New Dual Rankings process is implemented

[andegre]

A month ago, or so, I talked about implementing a "prediction engine", which pits two wrestlers against each other, and spits out who the engine thinks will win in a matchup (returning the projected points for each wrestler).

I went a step further and introduced this methodology to the Dual Rankings process.

So, for every team, I took the starters at each weight and ran them through the prediction engine. Save the results for each match so I could get the sum of the points, like if it was an actual dual. That's then used to determine the winner for that dual. Rinse-repeat for each school against every other school.

This process takes A LONG time, but I feel this gives the MOST accurate representation of a teams' "dual strength".

tl;dr

If a team does not have a starter set at the corresponding weight, then they would forfeit that match and award the other team the 6 points.

If two teams tie in a dual, then it defaults to the traditional team ranking as the tie-breaker.

Lastly, if two teams tie for ALL of their matches (ie both teams went 55-20), then again, the old-style team rankings determine who is ranked higher.

Check out the results here: http://testwrestlestat.azurewebsites.net/rankings/dual

 

[RoarLions1]

Interesting andegre. It'll mean more once all the line-ups are set, and freshmen rankings settle a little. Good work taking this tool to a new level...what's next??

As an aside, me and a friend tried to boil dual meet prognostication to a quick-and-dirty mathematical model years ago...but failed miserably.

 

[andegre]

Interesting andegre. It'll mean more once all the line-ups are set, and freshmen rankings settle a little. Good work taking this tool to a new level...what's next??

As an aside, me and a friend tried to boil dual meet prognostication to a quick-and-dirty mathematical model years ago...but failed miserably.

What's next? Good question, still a few ideas that I have, but the main thing I'm trying to get done this summer is a "to do" and "approval" process. I'm going to open up the site to select users (that request and I fell comfortable with) to help me take care of maintenance type work. Including merging duplicate wrestlers, dual result entries, tournament entries, feature match selection, etc. The big part that is needed is the approval process. When X user changes the starter at 285 for Penn State to Jan Johnson (because they want the rankings to drop as a result), the approval process/queue will let these select users go and "confirm" that change. Approve or deny, and move on. Right now, about 50% of this stuff (including changing starters, deleting wrestlers, etc) just sends me an email. If I think it's wrong, I have to go back through and do the changes manually. Witht he approval process, you'll be able to just click approve or deny and the system will automatically accept or revert the change that just happened.

 

[andegre]

Interesting andegre. It'll mean more once all the line-ups are set, and freshmen rankings settle a little. Good work taking this tool to a new level...what's next??

As an aside, me and a friend tried to boil dual meet prognostication to a quick-and-dirty mathematical model years ago...but failed miserably.

Oh, and yes. It will definitely be more accurate as we get match results for these younger wrestlers. For example, on the The Open Mat forums, SetonHallPirate asked who Penn State's loss was to. I manually ran PSU thru the prediction engine to find out it was to Iowa. I then wanted to know how, so I ran each starter through the engine, the main match that tipped the scale was Gilman vs Suriano for the sole reason that Suriano is a true freshman having no match results, and thus the default ranking. Looking at Gilmans history, he usually smokes guys in the 1200 rating range, which is why it predicted a fall.

Here's a link to that thread with the breakdown for PSU vs Iowa through the prediction engine: http://forum.theopenmat.com/index.php/topic,34857.0.html

 

[tomsc]

Oh, and yes. It will definitely be more accurate as we get match results for these younger wrestlers. For example, on the The Open Mat forums, SetonHallPirate asked who Penn State's loss was to. I manually ran PSU thru the prediction engine to find out it was to Iowa. I then wanted to know how, so I ran each starter through the engine, the main match that tipped the scale was Gilman vs Suriano for the sole reason that Suriano is a true freshman having no match results, and thus the default ranking. Looking at Gilmans history, he usually smokes guys in the 1200 rating range, which is why it predicted a fall.

Here's a link to that thread with the breakdown for PSU vs Iowa through the prediction engine: http://forum.theopenmat.com/index.php/topic,34857.0.html

Andegre, you asked for corrections. I see you still have Beitz on the PSU roster. He's no longer on the team, having graduated per my understanding. I don't think he's transferred as a grad somewhere else, but someone could correct me on that.

 

[andegre]

Andegre, you asked for corrections. I see you still have Beitz on the PSU roster. He's no longer on the team, having graduated per my understanding. I don't think he's transferred as a grad somewhere else, but someone could correct me on that.

Thanks, he's been removed.

 

[andegre]

Andegre, you asked for corrections. I see you still have Beitz on the PSU roster. He's no longer on the team, having graduated per my understanding. I don't think he's transferred as a grad somewhere else, but someone could correct me on that.

By the way, in regards to my response to Roar up above about "what's next", this is one of the things I'm talking about.

For this scenario, I would have two different options:
1) Allow any registered user to delete the wrestler from the roster, which would then get added directly to the "approval queue" for someone to confirm or deny
or
2) Instead of allowing the delete, have a button where the user would click "flag as not on roster anymore". Which would then add this to the "to-do" queue, which would then get worked by someone. Then after it was worked, it would move to the approval queue for someone else to confirm.

 

[RoarLions1]

I tip my hat to ya, andegre...

 

[tomsc]

I tip my hat to ya, andegre...

Yes, thanks for doing this and making it easy for us to access it. I know you got a little grief last year for posting the links to your website on the wrestling forum, but I for one enjoy the easy access to it.

 

[burnsie1983]

Andegre, you asked for corrections. I see you still have Beitz on the PSU roster. He's no longer on the team, having graduated per my understanding. I don't think he's transferred as a grad somewhere else, but someone could correct me on that.

What's next? Good question, still a few ideas that I have, but the main thing I'm trying to get done this summer is a "to do" and "approval" process. I'm going to open up the site to select users (that request and I fell comfortable with) to help me take care of maintenance type work. Including merging duplicate wrestlers, dual result entries, tournament entries, feature match selection, etc. The big part that is needed is the approval process. When X user changes the starter at 285 for Penn State to Jan Johnson (because they want the rankings to drop as a result), the approval process/queue will let these select users go and "confirm" that change. Approve or deny, and move on. Right now, about 50% of this stuff (including changing starters, deleting wrestlers, etc) just sends me an email. If I think it's wrong, I have to go back through and do the changes manually. Witht he approval process, you'll be able to just click approve or deny and the system will automatically accept or revert the change that just happened.

If you are still looking for folks for the approval process, I'd be happy to lend a hand. Love your site.

 

[andegre]

If you are still looking for folks for the approval process, I'd be happy to lend a hand. Love your site.

Thank you, I'll make a note. It probably will be closer to mid to late summer before it's ready....and I'll have to go through a testing phase (with your help and others) before I turn it on in prod.