homeforumsrankingsprofilesahpavideosblogstips
rulesnewsphotosdownloadslinkscontact us
username
password
new user registration
forgot password?
air hockey chat forums
Forums Home | Log in for Private Messages | Search | View New Posts (Mark All Read) | User List
Forums Home / Tournaments and Challenges / Pedro seeding questions ( View Older Thread | View Newer Thread)

First | 1 | 2 | 3 | Last
- 10 Jun 2012
Total Posts: 658
Is pedro going to be seeded according to his current rating? It just seems so low. Just checking.
 
goran - 10 Jun 2012
Total Posts: 428
Yep, that reflects his current skill. We have match data against other known Venezuelans. All the players rated above him had a top 10 rank within the last year except George. 10 different players above him held a top 5 rank within the past year.
 
TWeissman - 10 Jun 2012
Total Posts: 210
His rating most definitely does NOT reflect his current skill level Goran.

I don't think it would be a good idea to seed him where he is on the ratings. You will basically be punishing someone by making them play against him early in the tournament.

Out of curiosity, how did he go from 1824 to 1630 between Dec 2 and Dec 6, 2011?
 
- 10 Jun 2012
Total Posts: 658
Hey, As long as I don't have to play him, I don't really care where he is seeded. ;)

Really though, for as good as the rating system is, I think this is one area that presents a negative.
 
Q - 10 Jun 2012
Total Posts: 497
TWeissman said:

Out of curiosity, how did he go from 1824 to 1630 between Dec 2 and Dec 6, 2011?


It was the challenge match loss to Daniel Ordonez that did that. http://airhockeyworld.com/match.asp?matchid=25639. A 4-3 loss at their rating separation at the time would be a 150-200 point change for the loser depending on the progression of the sets through the match and the full CM bonus.

I believe the reason it has his Old rating listed in that match as 1630 is an artifact of the "+1 rule" calculation step. The old rating in that column should still have been listed as 1824. Travis may be able to expand on this.
 
TWeissman - 10 Jun 2012
Total Posts: 210
goran said:
Yep, that reflects his current skill. We have match data against other known Venezuelans. All the players rated above him had a top 10 rank within the last year except George. 10 different players above him held a top 5 rank within the past year.


I was not aware that Jose, Wil, Jimmy and Evan all had top 10 ranks in the last year. Or, was your statement inaccurate?

The match data you have with Pedro against other Venezuelans is sparse, and he won almost all the matches. This doesn't tell us much. The two guys which are "known" quantities it looks like are Javier Pullido and Dionisio Diaz, and Pedro won all his posted matches against them, except a single 2 out of 3 game set against Javier (for which he lost only 4 rating points). I'm not sure how this gives us any real information about his skill. The real info we have is how he performed the last time he was actively competing in the US. And, based on that performance, he is far better than a 1649 rating.

In my opinion, if Pedro is seeded in the 20's, that will be a mistake and undermine the use of the ratings for seeding.
 
TWeissman - 10 Jun 2012
Total Posts: 210
Q said:
TWeissman said:

Out of curiosity, how did he go from 1824 to 1630 between Dec 2 and Dec 6, 2011?


It was the challenge match loss to Daniel Ordonez that did that. http://airhockeyworld.com/match.asp?matchid=25639. A 4-3 loss at their rating separation at the time would be a 150-200 point change for the loser depending on the progression of the sets through the match and the full CM bonus.



Holy shit, that's a major problem with this rating system, especially given the extreme lack of data in some circumstances. That is generalizing from an N of near 1...

ELO systems typically have max losses to prevent that sort of thing. The max win or loss is based on the weight of importance for the match.



 
goran - 10 Jun 2012
Total Posts: 428
If we were to seed with the ranking, and put Pedro at the bottom of the largest group of ultra pro, me, geoffroy, Robbins, Phil, that would put Pedro at the 27th seed. The ten players to hold top 5 are the current top 5 + Tim, Brian, August, Syed, Pete lippencott.
 
goran - 10 Jun 2012
Total Posts: 428
There is a Max loss based on the weight. Pedro lost a challenge match to an unknown player aka a beginner. Tim can you provide us with a list of how the tournament would be seeded using the traditional ranking method?
 
goran - 10 Jun 2012
Total Posts: 428
TWeissman said:
goran said:
Yep, that reflects his current skill. We have match data against other known Venezuelans. All the players rated above him had a top 10 rank within the last year except George. 10 different players above him held a top 5 rank within the past year.


I was not aware that Jose, Wil, Jimmy and Evan all had top 10 ranks in the last year. Or, was your statement inaccurate?

The match data you have with Pedro against other Venezuelans is sparse, and he won almost all the matches. This doesn't tell us much. The two guys which are "known" quantities it looks like are Javier Pullido and Dionisio Diaz, and Pedro won all his posted matches against them, except a single 2 out of 3 game set against Javier (for which he lost only 4 rating points). I'm not sure how this gives us any real information about his skill. The real info we have is how he performed the last time he was actively competing in the US. And, based on that performance, he is far better than a 1649 rating.

In my opinion, if Pedro is seeded in the 20's, that will be a mistake and undermine the use of the ratings for seeding.


Correct Tim, those players had no rank in the past year. The 10 players with a top 5 rank is true.
 
goran - 10 Jun 2012
Total Posts: 428
I have been playing air hockey for 10 years and never seen Pedro play. How does it work in other sports like tennis or golf when some one who was great ten years ago comes back, do they seed them at the top?
 
goran - 10 Jun 2012
Total Posts: 428
If ratings don't change, and everyone who says they're playing plays, Pedro #23 would play Anthony. #10, winner plays Nizzi #6, then Ehab#3. I don't think Pedro would beat any of those 3 right now. If he did, it would be an upset. Either way, it should be a fun fantasy draft this year.
 
- 10 Jun 2012
Total Posts: 658
Don't you mean he would play the number 7 seed, not 6? :)
 
goran - 10 Jun 2012
Total Posts: 428
marknizzi said:
Don't you mean he would play the number 7 seed, not 6? :)


Darn your right Nizzi, I wanted to see you play Pedro. I think Tim would be the 7th seed. I have been wrong before.
 
- 10 Jun 2012
Total Posts: 658
Whew!!.....close one. Sorry Tim! lol ;)
 
ajflanagan - 11 Jun 2012
Total Posts: 509
I think it's important to remember that the ratings have only been provisionally approved by the USAA for the two major tournaments this year. What happens after this is still open for debate and a vote. There has been extensive discussion on the private USAA forums about the ratings. Many changes and adjustments have been made in an effort to better the system. If anyone has an issue with the functionality of the system, now is the time to air it out. The private USAA forums are available to discuss the system with the ratings committee. Also, there is a USAA meeting coming up in July.

With that said, I completely agree with Goran's assessment of Pedro's position on the ratings and his subsequent seed. To give him any other seed going into the July event would be completely 100% subjective and negate the purpose of the system itself.

The Vegas event showed the ratings to work almost perfectly for seeds. Of course, there are going to be anomalies. There are also going to be upsets. In those odd cases, the tournament itself will work it out and ultimately improve the accuracy of the ratings even more. It's a constantly self-improving system.

One thing is certain... if Pedro should be rated higher than he will be going into the July event, he most certainly will be after the July event.

The beauty of this system is you know exactly where you stand. There is no hiding from it. If a player doesn't like their seed, there is plenty of time to get out there and play some matches.

Andrew
 
TWeissman - 11 Jun 2012
Total Posts: 210
So, what is the max point change in a match Goran?

It looks like Pedro lost 195 points in a single match.

How does that make sense?

Definitely, given these sorts of issues it will be clear after the tournament why the ratings in their current form should not be used.
 
Q - 11 Jun 2012
Total Posts: 497
TWeissman said:
So, what is the max point change in a match Goran?

It looks like Pedro lost 195 points in a single match.

How does that make sense?


Using the current table of values (they may have been different prior to the recent update to the ratings system when it was voted on by the USAA):

If both players have an active status: The maximum points lost for a challenge match (for this case both players <2000 rating) would be (32x4)+64 = 192.

If at least one player is establishing, the k-values have a 2x multiplier which put the maximum change at (64x4)+128 = 384.

The change from a challenge match is only different from 4 of 7 game rating sets in that there is the CM bonus calculation and the "+1 check" for the winner.
 
goran - 11 Jun 2012
Total Posts: 428
TWeissman said:
So, what is the max point change in a match Goran?

It looks like Pedro lost 195 points in a single match.

How does that make sense?

Definitely, given these sorts of issues it will be clear after the tournament why the ratings in their current form should not be used.


Where would the rankings seed Pedro?
Where would the rankings seed Tim, Javier, Anthony, Jose?
 
Q - 11 Jun 2012
Total Posts: 497
As a follow up to my previous post: The old k-value for a 4/7 game set was 30 and the old k-value of the CM bonus was 90. I believe these were the values back on 12/6/11.
 
DRAGO - 11 Jun 2012
Total Posts: 46
Seems like this thread has gone off course from tournament preview to seedings discussion and the the sham that is the MiBS. I will be posting my own tournament preview and report from the last tournament in the next few days.

goran said:
If we were to seed with the ranking, and put Pedro at the bottom of the largest group of ultra pro, me, geoffroy, Robbins, Phil, that would put Pedro at the 27th seed. The ten players to hold top 5 are the current top 5 + Tim, Brian, August, Syed, Pete lippencott.


Is this rule in effect when using the MiBS? If it is, it could solve the problem but only when used correctly.

NONE of the players listed above are Ultra-Pros! Not even close really. Phil and Mark are not even in the classification below Ultra Pros (Pro) anymore.

The Ultra Pro's in this tournament are: Billy, Danny, Rehab, Snowflake, Young Wolf, Brain, and Marino. Pedro should therefore be seeded around the 9-10 area applying the above rule. Trying to say Limp___ and The Sieve are UltraPros (meaning they are on the same level as Billy and Danny is beyond preposterous).

(On a side note Pedro shouldnt even have a MiBS rating to be discussing, he was 100% caught lying and manipulating the system before (never punished by USAA) by reporting false challenge match results. Who believes phatom matches supposedly taking place in a 3rd world country are legit? - but this is a topic for later)

goran said:
...ratings don't change, and everyone who says they're playing plays, Pedro #23 would play Anthony. #10, winner plays Nizzi #6, then Ehab#3. I don't think Pedro would beat any of those 3 right now. If he did, it would be an upset. Either way, it should be a fun fantasy draft this year.


Goran completely misses the point, and it is the same point missed regarding seeding by the majority of the brainiacs in the USAA. Marino and Pedro should not be meeting in the round of 32! A case can be made in this tournament (as the 8/9 or 7/10) that they are a legit round of 16 match-up, but 32 is insane.

It doesn't matter who wins, this incorrect seeding will RANDOMLY PUNISH a player not named Marino or Otero. The loser of this match will be playing someone in the loser's bracket and 100% of the time will win. That player should have a chance to win the match versus a mid 20's seeded player (someone like a Mike Rosen or Tad Gibson), instead they are randomly matched up agaisnt an Ultra Pro level player and therefore are in a lower spinoff than they may deserve.

CONT -
 
DRAGO - 11 Jun 2012
Total Posts: 46
- CONT

quoted]goran said:
...I have been playing air hockey for 10 years and never seen Pedro play. How does it work in other sports like tennis or golf when some one who was great ten years ago comes back, do they seed them at the top?


No one is saying seed him at the top. About 32 seeds matter and Pedro should be somewhere in the 8-12 seeds. So he is being asked to be seeded somewhere in the top 25%-40% - not "the top". Golf has no seedings (except one tournament a year). Ten years in Air Hockey currently is the equivalent to about 6 months in Tennis (Look at the AH rankings from ten years ago, same players). Tennis plays 50+ tournaments a year so seeding is not as vital and it balances out over time. The major championships have exceptions to their seeding rules (usually use world rankings which are the result of 50+ tournaments over the course of the previous year). Serena Williams among others has been seeded much higher than her current world ranking (which was low due to injuries and inactivity). So to answer Goran's question, Yes - other sports would allow for this.

The Solution is simple. The USAA cannot correct all the problems with it's rankings or the MiBS before this tournament. Apply the "last and largest" criteria to Pedro (and any other player grossly mis-seeded) after seeding is tabulated with the MiBS and make sure your classifications are correct and not politically correct. Have a knowledgeable third party apply the classifications and "L&L" criteria and not a member of the tournament staff or active player - in other words - ME. The USAA should assign me this task before the tournament to ensure the seeding process is as accurate as possible and therefore the finishes are as fair as they can be. I am not going to hold my breath...

---DJ
 
TWeissman - 11 Jun 2012
Total Posts: 210
The K-value should not allow a swing of 192, much less 384, points in Air Hockey matches, especially matches which are not under some form of scrutiny.

In ELO rating systems used all over the world, like Chess, etc. there is tons of data to support the use of K-values at levels that create point swings between 8 and 64.

"If the K-factor coefficient is set too large, there will be too much sensitivity to just a few, recent events, in terms of a large number of points exchanged in each game. Too low a K-value, and the sensitivity will be minimal, and the system will not respond quickly enough to changes in a player's actual level of performance." - From the Wiki page on ELO rating systems.

 
travis - 11 Jun 2012
Total Posts: 530
TWeissman said:
So, what is the max point change in a match Goran?

It looks like Pedro lost 195 points in a single match.

How does that make sense?


I'm looking into this, as it looks like this was Daniel Ordonez's first match. If so, Pedro's rating should not have been affected.


Travis Luscombe
AirHockeyWorld.com Webmaster
http://twitter.com/air_hockey
 
- 11 Jun 2012
Total Posts: 658
How far back does the mitic rating system cover Pedro's matches? Does it take into account any of his tournaments from 98-2001? If not, maybe in a situation like this, we can implement that information to boost the rating.
I can remember back in 2007, having the disapointment of playing Jimmy Heliander in the round of 32, after he'd been absent from the sport for 3 years. As a result his seeding had dropped pretty significantly for that tournament, being away for the time he had.
I just think this is one area that rating system needs some tweaking.

 
TWeissman - 11 Jun 2012
Total Posts: 210
travis said:
TWeissman said:
So, what is the max point change in a match Goran?

It looks like Pedro lost 195 points in a single match.

How does that make sense?


I'm looking into this, as it looks like this was Daniel Ordonez's first match. If so, Pedro's rating should not have been affected.




My point is that even if his rating should have been affected, the swing is way too large.

 
goran - 11 Jun 2012
Total Posts: 428
Nizzi shouldn't be playing Jimmy in the round of 32. That tournament wasn't seed right. He should have been seeded 8-12. 9 sounds fair. We should just let Don seed the tournament, then maybe he'll start playing again.
 
TWeissman - 11 Jun 2012
Total Posts: 210
marknizzi said:
How far back does the mitic rating system cover Pedro's matches? Does it take into account any of his tournaments from 98-2001? If not, maybe in a situation like this, we can implement that information to boost the rating.
I can remember back in 2007, having the disapointment of playing Jimmy Heliander in the round of 32, after he'd been absent from the sport for 3 years. As a result his seeding had dropped pretty significantly for that tournament, being away for the time he had.
I just think this is one area that rating system needs some tweaking.




Niz,

Just click on Pedro's link in the ratings and you can see all the matches that are taken in to account.

Another interesting thing to do is to click on the Ratings Calculator and put in yourself and someone else and see how this rating system looks on the inside. There are a number of issues that need to be fixed, #1 of which is the so-called +1 rule, whereby if a lower rated player wins a match, they are given a new rating which is 1 point more than their opponent. This bastardizes a rating system and basically makes it no better than a ranking system. It also allows for point creation out of thin air...

For example, put your name and Jacob's name in the calculator. Imagine that Jacob wins a match against you 4-3 because you get injured in the middle and have to forfeit. Jacob shoots up near 700 points in rating. Then, you can just play Jacob again a month later when you are healed, win, and get your rating back, leaving Jacob still up there in the 2000's.

Don't get me wrong. I MUCH PREFER a rating system to what we have used for all these years. But, I want it to be a rating system modeled after ELO systems used world wide - not a ranking/rating hybrid which takes the worst of the two and meshes them together.

 
- 11 Jun 2012
Total Posts: 658
TWeissman said:
marknizzi said:
How far back does the mitic rating system cover Pedro's matches? Does it take into account any of his tournaments from 98-2001? If not, maybe in a situation like this, we can implement that information to boost the rating.
I can remember back in 2007, having the disapointment of playing Jimmy Heliander in the round of 32, after he'd been absent from the sport for 3 years. As a result his seeding had dropped pretty significantly for that tournament, being away for the time he had.
I just think this is one area that rating system needs some tweaking.




Niz,

Just click on Pedro's link in the ratings and you can see all the matches that are taken in to account.

Another interesting thing to do is to click on the Ratings Calculator and put in yourself and someone else and see how this rating system looks on the inside. There are a number of issues that need to be fixed, #1 of which is the so-called +1 rule, whereby if a lower rated player wins a match, they are given a new rating which is 1 point more than their opponent. This bastardizes a rating system and basically makes it no better than a ranking system. It also allows for point creation out of thin air...

For example, put your name and Jacob's name in the calculator. Imagine that Jacob wins a match against you 4-3 because you get injured in the middle and have to forfeit. Jacob shoots up near 700 points in rating. Then, you can just play Jacob again a month later when you are healed, win, and get your rating back, leaving Jacob still up there in the 2000's.

Don't get me wrong. I MUCH PREFER a rating system to what we have used for all these years. But, I want it to be a rating system modeled after ELO systems used world wide - not a ranking/rating hybrid which takes the worst of the two and meshes them together.




Yes, i agree, and also feel this is why I.M.O., Pedro's other matches from 98-2001 need to be accounted for somehow.
Good point about the +1 rule and Jacob scenario. And it probably doesn't take much imagination to see me getting injured and forfeiting, huh? ;)
 
travis - 11 Jun 2012
Total Posts: 530
I think one thing that should be considered as a future enhancement to the ratings is to have a distinction between the brand new players and the ones that used to play but are then deactivated. Inactive players' ratings should not be affected if they play a brand new player. So it would look like:

Three statuses: New, Active, Inactive (used to be active)

Active players' ratings are affected by playing against active players
Inactive players' ratings are affected by playing against active and inactive players
New players' ratings are affected by playing against new, active and inactive players

This would correct the scenario w/ Pedro, since he was an inactive player playing against a new player.

Travis Luscombe
AirHockeyWorld.com Webmaster
http://twitter.com/air_hockey
 

First | 1 | 2 | 3 | Last

Forums Home / Tournaments and Challenges / Pedro seeding questions