2023 D3 Men's Soccer National Perspective

[PaulNewman]

@stlawus, I get H2H in close cases but I don't recall debate about quality of ranked wins.  Sounds like you're saying Babson could be dinged for having "weaker" ranked wins than a team from another region...and that, if I'm reading you correctly that a team from say Region 5 with say only 2-3 ranked wins (but of especially high quality) could trump a Babson with 5-6 ranked wins. 

Would be great if someone like FW could weigh in.

And btw, I'm pretty sure I would prefer it if you are correct....just hasn't been my understanding over the years.

[stlawus]

@stlawus, I get H2H in close cases but I don't recall debate about quality of ranked wins.  Sounds like you're saying Babson could be dinged for having "weaker" ranked wins than a team from another region...and that, if I'm reading you correctly that a team from say Region 5 with say only 2-3 ranked wins (but of especially high quality) could trump a Babson with 5-6 ranked wins. 

Would be great if someone like FW could weigh in.

And btw, I'm pretty sure I would prefer it if you are correct....just hasn't been my understanding over the years.

That's the type of scenario I'm referring to, but probably not as extreme as Babson's case.  If all 3 wins from a Region 5 team are all extremely high quality, like being wins against the top 2 teams in a very strong region like Region I that would stand a very good chance of beating out Babson if their wins are all against bottom ranked teams. But having twice the amount probably negates it and Babson also has some very high quality results.    I'm more referring to when there's a difference of 1 or 2 wins.   If a team has 1 or 2 fewer than another team on the board, but their wins are all significantly high quality the other team will definitely be dinged.  That of course is if they are relatively even in the other criteria.

[PaulNewman]

@stlawus, I get H2H in close cases but I don't recall debate about quality of ranked wins.  Sounds like you're saying Babson could be dinged for having "weaker" ranked wins than a team from another region...and that, if I'm reading you correctly that a team from say Region 5 with say only 2-3 ranked wins (but of especially high quality) could trump a Babson with 5-6 ranked wins. 

Would be great if someone like FW could weigh in.

And btw, I'm pretty sure I would prefer it if you are correct....just hasn't been my understanding over the years.

That's the type of scenario I'm referring to, but probably not as extreme as Babson's case.  If all 3 wins from a Region 5 team are all extremely high quality, like being wins against the top 2 teams in a very strong region like Region I that would stand a very good chance of beating out Babson if their wins are all against bottom ranked teams. But having twice the amount probably negates it and Babson also has some very high quality results.    I'm more referring to when there's a difference of 1 or 2 wins.   If a team has 1 or two fewer than another team on the board, but their wins are all high quality the other team will definitely be dinged.  That of course is if they are relatively even in the other criteria.

Interesting...

If you are correct (or for that matter even if you're not) just goes to show how difficult it is to absorb and retain all the nuances.  I've been here since 2012 and I'd guess you've been here as long or longer, and after all this time, and all the articles about the process that get posted yearly, here with are with two long-term, high-volume posters who have significantly different understandings.  So definitely understandable when new and relatively new participants can be so surprised and even upset as they go through each step in the process.

I picked Babson because of the four or so ranked wins against teams at .500 and lower who would never have been ranked in any other region.  For me, that would be the stage to ameliorate the big divergences.

[stlawus]

@stlawus, I get H2H in close cases but I don't recall debate about quality of ranked wins.  Sounds like you're saying Babson could be dinged for having "weaker" ranked wins than a team from another region...and that, if I'm reading you correctly that a team from say Region 5 with say only 2-3 ranked wins (but of especially high quality) could trump a Babson with 5-6 ranked wins. 

Would be great if someone like FW could weigh in.

And btw, I'm pretty sure I would prefer it if you are correct....just hasn't been my understanding over the years.

That's the type of scenario I'm referring to, but probably not as extreme as Babson's case.  If all 3 wins from a Region 5 team are all extremely high quality, like being wins against the top 2 teams in a very strong region like Region I that would stand a very good chance of beating out Babson if their wins are all against bottom ranked teams. But having twice the amount probably negates it and Babson also has some very high quality results.    I'm more referring to when there's a difference of 1 or 2 wins.   If a team has 1 or two fewer than another team on the board, but their wins are all high quality the other team will definitely be dinged.  That of course is if they are relatively even in the other criteria.

Interesting...

If you are correct (or for that matter even if you're not) just goes to show how difficult it is to absorb and retain all the nuances.  I've been here since 2012 and I'd guess you've been here as long or longer, and after all this time, and all the articles about the process that get posted yearly, here with are with two long-term, high-volume posters who have significantly different understandings.  So definitely understandable when new and relatively new participants can be so surprised and even upset as they go through each step in the process.

I picked Babson because of the four or so ranked wins against teams at .500 and lower who would never have been ranked in any other region.  For me, that would be the stage to ameliorate the big divergences.

I've been here the same amount of time you have, but I won't pretend like I'm an expert insider or anything.  Committee members are different across sports obviously, but they all largely use the same exact criteria.  The basketball process is a lot more clearer since d3hoops routinely interviews the committee members and gets to hear how they selected teams when it's all said and done.  This scenario plays out a lot in basketball.  In fact, this exact scenario happened with the SLU women's basketball team 2 years ago.  The commitee chairwoman outright said that SLU's quality of wins were not as high quality as Washington and Lee's and thus W&L got the bid despite them all having very similar numbers.

If the committee just used the algorithm then there would be no reason for the committee to exist. The NCAA could just pick the teams with a computer.    It is my understanding that the committee definitely weighs the quality of wins, and this process plays out when ranking teams regionally as well.

Babson has a very gaudy SOS and have a win against Amherst and tie against Tufts.  I highly doubt there is a team with a similar amount of ranked wins that has results against top tier competition like that.  And that only assumes that team has a .601 SOS like Babson, which again I highly doubt.

[Freddyfud]

Just did the math for Albertus Magnus since they were first on the Region 1 data sheet.  From the latest sheet, a record of 9-2-5, in division win loss of .719, and in division SOS of .471.  For in division W/L using .5 for a tie you get 11.5/16 or .719, easy enough.  As for SoS I noticed if you take 2/3 of .719 you get .479, which exceeds the SoS without even adding the cumulative OOWP.  So .5 for a tie cannot be used.  Went ahead and grabbed AM's opponent records:

Opponent	Opp record	Opp Wins	Opp Ties	Opp total games
Springfield	4-6-5	4	5	15
Casleton	4-11-1	4	1	16
New Eng Coll	8-5-4	8	4	17
Colby-Saw	3-10-4	3	4	17
Mitchell	2-10-2	2	2	14
Norwich	6-2-7	6	7	16
Anna Maria	4-9-1	4	1	14
St Joe's CT	11-1-3	11	3	15
Westfield	6-7-1	6	1	14
Regis MA	6-9-0	6	0	15
Lasell	3-8-3	3	3	14
Emman MA	5-10-1	5	1	16
Elms	2-11-3	2	3	16
St Joes ME	10-3-3	10	3	16
JWU RI	10-2-4	10	4	16
River	3-8-5	3	5	16
Totals		87	47	247

By taking different multipliers for ties and based on 2/3 for OWP and 1/3 for OOWP, I calculated:

Tie multiplier	Weighted OWP	Weighted OOWP	SoS
50%	.479	.149	.628
33%	.444	.138	.583
none	.375	.117	.492

And of course I couldn't recalc the .471 on the data sheet.  FAIL.  Could be I misapplied one of the opponent names as I'm not familiar with the schools or abbreviations or my math is just off somewhere.    But it seems to me any factor of a tie is too great for SoS, at least in this one example.

[camosfan]

Way too esoteric for me!

[paclassic89]

Freddy, you might want to contact the NCAA if you think you found an error (which does happen sometimes).  Ties should be worth .5 for win % purposes. The winning % formula is basically this (W + 1/2T)/(W + L + T). The RvR is one of those highly subjective elements of the committee process i'm pretty sure.  Also, people got confused because PN ninja edited his original question to refer to RvR rather than SoS with regards to ties.

[Freddyfud]

Freddy, you might want to contact the NCAA if you think you found an error (which does happen sometimes).  Ties should be worth .5 for win % purposes. The winning % formula is basically this (W + 1/2T)/(W + L + T). The RvR is one of those highly subjective elements of the committee process i'm pretty sure.  Also, people got confused because PN ninja edited his original question to refer to RvR rather than SoS with regards to ties.

Yea I was still curious about SoS though.  And I see that formula works for in division W/L.  Problem was I noticed some teams on the data sheets with high in division W/Ls and lower SoS--to the point 2/3 of the OWP based on a 50% factor for ties is greater than the SoS even without adding the 1/3 for OOWP.   Just so happened AM was one of them and first on the list. Same for Dean further down the Region 1 list.  I checked Appendix D of the most current Pre Championships Manual as referred by Cristan Shirk and couldn't find anything mentioned for ties with respect to SoS. The appendix could just be carry over from the years without ties.  I also checked some of the rule change updates but couldn't find anything. Which made me wonder if the SoS calculation is also carry over even if the in division W/L is not.

[PaulNewman]

@stlawus, I'm guessing the actual divergence in our views is far more marginal than is coming across.

At any rate, no disagreement about Amherst and Tufts with Babson.  Impressive overall resume and I already called them a lock.  The issue I thought was more about how the ranked wins against the other ranked teams in their region are evaluated (who wouldn't have been ranked in any other region).  And you seem to be saying that a cmte can and will downgrade some or all of those other wins in comparison with another team (with otherwise very comparable credentials) team in another region that has the same number of ranked wins.  Now could a cmte especially when torn about who to pick for the last two spots in the field among four to five nearly identical profiles delve another layer deeper?  Sure, I can see that.  But are you saying something well beyond that?

[stlawus]

@stlawus, I'm guessing the actual divergence in our views is far more marginal than is coming across.

At any rate, no disagreement about Amherst and Tufts with Babson.  Impressive overall resume and I already called them a lock.  The issue I thought was more about how the ranked wins against the other ranked teams in their region are evaluated (who wouldn't have been ranked in any other region).  And you seem to be saying that a cmte can and will downgrade some or all of those other wins in comparison with another team (with otherwise very comparable credentials) team in another region that has the same number of ranked wins.  Now could a cmte especially when torn about who to pick for the last two spots in the field among four to five nearly identical profiles delve another layer deeper?  Sure, I can see that.  But are you saying something well beyond that?

Yes, that is what I'm saying.  Basically not all wins are equal, and the committee proceeds accordingly.   I would believe that if there were 2 pool C spots left and the teams had a negligible difference in SOS and Win% but one team had fewer RR wins but all much all significantly higher quality than the other team's, the former would get selected.

[Christan Shirk]

From the 2023 Division III Soccer Pre-Championships Manual, the RvR criterion is as follows:

• Results versus ranked Division III teams as established by the final ranking and the ranking preceding the final ranking.
  

NOTE: It does not say W-L-T records versus ranked teams. It does not say winning percentage versus ranked teams. It merely says results versus ranked teams. That gives the Regional Advisory Committees (RACs) and the men's soccer committee a lot of latitude and room for subjectivity.

The data sheets provide the W-L-T record versus ranked teams as a piece of data which is part of this criterion, but not the whole of it. What more could the data sheets provide?  How would you concisely capture which wins came against teams highly ranked versus lowly ranked?  You can't. Just like the data sheets can't provide information on head-to-head competition and results versus common Division III opponents.  The data sheets only capture the quantifiable/numerical criteria.

I'm pretty sure that committee members for basketball and for soccer have confirmed that they do consider things like whether results/wins came against highly ranked teams or lowly ranked teams.  The W-L-T record is a starting point, but the committees can and apparently do dig deeper than that as needed to compare teams.

I think it would follow that, if the committee can and does weigh some results/wins greater based on opponents' ranking within the same region, than the committee could also, and likely does at times, consider results/wins versus teams ranked in one region as having more weight than results/wins versus teams ranked in another region.  I don't recall this specifically being confirmed by a committee member, but the criterion, written so general and broad as it is, wouldn't prohibit it.

[Christan Shirk]

At present, the NCAA considers ties to be one-half win, one-half loss.  All their winning percentages and the derivative OWP and OOWP are based on this approach. 

Just did the math for Albertus Magnus since they were first on the Region 1 data sheet.  From the latest sheet, a record of 9-2-5, in division win loss of .719, and in division SOS of .471.  For in division W/L using .5 for a tie you get 11.5/16 or .719, easy enough.  As for SoS I noticed if you take 2/3 of .719 you get .479, which exceeds the SoS without even adding the cumulative OOWP.  So .5 for a tie cannot be used.

How the NCAA is treating ties (as one-half win, one-half loss) isn't what's tripping you up with trying to compute SOS.  I'll try to guide you through it.

SOS is 2/3 OWP + 1/3 OOWP.  But you are taking 2/3 of Albert Magnus' winning percentage (see red underlined portion of quote above), not 2/3 of their opponents' cumulative winning percentage (OWP), which still wouldn't be completely correct, but much closer.  What's precisely needed is their opponents' cumulative winning percentage with results against Albertus Magnus removed.  I'll illustrate this below.

Went ahead and grabbed AM's opponent records:

Opponent	Opp record	Opp Wins	Opp Ties	Opp total games
Springfield	4-6-5	4	5	15
Casleton	4-11-1	4	1	16
New Eng Coll	8-5-4	8	4	17
Colby-Saw	3-10-4	3	4	17
Mitchell	2-10-2	2	2	14
Norwich	6-2-7	6	7	16
Anna Maria	4-9-1	4	1	14
St Joe's CT	11-1-3	11	3	15
Westfield	6-7-1	6	1	14
Regis MA	6-9-0	6	0	15
Lasell	3-8-3	3	3	14
Emman MA	5-10-1	5	1	16
Elms	2-11-3	2	3	16
St Joes ME	10-3-3	10	3	16
JWU RI	10-2-4	10	4	16
River	3-8-5	3	5	16
Totals		87	47	247

Assuming these numbers are correct, Albertus Magnus' opponents have a cumulative W-L-T record of 87-113-47 (.447).  However, that includes the results that each of those opponents had against Albert Magnus.  Albert Magnus and Springfield tied, so Springfield's W-L-T record to be used is 4-6-4 (with the head-to-head tie removed).  Castleton lost to Albertus Magnus, so their W-L-T record needs to be adjusted to 4-10-1 (removing the head-to-head loss).  And so on. Removing the opponents' results against Albertus Magnus would put their cumulative record at 85-104-42 (removing 2 wins, 9 loses, and 5 ties).  So Albertus Magnus' OWP is .459.

By taking different multipliers for ties and based on 2/3 for OWP and 1/3 for OOWP, I calculated:

Tie multiplier	Weighted OWP	Weighted OOWP	SoS
50%	.479	.149	.628
33%	.444	.138	.583
none	.375	.117	.492

And of course I couldn't recalc the .471 on the data sheet.  FAIL.  Could be I misapplied one of the opponent names as I'm not familiar with the schools or abbreviations or my math is just off somewhere.    But it seems to me any factor of a tie is too great for SoS, at least in this one example.

As I pointed out above, you took 2/3 of Albert Magnus' winning percentage, not of their OWP.  And then you used 1/3 of their OWP instead of their OOWP. 

Their OWP as I showed above is .459 after removing the head-to-head results.  So the first part of the SOS is 2/3 OWP = 2/3 (.459) = .306

That much is pretty easy to calculate reasonably quickly.  However, OOWP is where this gets laborious.  You need to collect the W-L-T records for all the opponents of each one of Albertus Magnus' 16 opponents.  In other words, the W-L-T records of over 250 teams (some may be repeats). Now, the Manual doesn't explicitly address one question, but it appears that for this step, head-to-head results do not have to be removed.  Well, I'm not going to try to do this step to confirm the NCAA's SOS for Albertus Magnus, but my guess is that cumulative winning percentage of Albertus Magnus' opponents' opponents (all 250+ of them) is .495.

If that's correct, the second part of the SOS is 1/3 OOWP = 1/3 (.495) = 0.165

Putting it all together:

SOS = 2/3 OWP + 1/3 OOWP = 2/3 (.459) + 1/3 (.495) = .306 + .165 = .471


As you can see, you can't calculate SOS quite as easily and quickly as you were trying to do. The math isn't hard in and of itself, but the volume of numbers you need to collect after first compiling the list of each opponents' list of opponents makes it a bit of work.  At D3osccer.com, we would run the numbers for the regionally ranked teams so, when we were trying to predict the at-large berths, we knew the SOS values that the committee had in front of them.

[PaulNewman]

From the 2023 Division III Soccer Pre-Championships Manual, the RvR criterion is as follows:

• Results versus ranked Division III teams as established by the final ranking and the ranking preceding the final ranking.
  

NOTE: It does not say W-L-T records versus ranked teams. It does not say winning percentage versus ranked teams. It merely says results versus ranked teams. That gives the Regional Advisory Committees (RACs) and the men's soccer committee a lot of latitude and room for subjectivity.

The data sheets provide the W-L-T record versus ranked teams as a piece of data which is part of this criterion, but not the whole of it. What more could the data sheets provide?  How would you concisely capture which wins came against teams highly ranked versus lowly ranked?  You can't. Just like the data sheets can't provide information on head-to-head competition and results versus common Division III opponents.  The data sheets only capture the quantifiable/numerical criteria.

I'm pretty sure that committee members for basketball and for soccer have confirmed that they do consider things like whether results/wins came against highly ranked teams or lowly ranked teams.  The W-L-T record is a starting point, but the committees can and apparently do dig deeper than that as needed to compare teams.

I think it would follow that, if the committee can and does weigh some results/wins greater based on opponents' ranking within the same region, than the committee could also, and likely does at times, consider results/wins versus teams ranked in one region as having more weight than results/wins versus teams ranked in another region.  I don't recall this specifically being confirmed by a committee member, but the criterion, written so general and broad as it is, wouldn't prohibit it.

Cristan, thanks for weighing in...

I had always thought the big picture message was something like "whatever subjectivity you think might come into the process is far less than you think." 

My takeaway now is don't ever assume you've figured this stuff out!

So many questions...

Every year it seems like there is some focus on who might benefit (or not) from a team slipping into the bottom of the regional rankings or falling out, and sometimes with a focus that considers an addition or subtraction to be critical in evaluating a team's viability for selection.  But sounds like in actuality a team that comes into one of the last couple of spots in a region's regional rankings can be evaluated to have less value than another ranked win (or loss...or draw).

And once a window of subjectivity is opened up, how do cmtes avoid that becoming a very slippery slope?  On selection day, how would the cmte decide between two ranked wins against F&M and Stevens versus two ranked wins against St Olaf and UW-Platteville?  If I say Trine instead of UW-Platteville, could that make a difference?  And does a loss to a lower ranked regionally ranked team potentially carry more weight than a loss to one of the top few teams in a region?

Further, once you're considering the relative value of one team's three ranked wins versus another team's three ranked wins, what prevents the cmte from looking at whether the matches (win or lose) happened very early in the season or later?  Or giving a little extra credit to a team that finishes the regular season with a flurry versus slumping to the end?  Or even injuries?  Could the cmte, hypothetically, say, well, Messiah just got Kent-Loop back, so we have to or at least can factor that in?  Or could the cmte say, look, these two teams are so even, let's give Team A the nod over Team B because even though Denison fell out of the regional rankings we all know Denison is still a really good team, so that's still a "good win"?  I mean why not consider other games of 'significance' even though those matches fall outside the RVR matrix?

[Freddyfud]

At present, the NCAA considers ties to be one-half win, one-half loss.  All their winning percentages and the derivative OWP and OOWP are based on this approach. 

Just did the math for Albertus Magnus since they were first on the Region 1 data sheet.  From the latest sheet, a record of 9-2-5, in division win loss of .719, and in division SOS of .471.  For in division W/L using .5 for a tie you get 11.5/16 or .719, easy enough.  As for SoS I noticed if you take 2/3 of .719 you get .479, which exceeds the SoS without even adding the cumulative OOWP.  So .5 for a tie cannot be used.

How the NCAA is treating ties (as one-half win, one-half loss) isn't what's tripping you up with trying to compute SOS.  I'll try to guide you through it.

SOS is 2/3 OWP + 1/3 OOWP.  But you are taking 2/3 of Albert Magnus' winning percentage (see red underlined portion of quote above), not 2/3 of their opponents' cumulative winning percentage (OWP), which still wouldn't be completely correct, but much closer.  What's precisely needed is their opponents' cumulative winning percentage with results against Albertus Magnus removed.  I'll illustrate this below.

Went ahead and grabbed AM's opponent records:

Opponent	Opp record	Opp Wins	Opp Ties	Opp total games
Springfield	4-6-5	4	5	15
Casleton	4-11-1	4	1	16
New Eng Coll	8-5-4	8	4	17
Colby-Saw	3-10-4	3	4	17
Mitchell	2-10-2	2	2	14
Norwich	6-2-7	6	7	16
Anna Maria	4-9-1	4	1	14
St Joe's CT	11-1-3	11	3	15
Westfield	6-7-1	6	1	14
Regis MA	6-9-0	6	0	15
Lasell	3-8-3	3	3	14
Emman MA	5-10-1	5	1	16
Elms	2-11-3	2	3	16
St Joes ME	10-3-3	10	3	16
JWU RI	10-2-4	10	4	16
River	3-8-5	3	5	16
Totals		87	47	247

Assuming these numbers are correct, Albertus Magnus' opponents have a cumulative W-L-T record of 87-113-47 (.447).  However, that includes the results that each of those opponents had against Albert Magnus.  Albert Magnus and Springfield tied, so Springfield's W-L-T record to be used is 4-6-4 (with the head-to-head tie removed).  Castleton lost to Albertus Magnus, so their W-L-T record needs to be adjusted to 4-10-1 (removing the head-to-head loss).  And so on. Removing the opponents' results against Albertus Magnus would put their cumulative record at 85-104-42 (removing 2 wins, 9 loses, and 5 ties).  So Albertus Magnus' OWP is .459.

By taking different multipliers for ties and based on 2/3 for OWP and 1/3 for OOWP, I calculated:

Tie multiplier	Weighted OWP	Weighted OOWP	SoS
50%	.479	.149	.628
33%	.444	.138	.583
none	.375	.117	.492

And of course I couldn't recalc the .471 on the data sheet.  FAIL.  Could be I misapplied one of the opponent names as I'm not familiar with the schools or abbreviations or my math is just off somewhere.    But it seems to me any factor of a tie is too great for SoS, at least in this one example.

As I pointed out above, you took 2/3 of Albert Magnus' winning percentage, not of their OWP.  And then you used 1/3 of their OWP instead of their OOWP. 

Their OWP as I showed above is .459 after removing the head-to-head results.  So the first part of the SOS is 2/3 OWP = 2/3 (.459) = .306

That much is pretty easy to calculate reasonably quickly.  However, OOWP is where this gets laborious.  You need to collect the W-L-T records for all the opponents of each one of Albertus Magnus' 16 opponents.  In other words, the W-L-T records of over 250 teams (some may be repeats). Now, the Manual doesn't explicitly address one question, but it appears that for this step, head-to-head results do not have to be removed.  Well, I'm not going to try to do this step to confirm the NCAA's SOS for Albertus Magnus, but my guess is that cumulative winning percentage of Albertus Magnus' opponents' opponents (all 250+ of them) is .495.

If that's correct, the second part of the SOS is 1/3 OOWP = 1/3 (.495) = 0.165

Putting it all together:

SOS = 2/3 OWP + 1/3 OOWP = 2/3 (.459) + 1/3 (.495) = .306 + .165 = .471


As you can see, you can't calculate SOS quite as easily and quickly as you were trying to do. The math isn't hard in and of itself, but the volume of numbers you need to collect after first compiling the list of each opponents' list of opponents makes it a bit of work.  At D3osccer.com, we would run the numbers for the regionally ranked teams so, when we were trying to predict the at-large berths, we knew the SOS values that the committee had in front of them.

Makes sense now thanks for clarifying.  And yea pulling OWP is one thing, but forget OOWP without access to the database.  Probably not a surprise, but with all of these inputs this tells me SoS can be volatile for each ranking.  Perhaps as more games are played there is a mean regression.  From scanning the data sheets it seemed .600 was somewhat of a break point for teams likely to be ranked.  And I guess the next ranking will be out today.

[coach analytics]

You really need to watch SOS pretty closely.

Last year I was absolutely shocked that Vassar and Carnegie Mellon got in over Montclair State.  (and Montclair ST beat Vassar Head to Head!)

Here were the resumes of my bubble teams.

                              
                              
Bubble teams from 2022                              
                         Overall            
Those who made the tournament   Win%      SOS               RvR
                              
   Vassar               0.676      0.594      1-3-4
   Carnegie Mellon            0.719      0.577      1-3-1
   Lynchburg               0.750      0.559      2-0-3
   Catholic               0.722      0.572      2-1-2
   Case Western            0.700      0.604      2-2-2
                              
                              
Those who did not make the tournament                              
                              
   Montclair State            0.810      0.537      2-2-2
   Drew                       0.778      0.563      1-1-1
   Hamilton               0.625      0.590      2-4-1