Pablo Championship Probabilities

[Gene Mills Fan]

141??? bartlett 

[ugarles]

1 hour ago, Pablo said:

Ranking		133	Rating	Finals	Winner
1		133:Illinois_Lucas Byrd	8504	0.434	0.294
2		133:Iowa_Drake Ayala	8122	0.282	0.154
3		133:Little Rock_Nasir Bailey	8101	0.264	0.143
5		133:Northern Colorado_Dominick Serrano	7782	0.157	0.078
6		133:Virginia Tech_Connor McGonagle	7693	0.117	0.051
7		133:Ohio St_Nic Bouzakis	7585	0.095	0.039
11		133:Cal Poly_Zeth Romney	7510	0.079	0.034
14		133:Stanford_Tyler Knox	7460	0.068	0.028
10		133:Penn St_Braeden Davis	7517	0.063	0.027
12		133:Iowa St_Evan Frost	7481	0.068	0.026
9		133:Maryland_Braxton Brown	7550	0.057	0.025
13		133:Nebraska_Jacob Van Dee	7462	0.043	0.018
18		133:Rutgers_Dylan Shawver	7176	0.042	0.013
17		133:Indiana_Angelo Rini	7186	0.032	0.011
16		133:North Dakota St_Kyle Burwick	7286	0.027	0.010
19		133:Wisconsin_Zan Fugitt	7141	0.031	0.010
20		133:Lock Haven_Anthony Noto	7139	0.028	0.009
21		133:Northern Iowa_Julian Farber	7041	0.021	0.006
NR		133:NC St_Kai Orine	6973	0.018	0.005
25		133:North Carolina_Ethan Oakley	6866	0.012	0.003
23		133:Pennsylvania_Ryan Miller	6981	0.010	0.003
24		133:Minnesota_Tyler Wells	6876	0.012	0.003
27		133:Gardner-Webb_Takeo Davis	6720	0.008	0.002
29		133:Chattanooga_Blake Boarman	6707	0.007	0.002
32		133:Army_Ethan Berginc	6645	0.006	0.001
30		133:Bucknell_Kurt Phipps	6669	0.004	0.001
35		133:Cornell_Tyler Ferrara	6577	0.005	0.001
33		133:South Dakota St_Derrick Cardinal	6598	0.005	0.001
40		133:Missouri_Kade Moore	6456	0.003	0.001
45		133:Central Michigan_Sean Spidle	6363	0.003	0.000
51		133:Edinboro_Colton Camacho	6204	0.001	0.000
55		133:Cal Baptist_Hunter Leake	6031	0.000	0.000
121		133:Michigan_Nolan Wertanen	4597	0.000	0.000

I'm trying to translate the rankings and ratings in my head, @Pablo, so please let me know if I have this correct:

1) The Rankings column is literally in the order of the odds but they are numbered by the "objective" rating. I'm curious how/why these would diverge. I have an instinct (a "bad" draw by the Pablo numbers) but wouldn't mind a more fulsome explanation.

2) I quoted 133 because Kai Orine is NR but by my count he had 10 matches at 133 by the end of tournament season.

3) Again, looking at 133, I see that (a) there are 33 wrestlers, so I'm going to assume this is the tournament field and (b) there is no 4 or 8 (and probably others). Should I assume one of those missing is Ryan Crookham, and the full numerical list includes wrestlers who are either redshirting, backup or injured?

Thanks again for the work

[Pablo]

38 minutes ago, ugarles said:

I'm trying to translate the rankings and ratings in my head, @Pablo, so please let me know if I have this correct:

1) The Rankings column is literally in the order of the odds but they are numbered by the "objective" rating. I'm curious how/why these would diverge. I have an instinct (a "bad" draw by the Pablo numbers) but wouldn't mind a more fulsome explanation.

2) I quoted 133 because Kai Orine is NR but by my count he had 10 matches at 133 by the end of tournament season.

3) Again, looking at 133, I see that (a) there are 33 wrestlers, so I'm going to assume this is the tournament field and (b) there is no 4 or 8 (and probably others). Should I assume one of those missing is Ryan Crookham, and the full numerical list includes wrestlers who are either redshirting, backup or injured?

Thanks again for the work

OK, let me see if I can help you out

1) The "Rankings" column is their overall ranking in Pablo rankings, so that's against the whole D1 field (of qualifying wrestlers).  The wrestlers are listed in order of their win probability, but the Rankings column (the first one) is their rank within all of D1.  The differences in how they come out is largely due to differences in matchups due to differences between Pablo and seeds.

2) It's certainly possible (actually certain) that I have missed matches here and there, but the biggest issue with Orine is that he has matches at 141.  If you say he has 10, I'll believe you, but I only have 8 in the database, and 3 are at 141.  However, you can still estimate his ranking by looking at this rating.  His rating of 6973 puts him a smidge behind #23 Miller at 6981 but ahead of #24 Wells at 6876.

Similarly, Teague Travis wrestled some at 157 but dropped to 149 for the Big 12 tournament.  His rating of 7185 is between the guys at 20 and 22. 

***I'll just put in here my PLEA for a better source for scores than having to plod through results on Track event by event and team by team.  If ANYONE can come up with a solution, this will work so much better!!!!***

3) The top 40 at each weight are listed in the Rankings thread.  That list includes everyone eligible (>5 matches), whether they are red shirt or even non-D1.  This list includes all the wrestlers going to nationals, but only those wrestlers.  If there is a highly ranked person who isn't in the tournament, you won't see them here.  In 133 for example, Pablo has Davino from OhSt at #4, but Bouzakis is representing tOSU (in fact, one of Davino's wins in the database was over Bouzakis at Edinboro!).  So if you see a ranking missing in the list, you can cross reference against the top 40 in the other thread.

Edited yesterday at 05:23 PM by Pablo

[ugarles]

2 minutes ago, Pablo said:

2) It's certainly possible (actually certain) that I have missed matches here and there, but the biggest issue with Orine is that he has matches at 141. 

Re: Orine https://www.wrestlestat.com/wrestler/56908/orine-kai/profile

[Pablo]

Yeah, no.  I'm not going to chase down every stinking wrestler one at a time to get all their matches.  I've got 4000 wrestlers in the database.  It's not going to work.

I've tried to go through all the teams (and I did go through NCState's schedule) but matches have slipped through the cracks.

I've done my best, but until there's a better source, I'm going to miss things.

[ugarles]

Just now, Pablo said:

Yeah, no.  I'm not going to chase down every stinking wrestler one at a time to get all their matches.  I've got 4000 wrestlers in the database.  It's not going to work.

 

No worries! I was more interested in figuring out where the sources diverged than casting blame for the divergence. This isn't for publication in pursuit of tenure. I wouldn't run down and sort out the redshirts either!

[Pablo]

Just now, ugarles said:

No worries! I was more interested in figuring out where the sources diverged than casting blame for the divergence. This isn't for publication in pursuit of tenure. I wouldn't run down and sort out the redshirts either!

I'm not sure what your "divergence" concern is, though.  Orine's rating gives him a win probability that's similar to those rated similarly.  He's just behind Farber but his probability is better than Miller who's rated higher.  This is all just an issue of matchups they face in the brackets, though.  If we go with that 24ish ranking, that's actually consistent with his seed even.  So while there is some difference between the order in terms of probability and in terms of rankings, I don't think Orine is really outside it, so I don't quite understand your concern.

Given that it is only 5 matches, I'm actually pretty impressed that it came out as reasonable as it did.  However, I've seen that in other places.  My early calcs were done with a disaster of a dataset, where nothing was consistent.  I had multiple entries for the same guys, because sometimes they be in Track as Josh and sometimes it's Joshua, or you have the Michael/Mike/Mikey variations.  But amazingly, I'd run these calcs, and while Michael Smith had 12 matches and was rated 5300, Mike Smith would have 3 matches and would come in at 5150, and Mikey might have two matches and be sitting at 5500.  So a single wrestler with different names, but the resulting ratings were extremely close, considering.  

I don't know how many matches we need here to get legitimate results, but I don't know if it's as many as I've seen required in other sports.  In volleyball, for example, nothing is anywhere close until you get at least 10 or maybe even 15 matches.  That's a good sign.

[ugarles]

1 minute ago, Pablo said:

I'm not sure what your "divergence" concern is, though. 

You're thinking in terms of how it affects the ultimate accuracy; I was thinking more like a dreary bean counter. It isn't much of a concern; I'm bored at work.

[Pablo]

OK, I just figure if I can understand your concern, I can help.

So three things I'd say

1) The rankings are the ranks only of those in tournament as compared to all of D1

2) NR just means I don't enough matches at that weight, but you can still use the rating to see where they would fit

3) Differences between the order of probability vs their Pablo ranking is due to differences in competition they face in the brackets, which reflects differences between Pablo rankings and seeds.  If they were seeded by Pablo rankings () they'd line up better (not necessarily perfect because of anistropy, but close)

[Nard Dog]

Pretty cool project. I'm anxious to see how it compares with reality.

[Pablo]

2 hours ago, Nard Dog said:

Pretty cool project. I'm anxious to see how it compares with reality.

Well, back at conference tournaments, Pablo outperformed seedings in reflecting winners.  Not by a lot, but it was statistically significant, so it's off to a good start.

Unfortunately, I will be out of town next week (leaving Sunday) and won't be able to do a tournament debrief until the first week of April.

 

[MidniteToka]

On 3/17/2025 at 2:19 PM, Pablo said:

One thing to be careful about when it comes to these probabilities is to wonder, how improbable does something need to be to be surprising if it occurs?

Over the years of doing this, my line is usually about 15%.  If something has a 15% probability of occurring, I generally say, I won't be surprised if it happens.  Kind of like, if I roll a 6 sided die and get a 1, then that's not a big surprise.  However, if a roll a d20 I'm not expecting a 1.  So that's kind of my line.

Let's see how I would apply that to these rankings above.  Let's go with 285, since that is where the discussion was.

From the probabilities, I would say, "I would be surprised" if Stevenson isn't in the finals.  93%, yeah, that's a pretty solid favorite.  But for the other guy in the finals, "I wouldn't be surprised" if it is either Hendrickson or Kerkvliet, and even Trumble wouldn't be a big shocker, but he's a taller order.

For winning it all, yeah, Stevenson is a huge favorite, but "I wouldn't be surprised" if it was Hendrickson.  From a Pablo standpoint, it's hard to point to anyone other than those two, BUT if you take the field, hey, that's a 1/8 chance (and if it is the field, then Kerk would be the best pick), so close to being not unexpected.  Hey, if you have 10 weight classes, there is something like a 75% chance that some "12% chance of happening" event ends up happening.  The question, though, is at what weight class will that happen?  Hard to predict (actually, not possible to predict).

In the end, although these types of probability calculations are fun to do and think about (for me, at least), it's really hard, in the end, to assess it or to say it was "wrong," even after the fact.

But it does give you fun things to think about and to put some hard values on.  For example, what is the most wide open weight class according to Pablo?  Well, if we draw a line and say that anyone with a 5% chance of winning is "in the mix" (see above - while there is a 75% chance that a 12% event happening for 10 weights, there is only a 40% chance of a 5% even happening), that means that for each weight we have

125 7 guys in the mix

133 5 guys in the mix

141 8 guys in the mix

149 4 guys in the mix

157 3 guys in the mix

165 2 guys in the mix

174 3 guys in the mix

184 2 guys in the mix

197 3 guys in the mix

285 3 guys in the mix

YMMV on where you draw the line of "in the mix" but basically, we see that 141 and 125 are the most wide open, while 165 and 184 the most dominated.

One thing I like about this exercise is that it is a way to evaluate those questions that people like to ask when the brackets come out.  The most fun one to do is the "who has the easiest path to the finals," which is what you always hear about the basketball brackets.  I've got a really good metric that I can use to do assessment, but it's a lot of work to program, and doing it for 10 weight classes is not something I feel like doing this year.  But I don't think that is as much a question for wrestling brackets.  More common questions are things like "Which is the low seed to look out for?"  I can do that, but need to think about it.

Maybe you need a "Snowball Line" (Snowball chance in hell)?
You may know the Mendoza Line in baseball, batting average below .200.
Maybe the Snowball Line should be .02 or .01 or whatever you think appropriate.   Wrestler X has less than.  .0x chance of winning, he's below the Snowball Line.

[ugarles]

Did you post projected team standings based on the revamped Pablo? Could you?

[Pablo]

8 hours ago, ugarles said:

Did you post projected team standings based on the revamped Pablo? Could you?

I really don't know how to do that.  I get the idea that you can use some estimates based on final placements, but I really don't even have that.  I can absolutely use advancement through the championship bracket, and it would be trivial to add in a probability for finishing 2nd, but I am not up for going into the consolation bracket to score either advancement or even places.  That would take a simulation effort that I don't want to get into.  If you look at the video I linked the other day, you can see that I've done these type of big-ass simulations, but it's a lot of effort.

At some point in my life, if I can keep this going, this is something that I could do (although if you ask, I could give you the information you could use to do it yourself!  Not much more is needed; if anyone wants to try it, DM me).  

[Gene Mills Fan]

20 minutes ago, Pablo said:

I really don't know how to do that.  I get the idea that you can use some estimates based on final placements, but I really don't even have that.  I can absolutely use advancement through the championship bracket, and it would be trivial to add in a probability for finishing 2nd, but I am not up for going into the consolation bracket to score either advancement or even places.  That would take a simulation effort that I don't want to get into.  If you look at the video I linked the other day, you can see that I've done these type of big-ass simulations, but it's a lot of effort.

At some point in my life, if I can keep this going, this is something that I could do (although if you ask, I could give you the information you could use to do it yourself!  Not much more is needed; if anyone wants to try it, DM me).  

cant you just add the ten teamates win percentages and compare it to the other teams totals and make the list.

Edited 5 hours ago by Gene Mills Fan

[ugarles]

18 minutes ago, Gene Mills Fan said:

cant you just add the ten teamates win percentages and compare it to the other teams totals and make the list.

hell, can't anyone? Using standard tournament scoring, assuming everything is chalk according to Pablo rating, nobody gets bonus points, and I only account for medal winners (no advancement points for anyone Pablo-ranked below top-8), here are the top 10:

Penn St.: 153.5
Iowa: 87
Ok. St.: 77.5
Nebraska: 70.5
Ohio St.: 60.5
Illinois: 48
Minnesota: 47.5
N. Iowa: 43.5
Cornell: 35
Mizzou: 33.5

I did this by hand, so the odds I screwed up a little are incalculable.

[Pablo]

19 minutes ago, ugarles said:

hell, can't anyone? Using standard tournament scoring, assuming everything is chalk according to Pablo rating, nobody gets bonus points, and I only account for medal winners (no advancement points for anyone Pablo-ranked below top-8), here are the top 10:

Penn St.: 153.5
Iowa: 87
Ok. St.: 77.5
Nebraska: 70.5
Ohio St.: 60.5
Illinois: 48
Minnesota: 47.5
N. Iowa: 43.5
Cornell: 35
Mizzou: 33.5

I did this by hand, so the odds I screwed up a little are incalculable.

Did you use the ratings then ("chalk") to get the places based on outcomes through the consolation brackets?  Or did you just call them 1 - 8?

The biggest worries I would have about this approach is

1) the lack of advancement points for non-place winners, and

2) matchup effects due to differences in seeding.

But it's a great start.  In particular, "assuming everything is chalk according to Pablo" is the best approach.  Because while the race is not always to the swiftest, nor the fight to the strongest, that's the way to bet. 

If I were to do this, of course, I'd make it a lot more complicated and do a bunch of simulated tournaments to see what kind of ranges those scores can adopt and the likelihood of each.

ETA:  I've also been thinking if there is a way to estimate the likelihood of bonus points given a ratings difference.

Edited 3 hours ago by Pablo

[ugarles]

It was meant to be back-of-a-napkin; all of the flaws you point out are ones I was aware of - that's why I put all the methodology (lol) up front. 

4 minutes ago, Pablo said:

Did you use the ratings then ("chalk") to get the places based on outcomes through the consolation brackets?  Or did you just call them 1 - 8?

This is the same thing unless you mean "did I award advancement points in the consolation bracket." I did.

I learned from reading another thread today that my calclations are already a little off because I gave an advancement point for winning the title match and that's apparently wrong.

My scoring was:
1: 5+16=21
2: 4+12=16
3: 3+1+10=14
4: 3+.5+9=12.5
5: 2+1+7=10
6: 2+.5+6=8.5
7: 2+1+4=7
8: 2+.5+3=5.5

I am not going to go back and fix it, but if medal rounds do not get advancement points, all champs would lose a point for team score and all 3/5/7 would lose .5 for team score.

[jchapman]

Anyone want to attempt to analyze Draft King betting odds for individual Championships, in light of the Pablo predictions?

Here are some that stood out to me as possible value bets:

125:  Ramos +330 (.361 to win)

141:  Hardy +370 (.307)

149:  Lovett +250 (.319)

157:  Shapiro +250 (.416)

165:  Caliendo +750 (.188)

174:  Hamiti +450 (.220)

197:  Barr +600 (.367), Buchanan +300 (.389)

 

Obviously Barr's health is a wildcard.

Edited 1 hour ago by jchapman

[Wrestleknownothing]

1 hour ago, ugarles said:

It was meant to be back-of-a-napkin; all of the flaws you point out are ones I was aware of - that's why I put all the methodology (lol) up front. 

This is the same thing unless you mean "did I award advancement points in the consolation bracket." I did.

I learned from reading another thread today that my calclations are already a little off because I gave an advancement point for winning the title match and that's apparently wrong.

My scoring was:
1: 5+16=21
2: 4+12=16
3: 3+1+10=14
4: 3+.5+9=12.5
5: 2+1+7=10
6: 2+.5+6=8.5
7: 2+1+4=7
8: 2+.5+3=5.5

I am not going to go back and fix it, but if medal rounds do not get advancement points, all champs would lose a point for team score and all 3/5/7 would lose .5 for team score.

Placement matches do not get advancement

[jchapman]

20 minutes ago, jchapman said:

Anyone want to attempt to analyze Draft King betting odds for individual Championships, in light of the Pablo predictions?

Here are some that stood out to me as possible value bets:

125:  Ramos +330 (.361 to win)  +330 is implied probability of .233

141:  Hardy +370 (.307)    IP = .213

149:  Lovett +250 (.319)  IP = .286

157:  Shapiro +250 (.416)  IP = .286

165:  Caliendo +750 (.188)  IP = .118

174:  Hamiti +450 (.220)  IP = .182

197:  Barr +600 (.367), Buchanan +300 (.389)  IP = .143, .250

 

Obviously Barr's health is a wildcard.