore's system to aid them making their own picks. Either that or you believe it's just a matter of coincidence.

4) "Too many years of having underlings proof everything I do because I'm told that my time is to [sic] valuable to worry about spelling and grammar"
This is just hilarious. Other than sounding buffoonishly infatuated with your own self-worth, I'm not sure what sector you work in where you can make this claim? I've been a director for a number of years and besides not referring to employees as the politically incorrect "underlings" I wouldn't make the claim you can suffice in the business world having said "underlings" proof all documents including E-mails. The CEO of my company wouldn't make this claim. I'm calling BS. It's one thing to type shorthand and/or make occasional errors on an internet forum but to make that claim in the corporate world is unrealistic and naive.


The bottom line is BCMoore has a proven track record of routinely picking >80% of games. If you read BCMoore's link you would have seen that average is higher than models used by the BCS process. You're entitled to your subjective opinion regarding computer prediction analysis but the track record stands for itself and most individuals who understand an ounce about prediction in sports would agree that >80% is a very viable model.

This post was edited on 12/11 12:29 PM by CP84

 

[cidhawkeye]

Time for you two to jump into the octagon together.

 

[spook78]

I know who I'd pick in that one, And I wouldn't need Moores prediction either. Althou he'd probably throw it out cuz his computer would predict a rout if it had the facts.

 

[CP84]

Originally posted by spook78:

I know who I'd pick in that one, And I wouldn't need Moores prediction either. Althou he'd probably throw it out cuz his computer would predict a rout if it had the facts.

Oh boy, you can't even have debate without someone turning it into an internet tough guy competition.

 

[DarkThunder#61]

 

[bcmoore87]

FWIW:

http://en.wikipedia.org/wiki/Survivorship_bias

Entertaining thread - good work to all (and keep it respectful!)

 

[ghost80]

CID,

He wouldn't have a chance against me in the Octagon.

You forget that I'm a ghost. Unless ECTO-1 was sitting ringside I'd slime C3P0 at will until he shorted out and his golden hide lit up like a Christmas Tree.

I have learned something from CP. If I coddled my "underlings" the way he apparently does, I'd be able to sit at my desk too at 9:59AM and post a temper tantrum on Iowa Preps. Word of warning CP... if your "associates" find out what you are doing while they are busting their hump to try and make you look good....you'll have someone ripping harder on you than I have.