RPI Forecast

[DD3]

Timmerman tweeted earlier that RPI Forecast favors the Bills to win their next 13 games. I also thought their probabilities on final record were interesting:

http://www.rpiforecast.com/teams/Saint%20Louis.html

[GOSLU68]

Wow, no wonder you get no replies that is a lot to digest. This is Timmerman's work or yours? If we win the next 13 and come out with the record predicted that means we lose the last seven straight and Crews gets lynched. I doubt it will work out that way.

This should be a link or banner or goal at the top of the page.

[Old guy]

Wow, no wonder you get no replies that is a lot to digest. This is Timmerman's work or yours? If we win the next 13 and come out with the record predicted that means we lose the last seven straight and Crews gets lynched. I doubt it will work out that way.

This should be a link or banner or goal at the top of the page.

I do not think it works out the way you said. The winning probabilities in that table are most likely calculated as of now, in other words using the team's current stats to calculate the winning probabilities for each game. However whenever you win a game your team's stats change, even if just a little. If we win the next 13 games, the chances are that our stats will be improved, perhaps quite significantly, and therefore we are not likely to lose the following 7 games straight.

[kshoe]

Wow, no wonder you get no replies that is a lot to digest. This is Timmerman's work or yours? If we win the next 13 and come out with the record predicted that means we lose the last seven straight and Crews gets lynched. I doubt it will work out that way.

This should be a link or banner or goal at the top of the page.

Yup, Timmerman is running an RPI forecast website in his spare time!

This is the best site on the web as far as RPI goes. Too many sites worry about what is your current RPI when all that really matters is what is your RPI on selection Sunday. This site does that (ignoring the conference tourneys at this point) in addition to telling you what the current RPI is.

The key take aways for SLU are:

- If we finish 22-8 we'll have an RPI of about 35. That would likely be good enough for an at large bid. There is a 78.6% chance of finishing that good or better.

- The most likely outcome is finishing 23-7 but that is expected to happen only 23% of the time.

- Just look at our schedule. If we find a way to win the next 4 games (with 3 road games its going to be hard) I challenge someone to tell me when our next loss will be. Maybe the Philly trip to St. Joes and LaSalle? Maybe just a random upset. If we get through these 4 games I could see us heading into the March 1 game at VCU with a 26-2 record and a 13-0 conference record. But these next 4 are going to be hard.

[RUBillsFan]

- Just look at our schedule. If we find a way to win the next 4 games (with 3 road games its going to be hard) I challenge someone to tell me when our next loss will be. Maybe the Philly trip to St. Joes and LaSalle? Maybe just a random upset. If we get through these 4 games I could see us heading into the March 1 game at VCU with a 26-2 record and a 13-0 conference record. But these next 4 are going to be hard.

Really sucks that 2 of our last 3 games are at VCU and at UMass. I believe the NCAA committee sometimes looks at the record in the previous 10 games. I'm not sure they'll dig in enough to realize that 2 of the toughest games on our schedule are in the last 3.

If SLU is 26-2 going into those games, I will be over the moon.

[Zink]

I do not think it works out the way you said. The winning probabilities in that table are most likely calculated as of now, in other words using the team's current stats to calculate the winning probabilities for each game. However whenever you win a game your team's stats change, even if just a little. If we win the next 13 games, the chances are that our stats will be improved, perhaps quite significantly, and therefore we are not likely to lose the following 7 games straight.

Woof. I am on my phone and can't multi-quote, but this is directed at GOSLU68 as well...

Being favored does not mean that this predictor model expects us to actually win every game we are favored in. Each game is an independent event with a probabilistic outcome. Consider this illustration - we could be favored in all of our remaining games, with a 51% chance of winning each game individually. Despite being favored for each game, the model would NOT be predicting that we win out, rather it would result in an expected record of 9-9. In the same way, our predicted record builds in the probability of each game and calculates the likelihood across all possible outcomes, arriving at the most likely scenario given the current ratings. The way they estimate our final record as predicted by this model would basically be to add up each game's win probability (%/100) and add that sum to our current record.

I hope this brief explanation clarifies the picture somewhat...

[slu72]

We will lose a "random" game or two. Hopefully not to that a-hole Hurley.

[The Wiz]

Woof. I am on my phone and can't multi-quote, but this is directed at GOSLU68 as well...

Being favored does not mean that this predictor model expects us to actually win every game we are favored in. Each game is an independent event with a probabilistic outcome. Consider this illustration - we could be favored in all of our remaining games, with a 51% chance of winning each game individually. Despite being favored for each game, the model would NOT be predicting that we win out, rather it would result in an expected record of 9-9. In the same way, our predicted record builds in the probability of each game and calculates the likelihood across all possible outcomes, arriving at the most likely scenario given the current ratings. The way they estimate our final record as predicted by this model would basically be to add up each game's win probability (%/100) and add that sum to our current record.

I hope this brief explanation clarifies the picture somewhat...

This is correct. The cumulative probability of winning the next 13 is much less than the individual game probabilities. To simplify it further....let's say the Bills were favored in the next 2 games, each with a 51% chance of winning....the chances of winning both games games is 26% or about 1 in 4. Even a couple of games where we have a good chance to win (71%) ...winning both is just 50-50. So the chances of the Bills winning the next 13 games is about 2%....though I am not sure why we are focusing on 13 since Timmerman in his own table shows us being favored in the next 15 games. Our chances of winning the next 15 games is 1.4%. The chances of the Bills winning the next 10 games (best 2 loss Bills start in history) is 6.8%

I have no problem with the methodology that Timmerman/Sagarin use. I differ mainly on the point spread/prob %. At the high end ...say around 12 pts ...I show a 96% chance of winning vs T/S at about 85%....on a lower spread ...about 7.5 pts I am at 83%...T/S at 75%....our numbers come together at around 3.5 pts (about 62%) under 3.5 pts I am actually a littlle more conservative (lower % than T/S).

Finally....these are just numbers ....you've got to play the game......lest anyone forget the dreaded RI game

Overall .....good stuff.....Let's just beat Vandy.

[kshoe]

Wiz, it's not timmerman's model. That was just a joke.

He does reference the site from time to time in his articles.

[The Wiz]

Wiz, it's not timmerman's model. That was just a joke.

He does reference the site from time to time in his articles.

I would have been shocked if Timmerman had made any of the calculations

[Old guy]

But that is not the point each individual game will be played and the probability to win the particular game will not be readily determinable until prior to the game. So the probability to win 15 games in a row may be 1.5% at this time, but taken individually we may indeed have a much greater probability to win the whole spread of 15 games than what is determinable at this point in time. Probability changes with time and with results.

[The Wiz]

But that is not the point each individual game will be played and the probability to win the particular game will not be readily determinable until prior to the game. So the probability to win 15 games in a row may be 1.5% at this time, but taken individually we may indeed have a much greater probability to win the whole spread of 15 games than what is determinable at this point in time. Probability changes with time and with results.

My take on this thread is that it is about collective probability....probability of multiple consecutive events happening

Sagarin's chart in the OP shows the entire rest of the season and all the possibilities.

The chances of rolling a 7 are 1 in 6....the chances of rolling 3 consecutive 7's is 1 out 252.....The collective probability is 4 tenths of 1 percent.....yet each roll is always 1 of 6 (16.67%)

[Zink]

But that is not the point each individual game will be played and the probability to win the particular game will not be readily determinable until prior to the game. So the probability to win 15 games in a row may be 1.5% at this time, but taken individually we may indeed have a much greater probability to win the whole spread of 15 games than what is determinable at this point in time. Probability changes with time and with results.

This is somewhat true, and is not what I took your prior post to mean. After each game becomes a 1 or a 0 (for win probability), of course we will have a better estimate of our final win total. Also, each team's rating will change over time as a result of their performance, changing each game's predicted outcome somewhat and hopefully becoming more accurate. But the model never actually predicted that we will win our next 13 and then lose the next however many, hence the confusion...

If we were to win the next 12, winning the 13th game would simply be whatever the win probability for that specific game, yes. BUT the reason we calculate the cumulative probability is because it should give a better view towards expected performance without assuming knowledge of events that are being predicted by the model.

Edit: The Wiz said this in a more straightforward way above - thanks! (Although you can't roll a 7 with a single die)

[moytoy12]

I'm no geologist, but I can tell you we are going undefeated the rest of the year.

[Old guy]

I am really quite happy about this discussion, thank you Wiz, thank you Zink.

[almaman]

me happy too. Great reference tool to watch all season. Our 2 losses suck as they may are looking better each week. Both undefeated and top ten.

[Old guy]

Thinking about the issues discussed here, I thought I would give an example as to how experience (ie games played and won, or in this particular example, mortality) affects the long term outcomes expected. From my area of expertise, the expected life span of a newborn US male is 76 years (2011 stats). So, lets move forward in time, at age 65, non smoking males in the US have an expected survival of 22.6 years. That is once you, a male non smoker, reach age 65 your expected survival has increased to 87.6 years, or 11.6 years over the expected survival of a newborn male. But that is not the whole story because at age 65, within the group of US surviving males (non smoking) there are a number that have a significant number of chronic degenerative diseases, and others that have few or no chronic degenerative diseases. Of course, barring accidental death (which occurs at all ages but which is NOT a significant cause of death over age 65 and over), those people with significant chronic degenerative disorders will survive less than those without the chronic degenerative disorders. Ultimately the longest survival within that group of 65 year old non smoking males will belong to those that have NO chronic degenerative diseases at age 65. These will live long lives indeed, with some of them becoming centenarians. The centenarians will, of course, have exceeded the life expectancy at birth by 24 or more years.

So, let's look at basketball now. All the teams have their own statistics and different styles of play and player talent. At this time we can calculate, as we have done, the cumulative probability, based upon current statistics and experience, for SLU to win the next 15 consecutive games. This is not a highly probable event at this time. However, SLU is very strong and has a very great potential, not fully shown so far because of the quality of most opponents we have been playing against so far, in the way they play and are organized. As we play in A10, our capability to win games of increasing difficulty will be demonstrated. Every time we win a game, we have a situation similar to the example I gave before. As people in the 65 year old die the survivors have a longer life expectancy. In basketball as teams win against other teams of increasing difficulty the winning team has gained something that will help them continue winning. So, the estimate of the cumulative probability of a streak of 15 consecutive wins becomes progressively better every time we win a game in the series. Having only two losses against teams that stand undefeated at this time, only 8 teams are still undefeated in the NCAAB, is not bad at all. Considering how close we came to beat WSU is even better. I firmly believe we can win every single game in conference play, and in the A10 tournaments, barring injury of course.

And for all of you, Wiz, MB, Zink, making meaningful contributions to clarify the situation as it stands now, I have nothing to say except thank you, I really appreciate and enjoy your efforts. I particularly like the classification of teams as A+, A, A-, etc... I think this is very meaningful and quite an interesting piece of data to keep in mind.

[Zink]

Thinking about the issues discussed here, I thought I would give an example as to how experience (ie games played and won, or in this particular example, mortality) affects the long term outcomes expected. From my area of expertise, the expected life span of a newborn US male is 76 years (2011 stats). So, lets move forward in time, at age 65, non smoking males in the US have an expected survival of 22.6 years. That is once you, a male non smoker, reach age 65 your expected survival has increased to 87.6 years, or 11.6 years over the expected survival of a newborn male. But that is not the whole story because at age 65, within the group of US surviving males (non smoking) there are a number that have a significant number of chronic degenerative diseases, and others that have few or no chronic degenerative diseases. Of course, barring accidental death (which occurs at all ages but which is NOT a significant cause of death over age 65 and over), those people with significant chronic degenerative disorders will survive less than those without the chronic degenerative disorders. Ultimately the longest survival within that group of 65 year old non smoking males will belong to those that have NO chronic degenerative diseases at age 65. These will live long lives indeed, with some of them becoming centenarians. The centenarians will, of course, have exceeded the life expectancy at birth by 24 or more years.

So, let's look at basketball now. All the teams have their own statistics and different styles of play and player talent. At this time we can calculate, as we have done, the cumulative probability, based upon current statistics and experience, for SLU to win the next 15 consecutive games. This is not a highly probable event at this time. However, SLU is very strong and has a very great potential, not fully shown so far because of the quality of most opponents we have been playing against so far, in the way they play and are organized. As we play in A10, our capability to win games of increasing difficulty will be demonstrated. Every time we win a game, we have a situation similar to the example I gave before. As people in the 65 year old die the survivors have a longer life expectancy. In basketball as teams win against other teams of increasing difficulty the winning team has gained something that will help them continue winning. So, the estimate of the cumulative probability of a streak of 15 consecutive wins becomes progressively better every time we win a game in the series. Having only two losses against teams that stand undefeated at this time, only 8 teams are still undefeated in the NCAAB, is not bad at all. Considering how close we came to beat WSU is even better. I firmly believe we can win every single game in conference play, and in the A10 tournaments, barring injury of course.

And for all of you, Wiz, MB, Zink, making meaningful contributions to clarify the situation as it stands now, I have nothing to say except thank you, I really appreciate and enjoy your efforts. I particularly like the classification of teams as A+, A, A-, etc... I think this is very meaningful and quite an interesting piece of data to keep in mind.

Yeah, no problem! This is what forums are for, no?

And this example makes perfect sense. I too think we have a shot at making a long run, as our conference schedule is a tad backloaded. Hopefully we are healthy, non-smoking 65 year olds by then

[dlarry]

I would have been shocked if Timmerman had made any of the calculations

Maybe not but I bet he scrolled down the page and sorted the columns by himself. :-)

[Pistol]

Wichita State is in a tight one today, up 45-42 against Davidson with 15:59 to play. I'd expect the Shockers to close it out against a weak (4-8) Davidson team, but it's one to keep an eye on for the next hour.

[TheChosenOne]

Wichita State got the win 81-70.

[AlumniFan]

Wichita State got the win 81-70.

They are definitely good this year. That said, we definitely should have beaten them. I just hate that we lost that game.