The Wiz Posted December 13, 2015 Share Posted December 13, 2015 Now that I have a minimun amount of data , I thought this would be a good time to look at the season. For those following me on the board you know that the Bills are a C+ team with a current range of B to C-. Also the A-10 report card shows that we are not an elite A-10 team...but you knew that already. But what do these things mean for wins and losses. Let's take a look..... I am showing we should win 3 of the 4 remaining OOC games. That means we finish at 8-4 through Dec. Then in A-10 play I am showing we win 5 games......conference record...5-13...overall rec..13-17...a definite thud. But these numbers assume no change from today ...we play in March the same as we play today. But what is the potential for the season? What is the upside? What happens if JC coaches them up...the team grows (as in experience and knowledge not height) and matures...we cut down on TOs and mistakes...we continue to make FTs at 70%+...we get an extra couple of rebounds / gm....we get a good bounce occasionally...maybe even a good ref call (.hmmm ..let's not go crazy) The potential is another 5 wins...conf rec....10-8....overall 18-12. Most would agree this would be a successful season. We are not talking about major changes, just a few small things here and there. ...an extra few throw ..an extra rebound... an extra timeout ( everyone can learn)... a few less turnovers.. Now before people go running off and saying...The Wiz said we are going to be 18-12... remember this is about potential. Bottomline...if we improve 18-12 ....if we stay the same....13-17...if we have less than 13 wins then I would considerr that regression. Should be an interesting season. Quote Link to comment Share on other sites More sharing options...
bonwich Posted December 13, 2015 Share Posted December 13, 2015 Bottomline...if we improve 18-12 ....if we stay the same....13-17...if we have less than 13 wins then I would consider that regression. I would have thought that you, of all people, would consider this regression: Quote Link to comment Share on other sites More sharing options...
The Wiz Posted December 13, 2015 Author Share Posted December 13, 2015 I would have thought that you, of all people, would consider this regression: I like Pi...cherry or apple Quote Link to comment Share on other sites More sharing options...
Old guy Posted December 13, 2015 Share Posted December 13, 2015 That is clever Bonwich, now can you say in English what it means. I mean it could be anything from the regression in the price of oil after the Fed increases rates to the rate of reproduction of fruit flies if the climate changes by +2 degrees next year. What is it? Quote Link to comment Share on other sites More sharing options...
billikenblue Posted December 13, 2015 Share Posted December 13, 2015 Now that I have a minimun amount of data , I thought this would be a good time to look at the season. For those following me on the board you know that the Bills are a C+ team with a current range of B to C-. Also the A-10 report card shows that we are not an elite A-10 team...but you knew that already. But what do these things mean for wins and losses. Let's take a look..... I am showing we should win 3 of the 4 remaining OOC games. That means we finish at 8-4 through Dec. Then in A-10 play I am showing we win 5 games......conference record...5-13...overall rec..13-17...a definite thud. But these numbers assume no change from today ...we play in March the same as we play today. But what is the potential for the season? What is the upside? What happens if JC coaches them up...the team grows (as in experience and knowledge not height) and matures...we cut down on TOs and mistakes...we continue to make FTs at 70%+...we get an extra couple of rebounds / gm....we get a good bounce occasionally...maybe even a good ref call (.hmmm ..let's not go crazy) The potential is another 5 wins...conf rec....10-8....overall 18-12. Most would agree this would be a successful season. We are not talking about major changes, just a few small things here and there. ...an extra few throw ..an extra rebound... an extra timeout ( everyone can learn)... a few less turnovers.. Now before people go running off and saying...The Wiz said we are going to be 18-12... remember this is about potential. Bottomline...if we improve 18-12 ....if we stay the same....13-17...if we have less than 13 wins then I would considerr that regression. Should be an interesting season. Quote Link to comment Share on other sites More sharing options...
billikenblue Posted December 13, 2015 Share Posted December 13, 2015 No way do we end 10-8 in the conference..the eye test based on the quality of the A-10 and our inability to get better as you suggest does not support your theory. Love your stats Wiz but these are stats based on wins against very weak teams. Quote Link to comment Share on other sites More sharing options...
majerus mojo Posted December 13, 2015 Share Posted December 13, 2015 No way do we end 10-8 in the conference..the eye test based on the quality of the A-10 and our inability to get better as you suggest does not support your theory. Love your stats Wiz but these are stats based on wins against very weak teams. His stats said we'll go 5-13 in conference Quote Link to comment Share on other sites More sharing options...
bonwich Posted December 13, 2015 Share Posted December 13, 2015 That is clever Bonwich, now can you say in English what it means. I mean it could be anything from the regression in the price of oil after the Fed increases rates to the rate of reproduction of fruit flies if the climate changes by +2 degrees next year. What is it? Well, in English, it's a standard set of equations for multiple regression. And you're right, it could be anything. You really kinda sorta have to identify what the source of your data set is and what hypothesis you're seeking to test. Quote Link to comment Share on other sites More sharing options...
billikenblue Posted December 13, 2015 Share Posted December 13, 2015 His stats said we'll go 5-13 in conference Read his note again I was basing my comment on the potential with improvement etc to get 5 more wins Quote Link to comment Share on other sites More sharing options...
Old guy Posted December 13, 2015 Share Posted December 13, 2015 Billiken blue potential is not reality until the time comes when it becomes reality. For example, it is possible that SLU will win the NCAAB tournament this year, however it is highly improbable it will do so. Speculation, what happens if... is conjecture. It is fun to do at times but it is not real. Again for example what if you buy a Powerball ticket today and win the big jackpot on Wednesday. It is possible but it is very unlikely. However you can play mind games for as long as you want trying to figure out how you will spend the money you are going to win in Powerball next Wednesday if you buy the ticket today. In other words it is not statistics, it is conjecture. And by the way, there is nothing written in stone using statistics either. The closer we get to the end of the season the more likely we will be to be able to predict exactly where we are going to end it. At this time there are all kinds of possible, although not likely, outcomes that could happen IF and only IF certain obstacles can be overcome. It can be fun to speculate using numbers. Quote Link to comment Share on other sites More sharing options...
Major Majerus Posted December 13, 2015 Share Posted December 13, 2015 Well, in English, it's a standard set of equations for multiple regression. And you're right, it could be anything. You really kinda sorta have to identify what the source of your data set is and what hypothesis you're seeking to test. I thought you were a foodie; if you wanted to be an engineer, or worse; a statistician, how did you wind up writing about food, of which I've seen articles with your byline as far away as Tampa? Quote Link to comment Share on other sites More sharing options...
David King Posted December 13, 2015 Share Posted December 13, 2015 18-12 would be a disaster for the Bills! It would guarantee that Crews remains. 12-18 is much more preferred. Quote Link to comment Share on other sites More sharing options...
The Wiz Posted December 14, 2015 Author Share Posted December 14, 2015 No way do we end 10-8 in the conference..the eye test based on the quality of the A-10 and our inability to get better as you suggest does not support your theory. Love your stats Wiz but these are stats based on wins against very weak teams. Actually the stats are based on the whole season...good teams and the bad teams. So it looks at the losses to WSU and Lville too. So here is how I arrived at the projections. On the 5-13 conference forecast (scenario 1), it shows we will be favored in 5 games. In the second forecast...It shows the same 5 wins plus 8 losses. ..plus 5 games we won't be favored but might lose by 1 or 2 points. As this within the margin of erroro or what is known as the "luck" factor... A ball touches an opponent on the way out of bounds on a crazy hop and the Bils get the ball, ..a rebound we didn't see falls in our lap , a ref misses a call in the Bills favor...a free throw bounces on the rim 4 times and goes in....or JC coaches them and they get just a tad better...Through any number of circumstances there are 5 more games that can be won. Btw many forecasting models use 3-5 points as the X or unknown factor. . In my projection , I went more conservative and used 2 pts. For those that don't believe any of the above is possible, then you better stick with scenario 1 ...5-13. For those that see potential... we can do better than 5 wins maybe as high as 10. Quote Link to comment Share on other sites More sharing options...
bonwich Posted December 14, 2015 Share Posted December 14, 2015 I thought you were a foodie; if you wanted to be an engineer, or worse; a statistician, how did you wind up writing about food, of which I've seen articles with your byline as far away as Tampa? I have a degree in mathematical computer sciences from alma mater and one in econometrics from Wisconsin. It prepared me perfectly to be a food writer (especially the little store I found in Madison that sold me '59 and '64 Latour, '61 Beycheville and '70 and '75 Lafite. One needs hobbies, after all.) Quote Link to comment Share on other sites More sharing options...
WVBilliken Posted December 14, 2015 Share Posted December 14, 2015 Tough watching this team. A 500 season would be a big accomplishment but feasible with slightly fewer turnovers and slightly better rebounding. Quote Link to comment Share on other sites More sharing options...
Taj79 Posted December 14, 2015 Share Posted December 14, 2015 "The closer we get to the end of the season the more likely we will be to be able to predict exactly where we are going to end it." Genius, bay-bee, GENE-EEEEEEEEE-USSSSS! When is post time at the local track? Quote Link to comment Share on other sites More sharing options...
kwyjibo Posted December 14, 2015 Share Posted December 14, 2015 ...Through any number of circumstances there are 5 more games that can be won. Btw many forecasting models use 3-5 points as the X or unknown factor. . In my projection , I went more conservative and used 2 pts. For those that don't believe any of the above is possible, then you better stick with scenario 1 ...5-13. For those that see potential... we can do better than 5 wins maybe as high as 10. While most people are focused on the "point" accuracy of ranking teams, I have been long interested in the variance. The actual average absolute error of any prediction system this time of year is 9-10 points (see http://www.thepredictiontracker.com/bbresults.php). This will go down likely to somewhere closer to 7-8 points but the best predictors are no more accurate. Remember that this is the AVERAGE error, the mean square error of all the systems listed is over 100. I can see why you may not want to see all this variance in estimating a whole season (as the variance across 20 games is different than the variance across one game) but I think it is important for people generally to realize that this "inaccuracy" of all prediction systems is greater than many of the predictors like to let on (and hence do things like forecast with smaller acknowledgement of potential error). Quote Link to comment Share on other sites More sharing options...
hsmith19 Posted December 14, 2015 Share Posted December 14, 2015 Looking at the schedule, I don't see 10 potential A-10 wins in any kind of realistic sense. There might not be a top 25 team in the A-10 this year, but over half of the league could end up in the top 100. Where are the five extra wins you think are possible with improvement? I picked 13 at the beginning of the year, and that's what I'm sticking with. I think that would be a reasonable improvement, given the tougher schedule this year. Quote Link to comment Share on other sites More sharing options...
Old guy Posted December 14, 2015 Share Posted December 14, 2015 Kwjibo, please remember that there is nothing final or terminal about predictors, you can juggle them and modify them at will. You must remember to backtest your modified predictors against prior seasons results to attempt to find the validity of the results obtained using the modified predictors vs the standard predictors. Suggest that you use the initial 7 or 8 results for data and then see how well your modified predictor does with the rest of the season. Remember to use only a single variable change in each test run, if you change several items you will never be able to determine what caused the changes in the results. The site you provide is interesting and allows comparison between multiple existing prediction systems. However please remember that given enough time and effort, any such system can be improved to a greater or lesser extent since all predictive systems contain arbitrary assumptions which may be improved on. Quote Link to comment Share on other sites More sharing options...
slufan13 Posted December 14, 2015 Share Posted December 14, 2015 Every time I look at the schedule I come up with somewhere around 12-14 wins. So agree with hsmith that 13 still seems like a good number. Quote Link to comment Share on other sites More sharing options...
kwyjibo Posted December 14, 2015 Share Posted December 14, 2015 The site you provide is interesting and allows comparison between multiple existing prediction systems. However please remember that given enough time and effort, any such system can be improved to a greater or lesser extent since all predictive systems contain arbitrary assumptions which may be improved on. This missed the whole point of my post--that despite all the improvement and changes in how games are predicted (and Vegas incorporating computer models) there is still 7-8 points average error in predicting any single game. That was just a reminder of exactly how accurate these predictors are which I just wanted to state as Wiz mentioned correcting for a 2-5 pts."X" factor (which in his context may be fine). I wanted to be clear that the actual accuracy and variance of game predictions is much greater. Quote Link to comment Share on other sites More sharing options...
Old guy Posted December 14, 2015 Share Posted December 14, 2015 I entirely missed your point, apologize Kwyjibo. What I was trying to point out is that a lot of these improvements and changes in how games are predicted (and Vegas incorporating computer models) are still based upon value judgments, and these can always be improved. By the way the use of computers in predictions does not give you an automatic advantage. The actual prediction accuracy depends on the assumptions made and the data used, all the computer does is to make the calculations faster and easier. Quote Link to comment Share on other sites More sharing options...
The Wiz Posted December 15, 2015 Author Share Posted December 15, 2015 While most people are focused on the "point" accuracy of ranking teams, I have been long interested in the variance. The actual average absolute error of any prediction system this time of year is 9-10 points (see http://www.thepredictiontracker.com/bbresults.php). This will go down likely to somewhere closer to 7-8 points but the best predictors are no more accurate. Remember that this is the AVERAGE error, the mean square error of all the systems listed is over 100. I can see why you may not want to see all this variance in estimating a whole season (as the variance across 20 games is different than the variance across one game) but I think it is important for people generally to realize that this "inaccuracy" of all prediction systems is greater than many of the predictors like to let on (and hence do things like forecast with smaller acknowledgement of potential error). While "point accuracy" is an important variable, it is not the most important in my system. Getting the wins and losses right is the most important. For example, if the Bills were favored to win by 2 and lost by 1 that would be worse than if the Bills won by 10. ...even though in the first case the point spread was only off by 3. In my example in the 1st post above, I tried to find potential wins by expanding the spread only 2 points...and that doubled the number of Bills wins from 5 to 10. Can you imagine if I had gone to 8 or 9 pts...that would have had us winning almost all the games and a 1st place finish causing a complete board meltdown...Lol. I decided to go with 2 pts to go after the borderline games ...the ones where we had a good chance to win if things went right or we improved by a small amount....2pts still gives us a 46% chance of winning. Even though I have been doing this for a long time, one of the things that still amazes me is how many games hover around the point spread (for most of the game) . Even on the point spread misses, many times that happens in the final few minutes when the game gets out of hand. Bottomline,,,,The primary purpose of the forecast is... #1.. to pick the correct winner ...#2.. to guage the relative strength of the pick. When the Bills are picked to win by 1...you know it's going to be a nailbiter... If we pull away in the final 2 minutes to win by 10 that doesn't change the fact that for most of the game it was a handwringer. and #3...to use the spread as a trend forecaster....If we do better than the spread ...the Bills had a good game...if we beat the spread over a few games we are trending up....If on the other hand we continue to just make the spread then this is an indication that we are where we are supposed to be. The grades are supposed to be an easy way to translate the raw numbers into an understandable guage as to how the team is doing. Since much on a college campus is tied to grades , I thought it made sense to have a grading system for basketball too . Quote Link to comment Share on other sites More sharing options...
The Wiz Posted December 15, 2015 Author Share Posted December 15, 2015 Looking at the schedule, I don't see 10 potential A-10 wins in any kind of realistic sense. There might not be a top 25 team in the A-10 this year, but over half of the league could end up in the top 100. Where are the five extra wins you think are possible with improvement? I picked 13 at the beginning of the year, and that's what I'm sticking with. I think that would be a reasonable improvement, given the tougher schedule this year. I will give you the 5 projected wins as well as the 5 potential wins because if I don't someone for sure will ask. Projected conf wins as of Dec 14,,,which means it is subject to change Home...GM Duq St. B Away....GM, LaS Potential conf wins...this means we are picked to lose but it is close enough where we have a good chance. Home...Dav, Rich, Ford Away....Duq, St,B That adds up to 10 Quote Link to comment Share on other sites More sharing options...
Old guy Posted December 15, 2015 Share Posted December 15, 2015 Wiz, the idea of a median value surrounded by tails on both sides (deviations from the median) has not crossed the minds of a lot of people who think of rankings and spreads as a quasi magical determination (not approximation) of reality. Basketball, as any other sport, is a very complex system that depends upon a myriad of details that are highly variable, are not easy to approximate, and are not correctly reflected on the statistics of prior games. This is where judgment and expertise has to be used to fill gaps on the generally available data. Some people, yourself included, are very good at this, others are not so. Not one prediction system is exact, all systems provide approximate results only. Reality is not predictable, but approximations of reality can be achieved in a regular basis by people, like yourself and others, that have developed a feeling for these fuzzy details. 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.