Baseball Rankings - 3/3

[campo118]

Edit: read that wrong... I am assuming the SWAC lost half of the 48 games to other divisions?

[GoneGolfin]

Edit: read that wrong... I am assuming the SWAC lost half of the 48 games to other divisions?

Correct.

Brian

[jumboragncajun]

I do not think anyone has these calculations down perfectly yet. Warren has one set of numbers, Boyd has another, and mine differ slightly from both of them. I did peek under the covers of Warren's data (not possible with Boyd) and I think he may be calculating SOS incorrectly ... in particular it may be OWP (but also possibly OOWP). I sent him a note to discuss ... will see if he responds.

In any event, for those interested in the details as to how OWP is calculated for a given team ...

- Take the Div. I W/L record for each of your opponents and remove the results of your games against that team.
- Calculate the above winning percentage for each opponent to four decimals of precision.
- Weight that winning percentage by the number of games played against each respective opponent.

DO NOT sum the weighted W/L records of each of your opponents (non-Cajun games) and then calculate the percentage on that summed W/L record. This will yield a different result. This *seems* to be what Warren Nolan is doing.

Example:
Southern Mississippi is 6-6 overall. However, they are 6-3 in games not involving the Cajuns. Hence, for OWP purposes we use .6667 * 3 (weighted by three games played).

- Then take all of the calculated weighted percentages and add them. Divide that sum by the total number of games played. You then have your OWP.

Note: OWP calculations do not use the home/road weightings that WP calculations use.

Example using Cajun opponents ...

Eastern Illinois (2-10) ... 1-7 in non-Cajun games = .1250 * 4 (games) = .5000
Northwestern State (6-4) ... 6-3 in non-Cajun games = .6667 * 1 (game) = .6667
Southern Mississippi (6-6) ... 6-3 in non-Cajun games = .6667 * 3 (games) = 2.0000
LSU (9-2) ... 9-1 in non-Cajun games = .9000 * 1 (game) = .9000
Alabama (5-5) ... 4-3 in non-Cajun games = .5714 * 3 (games) = 1.7143
--
Total: 5.781/12 games = .4817 (note the above uses rounded figures to four decimal places)

This is obviously different than the summed 49-50 record (.4949) ... and in some cases, the difference can be much more significant.

Thus we have the following RPI calculation for UL ...

WP = .7937
OWP = .4817
OOWP = .6137
SOS = .5257
--
RPI = .5927

On Warren's site, if you look at the SOS page, he calculates an opponents winning percentage of .4949 (49-50) by summing the weighted W/L record of each Cajun opponent and calculating the winning percentage. His SOS comes to .5249. His OOWP calculation is thus .5849.

Brian

Ok ya'll: Sticky this thread, print out this post, tape it to your computer, make a spreadsheet...etc.. If it's March 3, and we are this deep in the numbers, I predict no less than 100 times GoneGolfin will be asked how to calculate this, or how x game or y game effected our rpi.

Thanks Brian

[GoneGolfin]

I ran my calculations tonight and was still coming up with something slightly different than Warren Nolan. After closer examination ...

- WN has the 3/14 result from Arkansas-Pine Bluff and Grambling State reversed. Grambling State lost that game 8-5. Thus Grambling State should have a record of 6-17, not 7-16. Arkansas-Pine Bluff is 7-13 vs. 6-14.

- WN is missing two games ... the 3/18 DH pitting SIU-Edwardsville vs. Illinois State. Illinois State won both of those games. Thus SIU-Edwardsville is 9-12, not 9-10. Illinois State is 15-6, not 13-6.

If I implement the above mistakes in my own calculations, my numbers match Warren Nolan's for UL.

Thus, we have the following correct RPI calculation for Louisiana ...

WP = .8917
OWP = .4871
OOWP = .5397
SOS = .5046
--
RPI = .6014

Warren Nolan currently has ...
SOS = .5052
RPI = .6019

The RPI ranking is unchanged at #10 because 9 and 11 are rather separated from #10.

Brian

I do not think anyone has these calculations down perfectly yet. Warren has one set of numbers, Boyd has another, and mine differ slightly from both of them. I did peek under the covers of Warren's data (not possible with Boyd) and I think he may be calculating SOS incorrectly ... in particular it may be OWP (but also possibly OOWP). I sent him a note to discuss ... will see if he responds.

In any event, for those interested in the details as to how OWP is calculated for a given team ...

- Take the Div. I W/L record for each of your opponents and remove the results of your games against that team.
- Calculate the above winning percentage for each opponent to four decimals of precision.
- Weight that winning percentage by the number of games played against each respective opponent.

DO NOT sum the weighted W/L records of each of your opponents (non-Cajun games) and then calculate the percentage on that summed W/L record. This will yield a different result. This *seems* to be what Warren Nolan is doing.

Example:
Southern Mississippi is 6-6 overall. However, they are 6-3 in games not involving the Cajuns. Hence, for OWP purposes we use .6667 * 3 (weighted by three games played).

- Then take all of the calculated weighted percentages and add them. Divide that sum by the total number of games played. You then have your OWP.

Note: OWP calculations do not use the home/road weightings that WP calculations use.

Example using Cajun opponents ...

Eastern Illinois (2-10) ... 1-7 in non-Cajun games = .1250 * 4 (games) = .5000
Northwestern State (6-4) ... 6-3 in non-Cajun games = .6667 * 1 (game) = .6667
Southern Mississippi (6-6) ... 6-3 in non-Cajun games = .6667 * 3 (games) = 2.0000
LSU (9-2) ... 9-1 in non-Cajun games = .9000 * 1 (game) = .9000
Alabama (5-5) ... 4-3 in non-Cajun games = .5714 * 3 (games) = 1.7143
--
Total: 5.781/12 games = .4817 (note the above uses rounded figures to four decimal places)

This is obviously different than the summed 49-50 record (.4949) ... and in some cases, the difference can be much more significant.

Thus we have the following RPI calculation for UL ...

WP = .7937
OWP = .4817
OOWP = .6137
SOS = .5257
--
RPI = .5927

On Warren's site, if you look at the SOS page, he calculates an opponents winning percentage of .4949 (49-50) by summing the weighted W/L record of each Cajun opponent and calculating the winning percentage. His SOS comes to .5249. His OOWP calculation is thus .5849.

Brian

[Turbine]

Nice detective work Brian.

[moorecajun]

Nice detective work Brian.

Is he one of the "True Detectives" everyone has been praising on here?

[ULdude]

Amazed how you deduced exactly where the Opponent's opponents records were off in WN. Pretty awesome.

[GoneGolfin]

I ran my calculations tonight and was still coming up with something slightly different than Warren Nolan. After closer examination ...

- WN has the 3/14 result from Arkansas-Pine Bluff and Grambling State reversed. Grambling State lost that game 8-5. Thus Grambling State should have a record of 6-17, not 7-16. Arkansas-Pine Bluff is 7-13 vs. 6-14.

- WN is missing two games ... the 3/18 DH pitting SIU-Edwardsville vs. Illinois State. Illinois State won both of those games. Thus SIU-Edwardsville is 9-12, not 9-10. Illinois State is 15-6, not 13-6.

If I implement the above mistakes in my own calculations, my numbers match Warren Nolan's for UL.

Thus, we have the following correct RPI calculation for Louisiana ...

WP = .8917
OWP = .4871
OOWP = .5397
SOS = .5046
--
RPI = .6014

Warren Nolan currently has ...
SOS = .5052
RPI = .6019

The RPI ranking is unchanged at #10 because 9 and 11 are rather separated from #10.

Brian

I received a note from Warren Nolan that he has made the above fixes.

Brian

[GoneGolfin]

Just as an addendum to this ...

While the NCAA obviated the existing bonus/penalty structure with the advent of the 2013 RPI (weighting home/road), there is a penalty system that does exist ... that is not accounted for in either Boyd's or Warren's calcs. While results against non-D1 opponents are not included in the Base RPI calculations ... a loss to a non-D1 school will result in a penalty of .0022. Forty Eight (48) such penalties were levied in 2013 for non-D1 losses by D1 teams. A few were from some of the bigger names (Florida Gulf Coast, Texas Tech, Villanova, Northwestern, McNeese State). The SWAC had half of them (24).

Brian

I have also reverse engineered the penalty formula that is still in place for the RPI. I have communicated this to Warren Nolan ... and he confirmed that he is not accounting for these penalties at present. Here is the formula I arrived at ...

A .0022 penalty is assessed for ...

- The first four non-D1 games played
- The sixth non-D1 game that is played
- Each non-D1 game played after six

Additionally ...

- Home loss to a non-D1 opponent is assessed a penalty of .0022
- Neutral site loss to a non-D1 opponent is assessed a penalty of .0017 (mid-point between .0022 and .0011 ... rounded up)
- Road loss to a non-D1 opponent is assessed a penalty of .0011

A tie with a non-D1 opponent is not assessed a penalty.

A good cross section of examples from 2013 ...

- Florida Gulf Coast only played (1) non-D1 opponent in 2013. They lost to Ave Maria at home and were assessed an RPI penalty of .0022.

- The same happened to Texas Tech (lost to Lubbock Christian at home) ... penalty of .0022.

- Dartmouth played (5) non-D1 games ... and lost one of them. The total penalty was .0039. This is a .0017 penalty for the neutral site loss to Slippery Rock ... and .0022 for playing four or more (but less than six) non-D1 games. The NCAA logged this as (2) penalties.

- Maine played (4) non-D1 games and was assessed a penalty of .0022.

- UMES played (1) non-D1 game ... a road loss to Wilmington (DE) ... resulting in a penalty of .0011.

- Prairie View had (2) losses vs. non-D1 opponents (one home, one road) and played (10) games vs. non-D1 opponents. They were assessed (8) penalties for a total of .0165. So, this seems to be ... (1) penalty for the first four non-D1 games ... another for the 5th and 6th ... and then a penalty for each one after the 6th. That would be (6) penalties total for non-D1 games played (.0132) plus (2) penalties for the losses (.0022 + .0011 = .0033). This gives us our .0165 total penalty.

- North Dakota played (8) games vs. non-D1 competition (all neutral site games), lost twice, and tied once. They were assessed (6) penalties for a total of .0122. It does not seem that a penalty was assessed for the tie (only losses are penalized). Thus we have ... .0088 or (4 * .0022 = .0088) for the number of non-D1 games played (4th, 6th, 7th, 8th) ... plus (2 * .0017 = .0034) for the neutral site losses ... for a total of .0122.

Brian

[moorecajun]

Brian, you amaze me with every single post! Sometimes I don't have time to read them as thoroughly as I should and I definitely am smarter when I finish, even if I only get half of your post!

Side note: I wonder who Ave Maria has on their team?



Keep up the good ... no, GREAT work ... we all appreciate it ... and I feel like I owe you some money.

[BeauCajun]

Side note: I wonder who Ave Maria has on their team?

Probably Jesus and the 12 apostles. I heard they picked up David as the closer for the FGC series. It's my understanding he can really sling the rock.

[GoneGolfin]

Brian, you amaze me with every single post! Sometimes I don't have time to read them as thoroughly as I should and I definitely am smarter when I finish, even if I only get half of your post!

Side note: I wonder who Ave Maria has on their team?



Keep up the good ... no, GREAT work ... we all appreciate it ... and I feel like I owe you some money.

Thank you moorecajun. That is very nice of you to say.

Brian