My unsolicited advice for those readers with a large ego is to become a forecaster. I am confident that the size of your ego will dramatically decrease when you see just how difficult it is to forecast. The reason I can write this so confidently is that for the last three years I have been a professional forecaster for a Fortune 500 company. In addition to forecasting over 2,000 items I have participated in plenty of training sessions aimed at improving forecast performance. One might think that with this experience that I might be an expert forecaster. I am not. I try my best to increase the probability that my forecast will be accurate, but a good portion of the times the actual demand doesn’t cooperate with the forecast.
Now that I have gone over my credentials (or lack of if you prefer) I am going to try my best to go over certain ways you can forecast the Jazz win total for next year. First I look over a few statistical tools for a top down forecast. For this I used total wins over the years where Coach Sloan was the head coach of the Jazz. After looking at those I looked at a bottoms up forecast by looking at the individual players on the roster for next year. For this analysis I used the Win Score per 48 minutes, which is just an estimate of the number of wins each player contributes to the team divided by the amount of time in a game. For more information on this stat please read the following from basketballreference.com.
Anyway, below are the forecasting techniques (in various levels of complexity) I used with the overall wins since the 1989-1990 season.
- Mean – A simple average of the 21 seasons is 51 wins per season
- Median – The middle number of wins was 53 wins in a season.
- 3 year moving average – The average number of wins over the past three years has been 52 wins per season.
- Regression Analysis - 47 wins (This type of data looks at past data points and tries to create a “line of best fit”that reduces the amount of error between the model and the data. )
- Simple Exponential Smoothing - 52 wins (This type of forecasting places greater emphasis on more recent data points. If you care I used an optimal alpha of 0.79)
- Linear Trend Analysis – 45 wins (This type of forecasting technique tries place the best straight line through the data.)
There are other models I could use, but in the interest of time I stopped at those six. Overall the models predicted a pretty small range between 45 and 53 wins. This speaks to the consistency of Jerry Sloan has had in producing winning teams. Now let’s see if the bottom’s up approach to the forecast matches what we got above.
Using the stat Win Score and more specifically WS/48 I looked at each player and tried to predict their WS/48 and the amount of minutes they will play. For some players like Deron Williams it was easy because I used the same minute (2802) and same WS/48 (0.177) to calculate the same Win Score of 10.3 as he had last year. That means that according to this statistic Deron Williams directly accounted for 10.3 wins for the Jazz last year.
There was no reason to mess with those numbers, but with other players I made assumptions detailed below.
- Paul Millsap will have the same WS/48 (0.151) but will play more (projected 2673) because of Boozer leaving
- Al Jefferson will have the same career WS/48 (0.119) and will play about the same minutes (2900) that he did in the 2007-2008 season.
- Andrei Kirilenko will have the same WS/48 (0.171) as last year, but will play more minutes (1802).
- Gordon Hayward was probably the most difficult player to forecast. I looked at the rookie year WS/48 for the past ten 9th picks in the draft. The average WS/48 was 0.0768, which included really good players like Andre Iguodala, Amare Stoudemire and Joakim Noah as well as busts like Patrick O’Bryant and Rodney White. The 9th pick of the draft seems to be very hit or miss with he lone exception being last year’s 9th pick Demar Derozan, whad a WS/48 of 0.066 over 1664 minutes. For Hayward I used the average WS/48 of the past ten 9th picks (0.0768) and then estimated he would play 1700 minutes.
- Raja Bell - I estimated 1500 minutes and then used his career WP/48 of 0.085.
- CJ Miles - I increased his minutes to what Wes Matthews played last year (2025) and kept his below average WS/48 of 0.061 the same. The Jazz could really exceed expectations if Miles improves his game since there will definitely be minutes available for him.
- For the other players (Price, Fesenko, Gaines, Evans andJeffers) I made a few minor adjustments, but nothing that would affect that Jazz win total by more than maybe a game or two.
I then made sure the minutes played for the team equaled 82 games X 5 players X 48 minutes per game or 19,680 total minutes. After all of those calculations and assumptions the total win score totaled up to50.1 wins. The 50.1 is in almost exactly between the 45 and 53 wins that we got above so that makes me think that it at least feels correct. Also, as a general rule I tend to agree with a bottoms up forecast approach since it takes more time and analysis to do. With all of that that being written my official prediction for next year’s Jazz team is 50 wins. Now I can just sit back, enjoy the season and see how wrong my prediction is since that is the life of a forecaster.