Round 12 Updated Big 12 Projections

K-State is closing in on a stranglehold on the Big 12 race. We really really really need Iowa St. to pull an upset and we pretty much need to win the rest of our games. It is not over but if K-State wins this weekend it will be close.







As for trends, Iowa St. has been trending down and K-State trending up. Right now Iowa St. is projected to have only a 30% probability of winning against K-State in Manhattan this weekend. A home loss by K-State would be very significant.

Agree CJ, if ISU doesn't beat KSU this Sat, don't see much chance of even a tie for KU. Is always some hope, since have been some crazy outcomes - ISU might be motivated for payback for their home loss to KSU.

any idea why your numbers are so dofferent from asteroids? am sure its bias but asteroids do seem more right to me. espec since with a ksu loss this weekend ( at some point their streak has to end since they are just not that good) and a ku win at home over ksu they are tied pending all remaining games. that just doesnt feel almost insurmontable to me.

not that anyone cares abt my opinion but probability aside for the moment it really feels like we go 5-1 from here and they go 4-3 so we both end up with 13 wins instead of us at 12 and them at 14 in your model so just a one game switch for each of us. while i like their players theyve never won before, their coach is terrible and the pressure i think will all be on them. plus they just lost one of their best players and some reversion to the mean after this streak is called for (remember they lost their first two games so what team are they?). lastly, on match 9 we play at 2pm and they play at 6pm. if they need that win to tie or beat us im betting oklahoma.

btw to add to the march 9 drama tech plays at isu at 2pm.

I believe Asteroid uses one of Sagarin’s base ratings. I do not know what probability distribution function he uses. He can answer that. I use the DPPI raw predictor with a probability distribution function unique to each teams standard deviation or consistency factor assuming a normal distribution (which is not really an assumption, it is demonstrable truth) My method is pretty standard and basic, nothing too complicated or exotic. Asteroid may use a normalized distribution rather than a unique distribution by team. That will certainly have an impact since the distribution varies pretty widely by team. As for Sagarin’s rating vserus DPPI it is pretty much a preference. I do not know Sagarin’s algorithm (nor does anyone else since he has not divulged it). As a grad of MIT his pedigree is obviously first-rate and he has been among the most notable statisticians for many years. My DPPI is far more nascent yet born from 40 years of being a student of the game. I am quite proud of my DPPI and it has proven reasonably efficacious. For me it has the advantage of i can run it anytime i want. I rather enjoy the variety of systems out there. I participate in Massey’s comparison with about 65 other computer models. The vast majority of those are all in pretty strong agreement with very slight differences. There are a few that seem to be pretty whacked out. But it is important to understand that the rating is only one aspect. The probability distribution function is equally important when calculating actual game probabilities.

So having said all that, I personally love that this sight has two different methodologies. The only absolute is that neither are correct because by definition they are probabilities which is only meaningful in the range but not the mean. Asteroid is brilliant in the way he processes the data and I always look forward to seeing how it compares to mine since that provides even greater insight. I wish there were 30 different models. That would provide a dimensional analysis that would give even better insight.

Quite frankly, I do not know where Asteroid finds the time to compile all of the data he does. He tracks 15 or so different methods and calculates the projections for each game using each method. He is obviously a workaholic or brilliant beyond description at compiling massive amounts of data. By contrast I track only one method and I have the source code and it consumes a fair amount of my time. Tracking all he does I would never sleep.

CJ - I personally love that this site has two different methodologies also. Great given the small board count.
Some (not you) say that KSU isn't that good. They got to the Elite 8 last year without Dean Wade in the NCAA tourney - they have the same team back (except for Amaad Wainwright's 2 points in E8 game). Contrast that with KU's E8 win - the two players now possibly available (Garrett and Vick) scored 18 of 85 points and got 5 of 46 rebounds. DOK had 9 & 8 and de Sousa had 4 & 10, but they are not available.
big g - the March 9th game times are 1pm CT for Baylor at KU and TT at ISU, and 5pm CT for OU at KSU. Imagine is Senior Day for all - KSU has 3 senior starters (Wade, Brown, & Stokes), so seriously doubt that OU will beat them.

Thx JR very compelling alternative point of view. Ksu is missing cartier as well which is a big loss at crunch time i think. While I love their seniors i do lack confidence in their coach and see him pressing more and more as he gets closer to dethroning us. They go thru long periods of no offense and that shld catch up with them. While i think they may have played their best conf games i think ours are still to come. We have our core now, coach now understands how to optimize their talent and they get him now. So im on record. Three way tie for first. And for ksu isee senior day just adding to the pressure for them on march 9 so they lose to okla to make it a tie. Ps sorry my game times were ET which is where i am. Love to basketball knowledge and passion on this board.

OK big g, so let it be written .. . .

RC

Good one Bay. I thought maybe you were going to post what time the March 9th games are in PST.
big g - I don't like KSU's coach either - he drives me crazy on the sideline. I agree that losing Diarra will hurt KSU some - the ESPN color guy on the KU-KSU game called Diarra the MVP of game. He was the one that solved KU's zone. In other games his main purpose was to spell Stokes and others.
I would take a 3-way tie for first.

I have a question for the prediction modeling team. Could they work into their models a "who actually plays" on historical games as well as who is expected to be available on future game. Some way to "assign" a figure to the value of a previous outcome based on the relevance to the current game. For example you could attribute a 12/13s weight to games before Dok was hurt or something along that line. Maybe say a % of those that actually played. It would be tough to track but might yield better results.