What's new
By:
USCHO Fan Forum

This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

  • The USCHO Fan Forum has migrated to a new plaform, xenForo. Most of the function of the forum should work in familiar ways. Please note that you can switch between light and dark modes by clicking on the gear icon in the upper right of the main menu bar. We are hoping that this new platform will prove to be faster and more reliable. Please feel free to explore its features.

Dramatic Changes To *This Season's* NCAA Selection Criteria?

Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

EastFan1 said:
So did any team make a jump up or down based on the new system?
Click to expand...
There were a few jumps just below the Pairwise bubble -- no changes to the tournament field just yet, although North Dakota picked up a lot of ground on Princeton for 7th. They were 0.0065 out of a playoff spot before, now they are just 0.0015 out. That's small enough where North Dakota could theoretically flip with Princeton even without playing a game.

Here are the changes, as it stands today, from the old PWR to the new PWR:

BSU: Up 1 to 9th
Colgate: Down 1 to 10th

BU: Up 2 to 12th
SLU: Down 1 to 13th
Cornell: Down 1 to 14th

SCSU: Up 1 to 15th
Mercyhurst: Down 1 to 16th

Everyone else stays the same although like NoDak some of them are closer to or farther from their nearest competition than they were under the old system.
 
Last edited:
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

robertearle said:
Bemidji, currently at #9 in adjusted RPI and #8 in unadjusted RPI
Click to expand...
Just so you're aware, "unadjusted RPI" is just a raw calculation of the Win%/OW%/OOW% without removing bad wins. I just left it in there informationally.

The QWB is calculated on the actual RPI, adjusted for bad wins, before adding a QWB. I'll change that field in the table to be the RPI that QWB is calculated off. wwhyte pointed out that it's a bit confusing as well.

EDIT: All set. The "Unadj. RPI" is now the original RPI without any QWB.
 
Last edited:
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

TonyTheTiger20 said:
You're just missing the last step which is to divide the total bonus earned by total number of games played

(Note that there's no home/road weighting in women's)
Click to expand...

Thanks. But that raises a couple more questions.

So I'm guessing you used the 0.18 number I came up with, divide by 24 games played equals 0.0075. That got rounded up to 0.008?

First, isn't that a pretty 'significant' amount of rounding error being introduced, at least in this one example?

Second, 0.18 implies that you used the 'adjusted' rank for the Bemidji component of the summed number. Should you not be using the 'unadjusted'? I'm contemplating the possibility of two (or three) teams at #11, 12,13 who have some sort of 'interrelation' to their adjusted ranks; team A at #13 gets their 'bonus' added, and it moves them to #12, which drops team B - a team A beat - from #12 to #13, which changes the bonuses team A earned. which triggers a recalculation, which decreases the team A bonus, which drops them back to #13 and team B back up to #12, which triggers a recalculation... ad infinitum. Your program "thrashes". If you (they) use the unadjusted RPI to base the bonuses, and therefore no 'recalculation', no such potential problem.

Thanks.


EDIT - letters crossing in the mail...

So 'adjusted' in your table is just 'adjusted for bad wins'? Or is it adjusted to 'bad wins AND QWB'? If the former, shouldn't you also have a 'final all-adjustments included' column that is the final rank?
 
Last edited:
robertearle said:
Thanks. But that raises a couple more questions.

So I'm guessing you used the 0.18 number I came up with, divide by 24 games played equals 0.0075. That got rounded up to 0.008?

First, isn't that a pretty 'significant' amount of rounding error being introduced, at least in this one example?
Click to expand...
There's no rounding -- I just limited the column to go only 4 decimals for space purposes but Excel still calculates using all the hidden decimals. I changed it to display a fifth decimal just now for the heck of it.

Second, 0.18 implies that you used the 'adjusted' rank for the Bemidji component of the summed number. Should you not be using the 'unadjusted'? I'm contemplating the possibility of two (or three) teams at #11, 12,13 who have some sort of 'interrelation' to their adjusted ranks; team A at #13 gets their 'bonus' added, and it moves them to #12, which drops team B - a team A beat - from #12 to #13, which changes the bonuses team A earned. which triggers a recalculation, which decreases the team A bonus, which drops them back to #13 and team B back up to #12, which triggers a recalculation... ad infinitum. Your program "thrashes". If you (they) use the unadjusted RPI to base the bonuses, and therefore no 'recalculation', no such potential problem.

Thanks.
Click to expand...

The QWB is calculated based on each teams ranking in the "original" RPI (as it was being calculated last week) without any quality wins bonus. QWB is not recalculated based on any changes in ranking from adding if in.

There's some confusion, I think, on what I was calling "unadjusted RPI" in the spreadsheet. That was originally a teams RPI without any changes and without removing "bad wins." Having that was less informative than putting the original RPI prior to adding QWB so I changed it, and it now reflects the originals QWB-less RPI, the ranking of which is from where QWB was calculated.
 
EDIT - letters crossing in the mail...
Click to expand...
HA yup

So 'adjusted' in your table is just 'adjusted for bad wins'? Or is it adjusted to 'bad wins AND QWB'? If the former, shouldn't you also have a 'final all-adjustments included' column that is the final rank?
Click to expand...
As it stands right now I have two columns:

"RPI" is final, all adjustments, bad wins removed, added QWB
"Unadj. RPI" is regular RPI, also has bad wins removed, but without QWB -- I.e., the RPI column minus QWB
 
Last edited:
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

TonyTheTiger20 said:
There's no rounding -- I just limited the column to go only 4 decimals for space purposes but Excel still calculates using all the hidden decimals. I changed it to display a fifth decimal just now for the heck of it.



The QWB is calculated based on each teams ranking in the "original" RPI (as it was being calculated last week) without any quality wins bonus. QWB is not recalculated based on any changes in ranking from adding if in.

There's some confusion, I think, on what I was calling "unadjusted RPI" in the spreadsheet. That was originally a teams RPI without any changes and without removing "bad wins." Having that was less informative than putting the original RPI prior to adding QWB so I changed it, and it now reflects the originals QWB-less RPI, the ranking of which is from where QWB was calculated.
Click to expand...

Thanks. So just to beat it to death... :-)

I still don't quite see how you get a value for Wisconsin of (currently listed as) 0.0802. If the Bemidji component is 4 x 0.02, the final QWB should be 0.0075; if the Bemidji component is 4 x 0.025, then the final QWB should be 0.0083333... Neither of which rounds to 0.0802. So...??

Sorry if I'm being a pain.

Edit more crossing...

The values in the table right now imply that the actual calculation of 'final RPI' used the 0.0083333; it is just the 0.0802 number that looks funny, but isn't used for anything(?).

Edit #2 (Except maybe the final answer should be .7009 instead of .7010 ?)

Edit #3 (Except if the 'base' RPI has a fragment beyond four decimal places that adds to the end of the repeating 3333 which makes the correct final round .7010. Now I'm just babbling...)
 
Last edited:
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

robertearle said:
Thanks. So just to beat it to death... :-) ... Sorry if I'm being a pain.
Click to expand...
Not at all, I live for this stuff! haha

I still don't quite see how you get a value for Wisconsin of (currently listed as) 0.0802. If the Bemidji component is 4 x 0.02, the final QWB should be 0.0075; if the Bemidji component is 4 x 0.025, then the final QWB should be 0.0083333... Neither of which rounds to 0.0802. So...??
Click to expand...
Individual team tabs are hidden on the online sheet but I can pull them up on my desktop version (it actually allows someone to download the online sheet to desktop where one can unhide all the tabs if so inclined).

For Wisconsin, they are getting credit for the following QWB results:

12/12 tie against North Dakota (a tie is 1/2 a win -- this could be where you're getting a different number) -- for 0.0125
1/15 win against BSU for 0.020
1/16 win against BSU for 0.020
10/23 win against BSU for 0.020
10/24 win against BSU for 0.020
12/4 win against UM for 0.050
12/5 win against UM for 0.050

Sum of all those is 0.1925
Divided by 24 games played = 0.0080208333...

Sheet is displaying 5 decimals so 0.00802

The values in the table right now imply that the actual calculation of 'final RPI' used the 0.0083333; it is just the 0.0802 number that looks funny, but isn't used for anything(?). (Except maybe the final answer should be .7009 instead of .7010 ?)
Click to expand...
Basically the last part -- it's all just how Excel is displaying values, it is rounding to 4 digits in the RPI column for display purposes but still using all of the decimals to calculate.

UW's unadj. RPI is 0.692947915
UW's QWB is .0080208333

Add them together and UW's final RPI is 0.7009687

Excel is displaying 4 decimals so it rounds to 0.7010

:cool:
 
Last edited:
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

TonyTheTiger20 said:
For Wisconsin, they are getting credit for the following QWB results:

12/12 tie against North Dakota (a tie is 1/2 a win -- this could be where you're getting a different number) -- for 0.0125
Click to expand...

Yes, I was entirely ignoring the 'Quality Tie Bonus' component. How silly of me! :o
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

Grant, I'm also a little wary of 'double rounding' type problems.

A couple years ago, I was looking at the 'raw numbers' in a political poll published by a well-know (in Wisconsin, anyway) pollster. And the response for a particular question asked worked out to (let's say) 46.47%, which he published as "46.5%". As is OK. But then the media took that 46.5% and rounded it up to "47%" And in doing so introduced a real error. Because they were then reporting a number 'double rounded' such that the result was being changed to the full integer further from the actual value; the reported value of "47%" was 0.53 away from the actual value, further than a reported value of "46%" would have been away from the actual value (0.47 away). Because of double-rounding. It actually took three or four emails to the pollster to make them realize the problem (and there wasn't much they could do going forward, other than try to warn the media not to round their numbers a second time.)

Of necessity, your spreadsheet software is rounding (or truncating) the infinite ...3333... on the Wisconsin bonus value. I assume it could also be rounding/truncating the last digits of the 'unadjusted RPI'. And then the printout is rounding the result for the printout. Now, the number of 'significant digits' on the printout are different enough such that there likely isn't any potential 'double rounding' problem, etc. But you might want to give it some thought, in case of some really 'degenerate' case of an almost tie, etc.
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

Can I tell my students that this is why math is important to life? :P
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

robertearle said:
Of necessity, your spreadsheet software is rounding (or truncating) the infinite ...3333... on the Wisconsin bonus value. I assume it could also be rounding/truncating the last digits of the 'unadjusted RPI'. And then the printout is rounding the result for the printout. Now, the number of 'significant digits' on the printout are different enough such that there likely isn't any potential 'double rounding' problem, etc. But you might want to give it some thought, in case of some really 'degenerate' case of an almost tie, etc.
Click to expand...
I hear what you're saying -- but it's not truncating or rounding anything. All it's doing is changing what is displayed for the viewer. Any calculations it does behind the scenes is not truncating or rounding any of the values while doing calculation.

As an example let's say it displays a value of 0.7010 to the viewer, but the actual value is 0.7009567098347698 or whatever. Even if I reference the cell that shows "0.7010" to make another calculation, Excel will still use the 0.7009567098347698 value stored behind the scenes, not the 0.7010.
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

gojackets said:
Can I tell my students that this is why math is important to life? :P
Click to expand...
YES! Because it IS! Isn't math AWESOME?!

I knew I was going to be a math major since I was in 2nd grade; isn't that the nerdiest thing you've ever heard? I love it.
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

TonyTheTiger20 said:
I knew I was going to be a math major since I was in 2nd grade; isn't that the nerdiest thing you've ever heard? I love it.
Click to expand...

"I love long division" -- actual quote from Grant age 27.

PS: I love long division too.
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

TonyTheTiger20 said:
YES! Because it IS! Isn't math AWESOME?!

I knew I was going to be a math major since I was in 2nd grade; isn't that the nerdiest thing you've ever heard? I love it.
Click to expand...

I was going to comment on the massive nerdiness of this thread. I often tell my students how when I was in high school I would draw oddly shaped 3 dimensional objects and then find the surface area and volume of them for fun!
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

gojackets said:
I often tell my students how when I was in high school I would draw oddly shaped 3 dimensional objects and then find the surface area and volume of them for fun!
Click to expand...
Oh that's fantastic... sometimes I would take two large numbers (like 10+ digits) and multiple them together by hand to pass the time. Or (as wwhyte mentioned) I'd do a big piece of long division.

Those were the days.
 
TonyTheTiger20 said:
Oh that's fantastic... sometimes I would take two large numbers (like 10+ digits) and multiple them together by hand to pass the time. Or (as wwhyte mentioned) I'd do a big piece of long division.

Those were the days.
Click to expand...

I need an actual data set to play with. I guess that's why I was a stats major rather than pure math.
 
Eeyore said:
I need an actual data set to play with. I guess that's why I was a stats major rather than pure math.
Click to expand...
I have both a math degree (Bachelor's) and stats degree (Master's). Working with actual numbers is why I made the switch from one to the other.
 
Re: Dramatic Changes To *This Season's* NCAA Selection Criteria?

I'm here to tell you all one thing right now. MATH and all related subjects SUCK. Have a nice day. :)
 
You must log in or register to reply here.
Back
Top