ESPN Crew is Unlistenable

[CU77]

Basketball (at least d1) uses it as a data point, but ultimately the committee looks at the overall bodies of work of the teams involved. It's not perfect, of course, but neither is straight RPI.

In D1 basketball, all official criteria were removed about 5 years ago. The committee is now charged with simply putting together the best teams, period. Carc would love it.

I think this would totally not work in lacrosse, precisely because there is less data. Much more opportunity for bias to sneak in and dominate the selections. The QuACkers would be happy to make the selections without knowing any game results at all! Because of course their faves must be the best.

So I say: pick a formula, then stick to it. And RPI is as good as anything else.

[ICGrad]

Basketball (at least d1) uses it as a data point, but ultimately the committee looks at the overall bodies of work of the teams involved. It's not perfect, of course, but neither is straight RPI.

And RPI is as good as anything else.

This is where we'll never agree. RPI has some advantages and some disadvantages; I have no doubt it could be significantly improved upon.

EDIT: I also think this is interesting on two fronts:

1: Basketball, with a season 2-3 times longer than lacrosse and more intra-conference games and overall datapoints and thus, one would assume, with a more accurate RPI, has completely abandoned RPI. I wonder why that is?

2: Why would choosing based on "the best teams, period" lead to more ACC and so-called blue bloods? It didn't this year, when the committee moved away from straight-RPI. Why would you assume that a committee comprised of knowledgeable representative from a ranges of conferences is always going to just take the Dukes and NDs of the world?

[CU77]

Why would you assume that a committee comprised of knowledgeable representative from a ranges of conferences is always going to just take the Dukes and NDs of the world?

Because I'm a Cornell grad. We will never forget 1970.

[ICGrad]

Why would you assume that a committee comprised of knowledgeable representative from a ranges of conferences is always going to just take the Dukes and NDs of the world?

Because I'm a Cornell grad. We will never forget 1970.

Newsflash: My guess is you wouldn't have been #1 in RPI in 1970, either, and would have lost out.

And have you forgotten 2019?

[Wheels]

LaxRef's strength of record:

https://lacrossereference.com/stats/str ... rd-d1-men/

The 8 at-larges based upon this metric would have been:
Rutgers
Virginia
Yale
Cornell
Princeton
Notre Dame
Army
Jacksonville

Jacksonville is tied with Duke, but the head-to-head win would go to Jacksonville.

That would mean no Brown, OSU, and Harvard this year.

[wgdsr]

then #7 wins don't count because of something we made up. they didn't "dig into numbers". they ran with their own stuff.

So you're saying that the committee should have gone straight RPI, and Duke should have gotten in, and ND should have gotten in.

That's fine. I mean, that's they way it's been the vast majority of the time, and we all sit here and point to all of the inconsistencies and limitations of RPI, again especially over such a small sample size, and how teams can have really pretty mediocre seasons, but as long as their loses were to teams with good records and/or high RPIs, such a team will still have a high RPI and skate into the tourney (looking at you, Hopkins, but also sort of to at Duke this year, no?)

Ultimately, I fully expected Duke and ND to both get invites and was surprised they didn't. But I don't think it's the crime of the century. I think there have been far bigger snubs over the years, and this one is getting as much attention as it is because it's Duke and ND.

i thought we hashed this already but maybe it was someone else.

if i was on the committee and not allowed to change how we were doing things (i.e. telling coaches exactly how we'd be prioritizing criteria and selecting teams pre-season), then i'd vote straight rpi. because that's how things had been done. imo.

ftr, i have railed against using straight rpi for years. it's not even a criterion. they finally beat me after 2019. they admitted that's what they're doing, and at least teams would know what was up. guess not!

anyway, the committee decided to change things up. fine, i guess, not like that's the 1st time it happened.

so then they say big wins and bad losses were the priority in separating teams at the bubble. again, fine.

then they put in a team based on that (harvard) and exclude another (duke). more than a little dubious on the bad losses thing, but so far so good... then they put in tosu who has absolutely no business being in over nd if what they said and did about priorities (big wins/bad losses) was accurate. there were 3 teams in the at large field that had zero top 10 wins, one of them was seeded, and notre dame had 2 of those wins. and another top 20. losses track, too.

but the committee just decided those wins didn't count. after using the same metric (ws and ls against rpi rank) for every other team. it's what you get i guess when there's no quantifiable guidelines, no accountability.
the changing of annual priorities isn't new and that's bad enough. the fabrication of what and for whom to use them is new. not easy to do, but this committee makes the pantheon.

[Gobigred]

Why would you assume that a committee comprised of knowledgeable representative from a ranges of conferences is always going to just take the Dukes and NDs of the world?

Because I'm a Cornell grad. We will never forget 1970.

And I will never forget 2007 when RPI caused a team to be seeded above unbeaten Cornell because it had a lot of "good losses." In the regular season that team lost to eventual NCAA seeds 1, 2, 5 and 8, and beat only seed 7. Yet RPI gave it RPI's third highest score, and the committee seeded it 3rd. How can a metric be taken seriously when it gives a team that has a 1 - 4 record against the top 8 teams the 3rd seed? You have to look at whom did a team beat and to whom did it lose, and adjust accordingly.

Why don't you, CU77, explain to us why you think RPI is the best solution for selecting and seeding?

[ICGrad]

Why would you assume that a committee comprised of knowledgeable representative from a ranges of conferences is always going to just take the Dukes and NDs of the world?

Because I'm a Cornell grad. We will never forget 1970.

And I will never forget 2007 when RPI caused a team to be seeded above unbeaten Cornell because it had a lot of "good losses." In the regular season that team lost to eventual NCAA seeds 1, 2, 5 and 8, and beat only seed 7. Yet RPI gave it RPI's third highest score, and the committee seeded it 3rd. How can a metric be taken seriously when it gives a team that has a 1 - 4 record against the top 8 teams the 3rd seed? You have to look at whom did a team beat and to whom did it lose, and adjust accordingly.

Why don't you, CU77, explain to us why you think RPI is the best solution for selecting and seeding?

Or, again, Duke this year: #7 RPI with a single game against a top 10 RPI team (which they won, over #9 Virginia), a win over the #13 team., two wins over the #17 team...and 6 losses, including 3 to teams ranked #20 or lower in the RPI.

When I asked how that translates to a #7 RPI (how can you be #7 when your highest win is over #9 and you have 3 losses to teams 20 or lower?) or when I suggest that their RPI was grossly inflated by games (not wins...games) against teams with good records, the answer I keep getting is that RPI can't be inflated, it's a formula, it's the maths, etc, etc...

Look, I get it. I'm a computer programmer for a Fortune 100 company. But sometimes algorithms kinda suck. Sometimes they're wrong, or have severe limitations that are exposed by edge cases. You don't just shrug and say: "It's the maths." (At least I don't; they'd fire my ass). You fix the algorithm. You come up with something better.

[Or you do what the committee did this year: You use the results as one datapoint, but you look at other data points as well.]

In Duke's case (or the case Gobigred cited above, or Hopkins in 2019, or Hopkins in 2016, etc...) it's not that they didn't pass the eye test; they didn't pass the smell test. The inputs produced outputs that were simply awful and made no sense. To just suggest we shrug and say "Oh well; that's RPI for ya" because it's easier or "objective" seems like flawed reasoning.

[Typical Lax Dad]

Can anyone let me know when the selection committee said RPI is the only criteria for at large bids. Not “We looked at RPI” but “we only look at RPI”.

[Henpecked]

When I asked how that translates to a #7 RPI (how can you be #7 when your highest win is over #9 and you have 3 losses to teams 20 or lower?) or when I suggest that their RPI was grossly inflated by games (not wins...games) against teams with good records, the answer I keep getting is that RPI can't be inflated, it's a formula, it's the maths, etc, etc...

That is what has always bothered me about RPI. You win the RPI game in the off season when you put your schedule together. You don't even have to be competitive in these games against high RPI teams. You just need to show up because goal differential does not enter into the equation.

People on the forum always argue that goal differential shouldn't enter the equation. But nobody has ever clearly explained why.

[ICGrad]

When I asked how that translates to a #7 RPI (how can you be #7 when your highest win is over #9 and you have 3 losses to teams 20 or lower?) or when I suggest that their RPI was grossly inflated by games (not wins...games) against teams with good records, the answer I keep getting is that RPI can't be inflated, it's a formula, it's the maths, etc, etc...

You win the RPI game in the off season when you put your schedule together.

This, exactly.

[Hoxwurth]

RPI is a terrible metric. Has been since the beginning of time.

Please tell me the specific objective formula that you think is better.

Lacrosse Reference's strength of record. Would have put a bunch of the Ivies in based on a tremendous season, but not over 85% of the conference.

Even better than that would be a strength of record computed against a NET-type metric that incorporates home/away and score differential. While LaxRef's Elo was (and I assume still is) well calibrated, the formula rests on a carryover metric from the prior season. Teams would understandably complain that their baseline was based on work outside the season.

[wgdsr]

RPI is a terrible metric. Has been since the beginning of time.

Please tell me the specific objective formula that you think is better.

Lacrosse Reference's strength of record. Would have put a bunch of the Ivies in based on a tremendous season, but not over 85% of the conference.

Even better than that would be a strength of record computed against a NET-type metric that incorporates home/away and score differential. While LaxRef's Elo was (and I assume still is) well calibrated, the formula rests on a carryover metric from the prior season. Teams would understandably complain that their baseline was based on work outside the season.

pass on net and elo. score differential is not any kind of barometer i'd like to see used.

[ICGrad]

RPI is a terrible metric. Has been since the beginning of time.

Please tell me the specific objective formula that you think is better.

Lacrosse Reference's strength of record. Would have put a bunch of the Ivies in based on a tremendous season, but not over 85% of the conference.

Even better than that would be a strength of record computed against a NET-type metric that incorporates home/away and score differential. While LaxRef's Elo was (and I assume still is) well calibrated, the formula rests on a carryover metric from the prior season. Teams would understandably complain that their baseline was based on work outside the season.

I saw the list that Wheels posted yesterday. An interesting at-large field; I'd like to see tis extended over several seasons, and know a bit more about the overall methodology.

For this year's field, though, it does seem to produce a more-fair at-large field.

EDIT: I wouldn't have objected to a field that included Army and Jacksonville; I like the idea of a field that's representative of a broader cross-section of the d1 lax universe.

[Hoxwurth]

RPI is a terrible metric. Has been since the beginning of time.

Please tell me the specific objective formula that you think is better.

Lacrosse Reference's strength of record. Would have put a bunch of the Ivies in based on a tremendous season, but not over 85% of the conference.

Even better than that would be a strength of record computed against a NET-type metric that incorporates home/away and score differential. While LaxRef's Elo was (and I assume still is) well calibrated, the formula rests on a carryover metric from the prior season. Teams would understandably complain that their baseline was based on work outside the season.

pass on net and elo. score differential is not any kind of barometer i'd like to see used.

I think most agree that wins and losses should be the most important category. The issue becomes differentiating between various wins and losses. Tweaking RPI will still lead to the problem where scheduling is a large determinant as opposed to results of playing. Score differentials are important for determining the best teams, and therefore the best wins.

I would still use strength of record (actual wins and losses) to determine which teams get bids, but the quality of those wins and losses needs to be based on something more than RPI, which didn't work in a 40-game basketball season.

Otherwise, how do you improve on RPI?

[xxxxxxx]

Make it 24 teams, the top 8 get a bye, lower 16 play on a Wednesday, schedule is the same just more action. Yes, 25 - 30 will still complain but it would be less. This year many teams that could have won games were left out, but the system is not designed to pick the top 18 teams. I'm sure the NCAA has some reasons to keep the field at 18 but I hate them anyway.

How about D3 playing back-to-back playoff games now on a Saturday and Sunday, they really care about the "Student-athlete".

Oh and to stay on the subject of this thread, the three amigos at ESPN are horrible and setting the game back every time they cover games.

[wgdsr]

RPI is a terrible metric. Has been since the beginning of time.

Please tell me the specific objective formula that you think is better.

Lacrosse Reference's strength of record. Would have put a bunch of the Ivies in based on a tremendous season, but not over 85% of the conference.

Even better than that would be a strength of record computed against a NET-type metric that incorporates home/away and score differential. While LaxRef's Elo was (and I assume still is) well calibrated, the formula rests on a carryover metric from the prior season. Teams would understandably complain that their baseline was based on work outside the season.

pass on net and elo. score differential is not any kind of barometer i'd like to see used.

I think most agree that wins and losses should be the most important category. The issue becomes differentiating between various wins and losses. Tweaking RPI will still lead to the problem where scheduling is a large determinant as opposed to results of playing. Score differentials are important for determining the best teams, and therefore the best wins.

I would still use strength of record (actual wins and losses) to determine which teams get bids, but the quality of those wins and losses needs to be based on something more than RPI, which didn't work in a 40-game basketball season.

Otherwise, how do you improve on RPI?

one solution i've thrown out in the past is a hybrid to elo, someone would need to back test for the proper numbers. no score differential, no preseason set. don't need to even look at prior to halfway thru the season...

+ points for wins and - points for losses, using a base of opponents' (eventual)end of year rpi. eliminates the cutoffs of 5, 10, 20.

70 teams. win against # 70 gets you say 10 points. then 1 more point for say the next 39 teams. from 30 to 21 it climbs by 2 for each. from 19 to 11 it climbs by 3. from 10 to 6 it climbs by 4, and from 5 to 1 it climbs by 5. so you get ~145 points for beating # 1.

for losses, you reverse it. maybe start a loss to # 1 at 20 points, add 1 to top 30 or more, possibly set different bump ranges for worse losses.

everyone finishes with a point total. smoothed to a 15 game average per game. a key point also is that with just 15 games more or less, you'll have to get your share of good wins not just ok ones or you won't be able to keep up with the joneses. and losses will nick you with only so many games to get them back.

[FannOLax]

At least Quint and Anish know that Rutgers is in New Jersey, the land of Springsteen. "As Shroff and Quint Kessenich breezed through Sunday’s first quarter, the broadcasters arrived in living rooms like they were sprung from cages, chrome-wheeled, real fuel-injected and ready to step over the line.

Throughout their two and half hours on air, the veteran broadcast team packed more than 20 references to Springsteen."

https://www.nj.com/rutgers/2022/05/rutg ... -four.html

[Typical Lax Dad]

When I asked how that translates to a #7 RPI (how can you be #7 when your highest win is over #9 and you have 3 losses to teams 20 or lower?) or when I suggest that their RPI was grossly inflated by games (not wins...games) against teams with good records, the answer I keep getting is that RPI can't be inflated, it's a formula, it's the maths, etc, etc...

You win the RPI game in the off season when you put your schedule together.

This, exactly.

It backfired on ND this season.

[CU77]

the changing of annual priorities isn't new and that's bad enough. the fabrication of what and for whom to use them is new.

Not new.

I think it was 2012 when committee chair Tony Seaman said Denver was helped by their strong out-of-conference record, which is absolutely not a criterion. I think it was Penn State that got left out that year.

I will repeat my mantra: without a quantifiable formula, this will happen every year.

AFAIK there is no agitation for change among the coaches, which is what it would take to make a change.