The Wiz's preseason predictions

[The Wiz]

Hi all....This is the official start of my basketball season.

For those that are new , this is the first of 3 forecasts....preseason...around the 8th game....and at the beginning of the A-10 season.  This preseason prediction is based on a Bayesian model....the next 2 on real data. I use a letter grade system which groups similar teams together.  As the season starts to unfold I use the first 8 games to meld the Bayesian model with  the real data model  and drop the Bayesian after the first 8 games....btw that is 8 real games....so no Harris or Rockhurst.  As the season goes on and I have more data , I become more accurate ( at least that is the way it is suppose to work).

So the big question on everyone's mind is how are the Bill's going to do?...The simple answer is...better.  At the end of last season  ( we finished with a D grade) I had said that when a team has a major change...5+ players...that historically the team can improve 2 letter grades...in this case a B.  But I also said  that a 2 letter improvement is a best case scenario. ....everything has to go just right...everyone shows up and is available for every game, team gels , the team adjusts ...to College BB, to school, to Ford etc.  We have already seen some blips. However at this point , I would still say max potential is a B.  However, I am going to start the Bills off with a C+.  Remember everything has to go perfect for a B.  While C+ may be disappointing for some it is a huge jump from D.  Again , this is preseason and the ability to change your grade, as all students know, is at the beginning of the semester. So a good start would set the tone for the year.  So.....now that you know the 8th game is important, I am going to change it this year. Why?  Because the 9th game is Murray St. ...and when we play them they will not have 8 games....So the first true data game will be the 10th game against Ore St.

So now let's move on to the A-10. Everyone thinks the A-10 is down this year. ....not quite. Over the years the A-10 has averaged around B+ ...last year it came in at B.  This year it is starting out at B again.  So what's happening?  The top of the league has weakened and the bottom has strengthened. So the league will look different this year....yet overall it will be about the same in difficulty.

Let's look at the the A-10 rankings

RI...........................................................A

Day.. St. B.. VCU...................................B+

Dav ...Rich... .St. J................................B

GW.........................................................B-

SLU...GM...LaS...UMass........................C+

For...........................................................C-

Duq.........................................................D+

 

So based on these grades ....RI...Big Dance.........Day , St. B , VCU.....NIT...........Dav ,  Rich , St. J...  1 or 2 of these NIT

 

We will revisit these forecasts in mid Dec....

To paraphrase an ancient Chinese philosopher...May you live in interesting times

This will be an interesting year .

 

 

 

 

 

 

 

 

 

 

 

[CenHudDude]

Thanks as always Wiz.  Looks like pre-season you are ranking us between 9th and 12th in the conference.

[Old guy]

Wiz, I posted this on the prediction thread for the Seattle game but it appears worthwhile to post it here again:

"Wiz if your Bayesian system assumes all hands will be on board, whatever it predicts may well come out wrong because it looks like 3 of the hands will be out of play. If you can figure out how  we are going to be without Bishop, Henriquez, and Graves (who was not going to be in anyways until the end of the 1st semester), it may be OK to publish your findings. Otherwise we are probably better off without the preliminary Bayesian prediction and should go with no prediction at all straight until after 8 games are played and you can base the results in actual game performance to date.

Your Bayesian system is working what amounts to be like an actuarial system using a limited number of samples, always a tricky situation because of the limited number of people involved. If you subtract 3 players out of the mix the current assumptions used in the system that all hands are present and available, the predictions may not work well or at all. Again it may be better if you wait for the first 8 games to be played before sharing your findings. This is particularly important for the Bills this year because the 3 missing players were presumably going to carry a good portion of our offensive capacity in court."

I agree with you that we will be better this year, the presence of Goodwin, French, Johnson, and Bess in the team will make this happen. But really, after the unexpected loss of a good portion of our offensive game with Bishop, Henriquez and Graves absence from play, I cannot see how a statistical system assuming all hands are available and present during the games can predict much of what may be happening for the first 8 games. Would it be possible to redo your calculations assuming no bench strength or depth, 7 players available for play (including Hines) for probably the full initial 8 games, with the possible early addition of Welmer to playing status a bit prematurely and due to the necessity to increase the warm body count. It would be highly appreciated if you could reformulate your initial preseason forecast to fit this new set of assumptions.

[The Wiz]

35 minutes ago, Old guy said:

Wiz, I posted this on the prediction thread for the Seattle game but it appears worthwhile to post it here again:

"Wiz if your Bayesian system assumes all hands will be on board, whatever it predicts may well come out wrong because it looks like 3 of the hands will be out of play. If you can figure out how  we are going to be without Bishop, Henriquez, and Graves (who was not going to be in anyways until the end of the 1st semester), it may be OK to publish your findings. Otherwise we are probably better off without the preliminary Bayesian prediction and should go with no prediction at all straight until after 8 games are played and you can base the results in actual game performance to date.

Your Bayesian system is working what amounts to be like an actuarial system using a limited number of samples, always a tricky situation because of the limited number of people involved. If you subtract 3 players out of the mix the current assumptions used in the system that all hands are present and available, the predictions may not work well or at all. Again it may be better if you wait for the first 8 games to be played before sharing your findings. This is particularly important for the Bills this year because the 3 missing players were presumably going to carry a good portion of our offensive capacity in court."

I agree with you that we will be better this year, the presence of Goodwin, French, Johnson, and Bess in the team will make this happen. But really, after the unexpected loss of a good portion of our offensive game with Bishop, Henriquez and Graves absence from play, I cannot see how a statistical system assuming all hands are available and present during the games can predict much of what may be happening for the first 8 games. Would it be possible to redo your calculations assuming no bench strength or depth, 7 players available for play (including Hines) for probably the full initial 8 games, with the possible early addition of Welmer to playing status a bit prematurely and due to the necessity to increase the warm body count. It would be highly appreciated if you could reformulate your initial preseason forecast to fit this new set of assumptions.

For one thing , Graves is not figured in the calculation because we know he won't be playing 1st semester.  After that no one really knows.  Do you know for sure that there will be 2 players missing for the next 8 games?  For that matter even the next game? There are always players missing on every team  during a season.  When a player who is scoring 8 points a game is out , you don't subtract 8 ppg off the final score. There will be subs who will replace the missing points.  If they don't then that will be factored into the system as we go  and it will correct. 

Since we know nothing about who will be missing or for how long (if at all) and since I will be replacing the Bayesian model quickly with real data,  I will leave you in the hands of Descartes...........

"It is very certain that,  when it is not in our power to determine what is true, we ought to act according to what is most probable."

--------------Rene Descartes  1637

[Old guy]

This is correct Wiz, I have absolutely no knowledge about when and if these players will be allowed to play. As a matter of fact I do not understand the reasoning behind their being held from play now. Therefore your assumptions may be correct and the model may work as expected. On the other hand there is a chance, that will grow with every single game they do not play in, that they will not be allowed to play during any of the initial 8 games, in which case, your model's assumptions will be incorrect and the model predictions faulty. 

Since no one knows for sure we can take your model as correctly stating the situation until further notice. At any rate after 8 games the Bayesian model will be replaced by an actual performance based model for this year which will correctly model how the team has played up to the last game. I honestly desire to be the one making the  wrong assumptions here, that is that the players will be allowed to play before long, however I fear this will not be the case. 

This is a quandary that cannot be solved statistically, we just have to wait and see.

[The Wiz]

5 minutes ago, Old guy said:

This is correct Wiz, I have absolutely no knowledge about when and if these players will be allowed to play. As a matter of fact I do not understand the reasoning behind their being held from play now. Therefore your assumptions may be correct and the model may work as expected. On the other hand there is a chance, that will grow with every single game they do not play in, that they will not be allowed to play during any of the initial 8 games, in which case, your model's assumptions will be incorrect and the model predictions faulty. 

Since no one knows for sure we can take your model as correctly stating the situation until further notice. At any rate after 8 games the Bayesian model will be replaced by an actual performance based model for this year which will correctly model how the team has played up to the last game. I honestly desire to be the one making the  wrong assumptions here, that is that the players will be allowed to play before long, however I fear this will not be the case. 

This is a quandary that cannot be solved statistically, we just have to wait and see.

The 2 models are being melded together starting with game 1 ...by game 8 it is all real data. Since you have stated you have no new  information to add....I will stay the course and let it do its thing. The system is self adjusting each and every game.

"It is likely that unlikely things should happen"

......Aristotle  325 BC

[Old guy]

That is ancient wisdom, far from statistics Wiz

[Major Majerus]

Please quit talking to the Wiz like u r an equal; u r not.

[WVBilliken]

Way off.  The A10 is more than a 2 bid league.  I see 3 to 4 bids.  RI for sure, like VCU to get in as well as St. Bonaventure.  I figure four teams.  Hopefully  SLU or some other team makes it four.

 

[almaman]

Can't remember worse preseason predictions for league since we got stuck in it. Only 2 teams received votes in one of the main polls none in top 25. 

[The Wiz]

17 minutes ago, WVBilliken said:

Way off.  The A10 is more than a 2 bid league.  I see 3 to 4 bids.  RI for sure, like VCU to get in as well as St. Bonaventure.  I figure four teams.  Hopefully  SLU or some other team makes it four.

 

Actually , we are fairly close to one another.....In the B+ category, a team usually has about a 1 chance in 3 of making it to the dance.... With 3 B+ teams most likely there will be another bid to the Dance....So there is your 2 bid league.  In addition it would be relatively easy for a B+ team to move to an   A-  in  a preseason forecast. So there is your 4 bid league.   So I would agree with you in that we are a 2-4 bid league at this point. And we also agree on VCU and St. B as possibilities. If you need a 4th bid I would recommend Dayton and then we would be in total agreement.   Stay tuned for the update in about a month to see how it shakes out.

[Old guy]

6 hours ago, Major Majerus said:

Please quit talking to the Wiz like u r an equal; u r not.

This is your expert opinion in the matter, I see.

[BillsBeliever!!!]

Say Rhode Island and St Bonaventure are in the top 20 by the end of the regular season and are around 26-5. If the Billikens were to win the conference tournament, would we be a 3 bid league?

[The Wiz]

56 minutes ago, BillsBeliever!!! said:

Say Rhode Island and St Bonaventure are in the top 20 by the end of the regular season and are around 26-5. If the Billikens were to win the conference tournament, would we be a 3 bid league?

See the quote from 706PM on Sunday from this thread.