John t whelan ranking simulator

[Numbers]

Re: John t whelan ranking simulator

While there were a lot of little details to get right, I think the only one where it's not fairly obvious what the NCAA intended is whether the removal of an adverse impact win from the index affects the QWB. Based on the language of the memo, my guess was that the QWB is a separate step from the calculation of the index, so you would still get it for wins excluded from the RPI calculation. I wouldn't be least bit surprised to turn out to be wrong, though.

Thanks for joining us Jim. I will copy the relevant part of the memo:
Once the total bonus points have been calculated they are divided by the total weighting of all games played per the RPI calculation (with road wins/home losses weighted with a factor of 1.2 and home wins/road losses weighted with a factor of 0.8). The resulting Quality Wins Bonus (QWB) is added to the original RPI to obtain the final RPI for each team. This is the value to be used in comparing any two teams, along with the other selection criteria.

Now, there are 2 parts:
1 question is: would you still get QWB from a removed game? This seems a bit irrelevant, seeing that any win against a team with a high enough RPI to get QWB pts won't be removed, so

moving on to 2) It seems you add all the QWB pts, and then divide by a #ofgames calculation. Obviously, the home/road weighting is in effect with this calculation. The part not spelled out is: Do you count games that were removed from the RPI calculation because of neg effect even though you won?

As an example, Minny has a win against Mercyhurst which is removed from the sched when calculating RPI. Do you count 0.8 for that game when deciding what number to divide the QWB by?

This is the question, and the answer does not seem to be explicit.

[JimDahl]

Re: John t whelan ranking simulator

While there were a lot of little details to get right, I think the only one where it's not fairly obvious what the NCAA intended is whether the removal of an adverse impact win from the index affects the QWB. Based on the language of the memo, my guess was that the QWB is a separate step from the calculation of the index, so you would still get it for wins excluded from the RPI calculation. I wouldn't be least bit surprised to turn out to be wrong, though.

[Numbers]

Re: John t whelan ranking simulator

What sort of strength of schedule would you like to see? I thought about trying to calculate KRACH, but didn't have a formula. Not to mention, those SoS calculations are effectively meaningless, unless you use it as a predicting tool to see who is likely to beat a certain team, something that the PWR is, in effect, designed to do.

Sorry, not clear in my post. I was thinking about the committee and their definition. I obviously spent awhile looking at the press release, trying to figure out how to calculate, and it wasn't perfectly clear.

My remark about SoS was from the committee's perspective. They know they need a SoS factor somewhere, because the schedules aren't balanced. Thus the birth of RatingsPI. And, the idea of coupling win%, oppwin%, and oppoppwin% seems 'right' and it's fairly easy to calc. Then, with removing negative impact games, and the QWB, and the home/road factor (and I see the justification for all of them - it's good), the calc just gets more and more complex. Witness a guy like Patman who is a programmer, and he claims the minutiae gives him indigestion.

It was just a comment on my part.

By the way, I can send you a spreadsheet for KRACH this week if you want. I made one a couple years ago for NHL, and for NCAA it would just need expanding. Since I am not a programmer (46 years old, wanted to do what Whelan does, but ended up in the ministry, so I didn't keep up with all the new languages), I just do every thing the tedious way - Excel will calculate KRACH - you just have to do the iterations manually.

[FlagDUDE08]

Re: John t whelan ranking simulator

Oh. Well, that works for me, then. Likely the Mercyhurst game is removed, so that leaves 8.6 bonus Pts and 8.6 games, so that is close enough to a rounding error to work for me.

But, like I said, all those levels make a very complicated calculation. And, things are not quite so perfectly spelled out - the press release doesn't specify whether or not to leave out negative impact games. Again, not nitpicking, but noticing that this calculation of "RPI" with all it's new definitions, is conceptually easy: We need to include strength of schedule, too, so let's do that. But, then it gets complicated with all the tweaking....

What sort of strength of schedule would you like to see? I thought about trying to calculate KRACH, but didn't have a formula. Not to mention, those SoS calculations are effectively meaningless, unless you use it as a predicting tool to see who is likely to beat a certain team, something that the PWR is, in effect, designed to do.

[Numbers]

Re: John t whelan ranking simulator

The .850 vs. .851 sounds like rounding vs. truncation in terms of display. The application stores the full decimal, but only outputs the first three places; I assume Java truncates instead of rounds. Also, when you are calculating the Quality Wins Bonus, are you taking into account the games that were removed in negative effect on the RatingsPI? One new feature with the 0.4 version is it left in a debug statement that when you select a team, it will tell you what games have been removed in the command line. Obviously you'll need to run from the command line instead of the .exe in order to get this.

Oh. Well, that works for me, then. Likely the Mercyhurst game is removed, so that leaves 8.6 bonus Pts and 8.6 games, so that is close enough to a rounding error to work for me.

But, like I said, all those levels make a very complicated calculation. And, things are not quite so perfectly spelled out - the press release doesn't specify whether or not to leave out negative impact games. Again, not nitpicking, but noticing that this calculation of "RPI" with all it's new definitions, is conceptually easy: We need to include strength of schedule, too, so let's do that. But, then it gets complicated with all the tweaking....

[FlagDUDE08]

Re: John t whelan ranking simulator

Thanks. I wondered about that. However, it still doesn't match what your calculator says. It would be a win% of .851 where your calculator gets .850 (I know that's miniscule, but it still indicates to me something different happening- maybe the win% on your program is without the weighting). And, the bonus doesn't come out right, either. I calculate a QWB of .915, whereas your calculator lists it as .98.

Not picking, just trying to see what is happening.

Thanks, FlagDude.

One thing that comes to mind - this is a very complicated calculation, and it depends on interpretation in some ways?

The .850 vs. .851 sounds like rounding vs. truncation in terms of display. The application stores the full decimal, but only outputs the first three places; I assume Java truncates instead of rounds. Also, when you are calculating the Quality Wins Bonus, are you taking into account the games that were removed in negative effect on the RatingsPI? One new feature with the 0.4 version is it left in a debug statement that when you select a team, it will tell you what games have been removed in the command line. Obviously you'll need to run from the command line instead of the .exe in order to get this.

[Numbers]

Re: John t whelan ranking simulator

I think you messed up the winning percentage potential. You take the total possible points to be had given the circumstance, and the weighting still follows the original paradigm that you described. The potential is 5 home wins (4.0) + 1 tie (1.0; doesn't matter where it was for a tie) + 3 away wins (3.6) + 1 road loss (0.8) = 9.4, not the 9.6 that you talked about previously. That should help your numbers a bit.

Thanks. I wondered about that. However, it still doesn't match what your calculator says. It would be a win% of .851 where your calculator gets .850 (I know that's miniscule, but it still indicates to me something different happening- maybe the win% on your program is without the weighting). And, the bonus doesn't come out right, either. I calculate a QWB of .915, whereas your calculator lists it as .98.

Not picking, just trying to see what is happening.

Thanks, FlagDude.

One thing that comes to mind - this is a very complicated calculation, and it depends on interpretation in some ways?

[FlagDUDE08]

Re: John t whelan ranking simulator

Patman,

Here is a link: http://www.uscho.com/2013/09/20/road...tion-criteria/

Basically, if I have it right:

1) Compute win% for all teams, and save this.

2) Re-compute, using 1.2/0.8 weighting scale for RoadWins&HomeLosses/HomeWins&RoadLosses (It says that ties are 1/2 win and 1/2 loss, so I assume that means that, for your win%, you get +0.4/1 for a home tie. Neutral Site Games get 1.0 weighting.). Example for Minnesota: # wins = 5*.8 (home wins), 1*.4 (home tie), 3 * 1.2 (road wins) =8.0. # games = 6*.8 (home) + 4*1.2 (road) = 9.6. Win% = 8/9.6 = .833

3) Compute RPI for all teams using 25% win% from #2 above, and 21%OppWin% and 54%OppOppWin% from #1 above.

4) Check to see if any games should be left out.

5) Build a list of the top-20 teams according to the results after #4.

6) Add Quality Wins Bonus according to the scale provided. Use the Home/Road system here as well. For example, Minny currently has 2 QWB results against BC, both at home. So, they count .8. It's a win and a tie, so (1.5*.8)xthe factor for BC= 1.2*3.5 or 4.2. Minny has a win at NoDame, so that counts 1.2*2.5 or 3.0. Also, a win at home against UNH = 0.8*1.25 or 1.00. And, 2 wins at home against Mankato = 0.8*2*.25=0.4. This gives a total of 8.6. Now, we divide by games played, and we did that calculation in #2 above and got 9.6. Minnesota's QWB is therefore 8.6/9.6 = .896. (Note that this does not quite agree with Flaggy's totals, so either he or I am doing something wrong).

Which, as you say, is lots of minutiae.

I think you messed up the winning percentage potential. You take the total possible points to be had given the circumstance, and the weighting still follows the original paradigm that you described. The potential is 5 home wins (4.0) + 1 tie (1.0; doesn't matter where it was for a tie) + 3 away wins (3.6) + 1 road loss (0.8) = 9.4, not the 9.6 that you talked about previously. That should help your numbers a bit.

[Numbers]

Re: John t whelan ranking simulator

I hate to be an *** but is it possible to get like a document or spec detailing the new rub. I want to compute it at some point but its the minutiae that causes me indigestion.

Patman,

Here is a link: http://www.uscho.com/2013/09/20/road...tion-criteria/

Basically, if I have it right:

1) Compute win% for all teams, and save this.

2) Re-compute, using 1.2/0.8 weighting scale for RoadWins&HomeLosses/HomeWins&RoadLosses (It says that ties are 1/2 win and 1/2 loss, so I assume that means that, for your win%, you get +0.4/1 for a home tie. Neutral Site Games get 1.0 weighting.). Example for Minnesota: # wins = 5*.8 (home wins), 1*.4 (home tie), 3 * 1.2 (road wins) =8.0. # games = 6*.8 (home) + 4*1.2 (road) = 9.6. Win% = 8/9.6 = .833

3) Compute RPI for all teams using 25% win% from #2 above, and 21%OppWin% and 54%OppOppWin% from #1 above.

4) Check to see if any games should be left out.

5) Build a list of the top-20 teams according to the results after #4.

6) Add Quality Wins Bonus according to the scale provided. Use the Home/Road system here as well. For example, Minny currently has 2 QWB results against BC, both at home. So, they count .8. It's a win and a tie, so (1.5*.8)xthe factor for BC= 1.2*3.5 or 4.2. Minny has a win at NoDame, so that counts 1.2*2.5 or 3.0. Also, a win at home against UNH = 0.8*1.25 or 1.00. And, 2 wins at home against Mankato = 0.8*2*.25=0.4. This gives a total of 8.6. Now, we divide by games played, and we did that calculation in #2 above and got 9.6. Minnesota's QWB is therefore 8.6/9.6 = .896. (Note that this does not quite agree with Flaggy's totals, so either he or I am doing something wrong).

Which, as you say, is lots of minutiae.

[FlagDUDE08]

Re: John t whelan ranking simulator

Flag, I took a look at the results file. It seems easy enough to alter myself. Do I have it right that it works like this:
1st column is school name. All other columns go in groups of 4:Opponent, Location, Date, Result. So, to add a result for an already scheduled game just requires changing the result column from "0" to whichever team won or "Draw". To add a game, like in the conference playoffs or a holiday tourney, just add all 4 columns at the end of the file. Right? Thanks.

I believe that's it, but just be aware that you will have to do it for both teams involved.

[Numbers]

Re: John t whelan ranking simulator

Flag, I took a look at the results file. It seems easy enough to alter myself. Do I have it right that it works like this:
1st column is school name. All other columns go in groups of 4:Opponent, Location, Date, Result. So, to add a result for an already scheduled game just requires changing the result column from "0" to whichever team won or "Draw". To add a game, like in the conference playoffs or a holiday tourney, just add all 4 columns at the end of the file. Right? Thanks.

[FlagDUDE08]

Re: John t whelan ranking simulator

Flaggy,

As one who inputs his own results, is there a way to input several results and then the computation doesn't happen until the end? I spent lots of time last night waiting for the calculation to happen as I inputted each game.

So, what I hope for is:
Button to add results or games
A different button to load the chosen results into the compiler like one that says "Calculate now"

At this point, I only have it individually. Obviously you could edit the text file, but the game is in two places; one for each team. I am going to start to look for more efficient calculation methods, including the suggestion you made, given the time to calculate is starting to get long, mostly because of the formula to remove wins that lower the rating. Thank you for the feedback.

[FlagDUDE08]

Re: John t whelan ranking simulator

I hate to be an *** but is it possible to get like a document or spec detailing the new rub. I want to compute it at some point but its the minutiae that causes me indigestion.

Take a look in the USCHO news archives; they had an article on this back in September.

[Patman]

It's quite alright, and I do appreciate the suggestion. There is a calculator to which I provide writing contribution, and I have been looking for some ideas of what to include; this is a good one. I'll also be looking at a pop-up chart of who bests who in PWR, though screen size is a challenge because all of the teams are now under consideration. The application does list this information individually. I can tell you that Michigan has received 0.53 bonus points.

I hate to be an *** but is it possible to get like a document or spec detailing the new rub. I want to compute it at some point but its the minutiae that causes me indigestion.

[Numbers]

Re: John t whelan ranking simulator

Flaggy,

As one who inputs his own results, is there a way to input several results and then the computation doesn't happen until the end? I spent lots of time last night waiting for the calculation to happen as I inputted each game.

So, what I hope for is:
Button to add results or games
A different button to load the chosen results into the compiler like one that says "Calculate now"