Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - Michel Bernstini

#1
Lacrosse, track, cross country, softball / Re: Lacrosse
January 28, 2020, 04:50:40 PM
FanLax has a good Forum.
#2
Men's soccer / Re: Big Dance
December 07, 2019, 08:06:37 PM
Should be a good second half.   :)
#3
Men's soccer / D1 Men Soccer
December 01, 2019, 10:29:59 PM
On Saturday, December 6th, the D1 Men will play their Quarterfinal games.  These, in order, are the best games:

Clemson #2 vs #3 Stanford
Virginia #1 vs #5 SMU
Georgetown #4 vs #8 Washington
UC Santa Barbara #6 vs #14 Wake Forest
#4
Men's soccer / Re: NESCAC
November 30, 2019, 07:45:38 PM
Quote from: Buck O. on November 28, 2019, 07:50:25 AM
Obviously, it's your model, and you can make whatever choices you make with regard to it, but if you're going to use it to support your assertions, you have to provide people with a reason to take it seriously.  With that in mind:

1. Choosing to focus on a certain fixed number of wins or losses is an arbritary distinction that will lead the model to produce unreasonable results, no matter whether you fix that number at three or some other figure.  For example, suppose Team 1 has wins over the teams ranked #1, #2 and #3 in the country, but all of its other wins are against teams in the 300s, while Team B has wins over teams ranked #1, #2, #4, #5, #6, #7, #8, #9, and #10.  Your model says that Team A has the more impressive slate of wins.  This is obviously wrong.  Similarly, by limiting its assessment to the three worst losses, your team may incorrectly conclude that Team C, with only three bad losses, has a less impressive slate of losses than Team D, even if Team D has a dozen losses to bad teams, just as long as the three worst losses aren't quite as bad.

And as a result of this, it appears that your model could rank a 3-13 team above a 13-3 team, even if they played the exact same schedule.  For example, suppose that Teams A and B both play Teams 1 through 16, with 1 being the best opponent and 16 being the worst.  If Team A has wins over 1, 2 and 4, and losses to everyone else, and Team B has wins over everyone except teams 14, 15 and 16, you'd rank A above B, because their worst three losses are the same and A's three best wins are better than B's.

Next, let's consider how you handle ties.  Suppose I have wins against teams #1, #2, #3 and I'm playing team #4 today.  My best tie is against team #300.  Under your system, I'll improve my ranking much more with a tie against #4 than with a win, because a win won't affect my slate of top three wins, but a tie will significantly improve my best tie.  Again, that doesn't make sense.

2.  Obviously NCAA tournament games are more important that non-tournament games because they determine who wins the NCAA tournament, but that does not mean that they are necessarily significantly more important for the purposes of determining which team is the best team.  Just because Conn beat Hopkins does not mean that Conn is better then Hopkins.  Upsets happen but that does not mean that a repeat matchup should be expected to lead to the same outcome.  And if the intent of your ranking system is to determine a relative ranking of teams, then the point should be to assess which team is better.  If you're simply trying to replicate the results of the tournament, then what's the point of even having a ranking system?

3.  It seems to me that the only "philosophy" that should matter is trying to produce the best set of rankings, rather than making arbitrary decisions as to which info to include or exclude without regard to whether including it or not improves the ability of your system to evaluate teams.  Close games could generally go either way and luck may be a significant factor in determining the winner, and a system that ignores that and pretends that a 2-1 win is the same as a 5-0 blowout is ignoring relevant information.

4.  Things don't work like that.  The better team doesn't always win.  If the best team in the country plays a schedule consisting wholly of top-20 teams (at neutral sites), it will be favored in each individual game, but that doesn't mean it is expected to get through that schedule undefeated.  Similarly, if the #20 team in the country plays a schedule wholly composed of other top-20 teams (again at neutral sites), it will be an underdog in each game, but it's almost certain that it will pick up several wins in the process even though it will probably have a losing record. 

More generally, you're missing the point.  If you rate teams based on their best three wins, teams with tough schedules have many more opportunities to rack up quality wins than do teams with weaker schedules.  Similarly, if you rate teams based on their worst three losses, teams with weak schedules have many more opportunities to accrue bad losses than do teams with strong schedules.  This, again, is a problem with ranking based on a certain fixed number of wins or losses.

1.
a. This is obviously wrong... I don't think it's so obviously wrong but, for the sake argument, let's say it's wrong.  Let's say that Team A has better wins but... Team B in Iowa traveled to the east coast play #1, #2, and #3 and, if they played #4, #5, and #6, they believe that they would've won those games, too, but... they can only travel so much.  Bottomline, based on what was put on their plate, Team B performed.  Which is (one of the reasons) why we need to draw a line somewhere.  We can't punish Team B just because they're based in, for example, Iowa.  D   
b. Similarly... 3 worst losses... Correct.  Team A could lose to teams #100, #99, and #98 and those are their only three losses while Team B loses to the same as well as #97, #96, etc. and, as long as they have the same Best Wins, though Team B should be ranked lower, they'll be ranked the same.  It's true.  I'm aware and, thus far, even with a gazillion HS games, this situation hasn't arisen.   
c. It appears that your model could rank a 3-13 team above a 13-3 team... It's true, but like b (above), it hasn't happened yet.  But you're right.  It's possible.

d. ties... I see "ties" and I smile.  With lacrosse, ties are basically a non-issue.  i needed to make a new system to include ties when I started using this system for soccer.  (Darn soccer!)  Ties aren't simple.  And what you're pointing out is one of the not-simples. :)  You're right.  And *that* does occur.

2.  "Just because Conn beat Hopkins does not mean that Conn is better then Hopkins... If you're simply trying to replicate the results of the tournament, then what's the point of even having a ranking system?"... First, It would be easy for me to adjust the system so play-off games are like any other game.  But... They're not.  (I want to come back to this.)  Second, I'm not trying to replicate the play-off system.  I'm trying to mirror a season.  But back to the first point.  Excuse for me reverting to lacrosse but it's a good example.  During the 2019 D2 Men's season, Merrimack had a good, not great season.  the, in the tourney, they beat the #6, #4, #2 ranked teams and then, in the finals, beat the #1 ranked undefeated Limestone.  So, the question is, Should Merrimack be ranked #2 or #1?  Based on their season, #2.  Based on their performance that season including when the stakes were higher, #1.  Personally, I think #1 is the correct answer.  but, make no mistake, then and now, i realize and realized that this is a very real question.  So let me ask a question. Let's say that Centre wins out.  They beat Amherst and then Tufts or Calvin.  And let's say, as a consequence, by the numbers, this makes them the #2 in the country.  Do you think they should be #2?  or #1?

3.  A 2-1 win is the same as a 5-0 blowout is ignoring relevant information... It's true.  And some systems look even beyond game scores into more details.  In theory, I should be doing that too.  But I don't.  I like ws and Ls.  But I hear you.  I just like Ws and Ls.

4.  Correct.  A tough (and an easy) schedule is a blessing and a curse.  And i believe that this system captures this duality.
#5
Men's soccer / Re: NESCAC
November 30, 2019, 07:11:49 PM
Quote from: YoungBuck on November 27, 2019, 01:35:27 PMI doubt even the great Ken Massey would say his ranking system should be the only method upon which a teams value is placed.

By all means, keep refining your results and improving your model, but be aware that the subjectivity that it fails to capture could be the difference.

Of course.  And I'm one of those humans.  :)  Of course the Rankings aren't the be-all-end-all.  :)
#6
Men's soccer / Re: NESCAC
November 27, 2019, 01:06:18 PM
One more thing.  Something I like about this ranking system (that I don't like about a lot of others) is that (1) It's not someone's or someones' opinion, (2) it's 100% simple and understandable.  i.e. At the end of the season, when Clemson doesn't get to play in the CFB Semifinals or St. Lawrence is left out of the Tourney or whatever, the players, coaches, and otherwise can see exactly why.  And, as a consequence, Clemson will improve their Out-of-Course Strenth of schedule, etc., and (3) It's fair.  You move up or down based on your team's resume and your resume is based on your best and worst performances.  And that is fair.
#8
Men's soccer / Re: NESCAC
November 27, 2019, 12:36:30 PM
Quote from: Buck O. on November 27, 2019, 10:46:39 AM
1.  What's magical about the number three?  Why is it that the best three wins and the worst three losses are all that matters?
2.  Similarly, what's magical about playoff games?  Why should they receive greater consideration in determining who should be ranked where?  UMBC beat UVA in the 2018 NCAA basketball tournament, but that didn't make them better than UVA.
3.  Why not use scores?  There's lots of valuable information there that you're discarding.  This is particularly true with respect to teams that may otherwise be hard to rank because of limited info.  For example, if a team with an excellent record, but which has played a pretty weak schedule, has played only one really good team, I'm going to look at them rather differently if they lost that game 5-0 than if they lost 2-1.  Now, one can go too far here--you shouldn't lean too far on one game in making assessments--but nevertheless focusing solely on W/L/T will reduce the accuracy of your assessments.
4.  There seems to be a significant bias in favor of teams with tough schedules inherent in your method.  To use an extreme example, suppose my schedule is exclusively against top-20 teams and I go 3-13.  If I understand your method correctly, you're going to conclude that I'm awesome because my best three wins are against top-20 teams and my worst three losses are also against top-20 teams.  I might very well be in the top 20.  But, at the end of the day, I did only win three times in 16 attempts against that level of competition.  I don't belong in the top 20.

1. Just a good number and we need to draw a line somewhere.  You can't do all Ws count because weaker schedules have a large disadvantage.  This is particularly true for HS teams in non-hotbed regions.  you can't do 1 win because 1 game is only 1 game.  So, the question is, How many games does a team need to win to prove that they're legit.  1, too few.  2, better.  And... 3 just seems to be a good number.  (Also, allows for wins to count 3x more than a tie.  i.e. One tie is counted.  Three wins and three losses are counted.  And the "3 Points for a Win, 1 Point for a Loss" formula is all but universal.)
2. UMBC is a unique example.  None the less, they earned the spot.  Bottomline, play-off games are not the same as regular season games.  they're later in the season and... They simply mean more.  The team that wins the National C'ship Tourney should be #1 (as long as they layed the top teams to win it all).
3. Just a philosophical difference.  Teams play to win, not to win by as many points as possible.  Especially teams who play 3 or so games per week.  Lots of teams pace themselves and they shouldn't be punished for doing so. 
4. I disagree.  If a team beats #18, #19, and #20 and loses to #1-#16, they should be the #17 team in the country.
#9
Somebody may have already mentioned this but the Messiah coach wrote a book about why they're so successful and, so to speak, faith plays no small part.
#10
Men's soccer / Re: World Cup and European leagues
November 26, 2019, 01:13:41 PM
Champions League day...
#11
Men's soccer / Re: NESCAC
November 26, 2019, 12:08:58 PM
Yes!  :)  Of course it's not just Ws and Ls.  It's Ws and Ls and who those Ws and Ls are against.  Ws and Ls is short-hand.  With this said, based on Ws and Ls (and Ties) and who those Ws and Ls (and ties) are against, Why would you put so many teams ahead of Conn College?  e.g. Messiah (and several more).  Based on their season, I don't see how they go ahead of Conn College.  Do you just not like the camels?  Keep in mind, in the play-offs, Conn College just beat JHU.

Separately... Champions League at 1!  :)
#12
Men's soccer / Re: NESCAC
November 26, 2019, 11:21:05 AM
Quote from: PaulNewman on November 26, 2019, 10:57:28 AM
it's YOUR ranking system and yours only.  Just above you replied it was based on Ws and Ls, but now.....there's more?  How about you throw in the steak knives and then we'll talk.  If you're gonna sell something at least give it some prima facie validity.

If it was based solely on Ws and Ls, it'd simply be Winning % and, as already spoken to, this is a very dull tool to assess teams.  I already put the link to how the system works.  It's based on Best Ws and Worst Ls.  Not solely Ws and Ls.  A tool based solely on Ws and Ls (that isn't a closed system where every team plays basically the same schedule thus countering any SOS issues) is silly. 
#13
Men's soccer / Re: NESCAC
November 26, 2019, 11:11:27 AM
Quote from: PaulNewman on November 26, 2019, 10:57:28 AM
So you're selling a business lol?

No.  I'm sharing a passion.  In this case, a passion for D3 Men's Soccer.  A business involves $.  My passion doesn't involve $.  Just interest and passion.
#14
Men's soccer / Re: NESCAC
November 26, 2019, 11:08:59 AM
Quote from: PaulNewman on November 26, 2019, 10:38:49 AM
Please explain for example how you current have Conn ahead of Calvin who has a FAR SUPERIOR W-L record and has been in four out of the last five final fours?  And if Conn is #3 does that make Midd #4?

My Rankings need a starting point.  If the starting point is the NCAA Rankings, then Conn College is ahead of Calvin.  If the starting point circles back to my own rankings, then Calvin is #2, then Tufts, then Conn College at #4.  (I prefer the latter so that's what the current rankings are based off of.)  But to answer your question, this is exactly my point.  W-L Record is a legit metric to decide which team is better.  But obviously, W-L while ignoring SOS (strength of schedule) creates very real issues.  Calvin's SOS is #82 in the nation.  Conn College's is #8.

Four out of the five last final fours... Another great example and, personally, I believe that this is the greatest problem for the mainstream rankings: teams' brands.  Teams' brands and group think.  College Football is a good example.  (I'm not a big CFB fan but the rankings are controversial so I couldn't resist.  Across the board, Clemson and Georgia are #3 and #4.  But, objectively, it makes no sense.  At least with Clemson, you can argue that they deserve to be there because of their game scores (though I do not agree.  Amherst beat Tufts 2-1, Babson beat them 2-0.  Grossly simplifying the metric but... this doesn't mean babson is better than Amherst.  I like Ws and Ls, not game scores) but... It makes no sense.  Some teams can have a terrible loss and then season is over.  A team like Georgia can lose to a south Carolina and it's treated very differently.  For obvious reasons, this isn't fair, isn't right, and, for me, isn't an accurate ranking.  In terms of Group Think, I don't think it's a coincidence that ESPN or whoever forgives Georgia for a loss and then the rest follow suit.  Bothers me...

No, Middlebury is #9.
#15
Men's soccer / Re: NESCAC
November 26, 2019, 10:47:49 AM
re: the "delusional" comment above, as HiYa spoke to, one cannot lean to heavily on objective metrics and... Nor too heavily on subjective opinions.  But objectivity and subjectivity both have a place.  One can evaluate D3 teams in innumerable ways: win %, Goal differential, strength of schedule and win %, etc.  I happen to enjoy playing around with this stuff and did for 2 seasons with lacrosse (mostly D1 and d3 mens but then... Men's D1, D2, D3, Women's D1, D2, D3, NJCAA, NAIA, MCLA, WCLA, every Boys and Girls HS team, etc.) and... There are benefits and flaws to every metric but some are more flawed than others.  Anyway, you may not agree with Conn College at #3 but it's not based on my subjective opinion.  It's based on a rational system.  So, as much as we pause and say, What's the best stat for a goalie?, we can do the same for the teams.  And teams are much easier.  The heart of my rankings are (1) Each team's three Best Wins (2) Three Worst Losses as well as (3) one Best Tie and one Worst Tie.  So a team can be amazing, undefeated, best in the nation in everyone's eyes, but if they haven't beaten good teams to demonstrate this, then they won't be #1 in the rankings.  Best Wins, Worst Losses, Best Tie, Worst Tie.  (Oh, one more thing.  Come play-offs, when you beat a team, you automatically step in front of them in the Rankings.  i.e. RPI, Centre, or whoever may not "deserve" to be #8, #10, etc. but if they beat the #9, #11, etc. team in the playoffs, they get the spot.)  You may not love the system but it's fair and it most certainly isn't delusional.  :)

This is the D3 Men rankings: http://app.endlesssoccer.com/category/ranking/54/2019
This is the system behind the rankings: http://endlesssoccer.com/2019/10/23/endless-soccer-computer-ranking-how-it-works/

And, because I find these fun, helps to keep me organized, etc.:
Best Games: http://app.endlesssoccer.com/category/best/games/54/2019
Best Upsets: http://app.endlesssoccer.com/category/best/upsets/54/2019
Best Upcoming Games: http://app.endlesssoccer.com/category/upcoming/games/54/2019