John t whelan ranking simulator

[FlagDUDE08]

Re: John t whelan ranking simulator

I have a headache.

So much for Simple Math (tm).

If they're going to call it RPI, we might as well own up to it.

Re: John t whelan ranking simulator

Very true. I'm still trying to get some of these figured out. I wonder how USCHO is calculating the weighting, because I'm trying to get to RPI's RatingsPI according to USCHO (even off by a factor of any multiple of .0025 to account for a mistake in the quality wins).; using it because they've both won and lost games, and there's no adjustment for negative effect.

I think USCHO is calculating the weighting wrong. The OWP and OOWP deifnitely add up to .42375, and you have to get to something ending in 81, 31, 56, or 06 to account for quality wins as RPI is listed as .6581; this means you must shoot for something ending in 44, 94, 19, or 69 (while accounting for some rounding). What USCHO did to get there was that they counted away games as 1.2 and home game as 0.8, regardless of result. This is NOT the case, according to how RatingsPI is handled for weighting. http://rpiratings.com/WhatisRPI.php If you win away or lose at home, the weight of 1.2 is used. If you lose away or win at home, the weight of 0.8 is used. Going back to my example of RPI... If you do it according to the site I cited, it is 3.2/4.0, or .2000. If you count the Sacred Heart game in Bridgeport as neutral site, it's 3/3.8, or .19736 points. The way USCHO did it is 3.2/4.4, or .18181.

I have a headache.

So much for Simple Math (tm).

[FlagDUDE08]

Re: John t whelan ranking simulator

I tried brandishing a PWR probability thing. Got all sorts of things wrong that I was doing daily corrections.

You can use all the words one wants to use, but something always pops up when you're working with plain language.

Very true. I'm still trying to get some of these figured out. I wonder how USCHO is calculating the weighting, because I'm trying to get to RPI's RatingsPI according to USCHO (even off by a factor of any multiple of .0025 to account for a mistake in the quality wins).; using it because they've both won and lost games, and there's no adjustment for negative effect.

I think USCHO is calculating the weighting wrong. The OWP and OOWP deifnitely add up to .42375, and you have to get to something ending in 81, 31, 56, or 06 to account for quality wins as RPI is listed as .6581; this means you must shoot for something ending in 44, 94, 19, or 69 (while accounting for some rounding). What USCHO did to get there was that they counted away games as 1.2 and home game as 0.8, regardless of result. This is NOT the case, according to how RatingsPI is handled for weighting. http://rpiratings.com/WhatisRPI.php If you win away or lose at home, the weight of 1.2 is used. If you lose away or win at home, the weight of 0.8 is used. Going back to my example of RPI... If you do it according to the site I cited, it is 3.2/4.0, or .2000. If you count the Sacred Heart game in Bridgeport as neutral site, it's 3/3.8, or .19736 points. The way USCHO did it (including Sacred Heart in Bridgeport as away) is 3.2/4.4, or .18181. Looking at the rules for neutral site as of 2007-08, if Atlantic Hockey refs were provided and SHU did all the hosting duties, that should be an away game for RPI.

[Patman]

Re: John t whelan ranking simulator

I noticed a difference between USCHO's and my calculations for RatingsPI, and could be a big difference as to why some of these numbers seem messed up. As my example, I'd like you to take a look at AIC's rating. I have them listed as 57.00, while USCHO has them listed as 75.00. Let's break it down.

WP: AIC has played one game vs. Providence and lost. Doesn't really matter where it is, that will be 0.00.
OWP: Providence has played two other games, each against Mankato, both of which Providence won. That will be 21.00.
OOWP: All we need to do is calculate Providence's OWP, and that's two parts Mankato, one part AIC. Where USCHO and I differ is that USCHO does not take opponents' games against AIC into account when calculating AIC's OOWP. There is some conflicting information, as the definition for OOWP I have, although sourced as being related to basketball so I don't know if the definition holds for college hockey, is that the team's OOWP includes games against the original team. http://en.wikipedia.org/wiki/Rating_Percentage_Index Mankato won all of their other games thus far. If you include AIC, it becomes 36.00, while if you do not include them, it becomes 54.00.

Of course, now I see information at http://rpiratings.com/WhatisRPI.php that you don't include games against the original team. Hello, confusion!

I tried brandishing a PWR probability thing. Got all sorts of things wrong that I was doing daily corrections.

You can use all the words one wants to use, but something always pops up when you're working with plain language.

[FlagDUDE08]

Re: John t whelan ranking simulator

I noticed a difference between USCHO's and my calculations for RatingsPI, and could be a big difference as to why some of these numbers seem messed up. As my example, I'd like you to take a look at AIC's rating. I have them listed as 57.00, while USCHO has them listed as 75.00. Let's break it down.

WP: AIC has played one game vs. Providence and lost. Doesn't really matter where it is, that will be 0.00.
OWP: Providence has played two other games, each against Mankato, both of which Providence won. That will be 21.00.
OOWP: All we need to do is calculate Providence's OWP, and that's two parts Mankato, one part AIC. Where USCHO and I differ is that USCHO does not take opponents' games against AIC into account when calculating AIC's OOWP. There is some conflicting information, as the definition for OOWP I have, although sourced as being related to basketball so I don't know if the definition holds for college hockey, is that the team's OOWP includes games against the original team. http://en.wikipedia.org/wiki/Rating_Percentage_Index Mankato won all of their other games thus far. If you include AIC, it becomes 36.00, while if you do not include them, it becomes 54.00.

Of course, now I see information at http://rpiratings.com/WhatisRPI.php that you don't include games against the original team. Hello, confusion!

[Patman]

Thanks for downloading and I'm glad you enjoy it thus far. Conference tournaments is something I will consider putting in automatically, but for the time being, the games would need to be entered manually. It would require me to do some reformulation including with the data file because I do not include in what league each team plays, nor do I give any consideration to a shootout, which is used in calculating the league points for several leagues. It is something that I am willing to work on, as one of my goals was to actually run through scenarios automatically.

Obviously this is sort of a work in progress, but given the absence of another tool out there, I have decided to introduce this tool to all of you.

The main reason I haven't even tried to broach it is dealing with tie-breakers. I'm a math guy, I'm not trained in algorithms and I can get lost in indexing

Even in season tournaments are easier to handle, in my opinion

Re: John t whelan ranking simulator

Actually, the CSV file is not game based, but rather team-based. The basis for the file was the spreadsheet that I did during the off-season for all the teams' schedules. The easiest way to insert a game at this point is to use the "Add Game" button on the application and then save the data file. As with the spreadsheet, though, you'll need to ensure that the team names are correct.

Even easier

[FlagDUDE08]

Re: John t whelan ranking simulator

It's a .csv file so it would mean learning the string used to create a game and plugging it in (same as we did with the tool at slack.net)

Actually, the CSV file is not game based, but rather team-based. The basis for the file was the spreadsheet that I did during the off-season for all the teams' schedules. The easiest way to insert a game at this point is to use the "Add Game" button on the application and then save the data file. As with the spreadsheet, though, you'll need to ensure that the team names are correct.

[FlagDUDE08]

Re: John t whelan ranking simulator

Thanks FlagDude,
I got it to download. Looks like a good tool. And, that you are going to update with other games is good, too. If you ever used Whelan's tool, you know that the great benefit of it is that for the conference tourneys, you can enter any game combination you want. In other words, you could play the conf tourneys multiple times with your own predicted results, a full week or 2 ahead of the last weekend. This allowed you to find things like, "If this series goes 3 games, then Mankato loses the compare with No Mich...." or things like that. Do you have capability for this in your program? Another good piece would be the ability to see just which compares are being won. As you know, some get very tight at the end of the year, and some schools' chances depend on a particular set of results, which can be found if you know where to look.

Thanks for the work. Those of us who use it appreciate it.

Numbers

Thanks for downloading and I'm glad you enjoy it thus far. Conference tournaments is something I will consider putting in automatically, but for the time being, the games would need to be entered manually. It would require me to do some reformulation including with the data file because I do not include in what league each team plays, nor do I give any consideration to a shootout, which is used in calculating the league points for several leagues. It is something that I am willing to work on, as one of my goals was to actually run through scenarios automatically.

Obviously this is sort of a work in progress, but given the absence of another tool out there, I have decided to introduce this tool to all of you.

[FlagDUDE08]

Re: John t whelan ranking simulator

I think I may have where my formula is confused, or at least one portion, and that is the calculation of a tie. Obviously for a win it is 0.8 out of 0.8 if home and 1.2 out of 1.2 if away; a loss is 0 out of 0.8 if away and 0 out of 1.2 if home, and I know I need to change these, but how would I quantify a tie? Would it be:

Away: 1.2 + 0 / (1.2 + 0.8) / 2
Home: 0.8 + 0 / (0.8 + 1.2) / 2

Thanks FlagDude,
I got it to download. Looks like a good tool. And, that you are going to update with other games is good, too. If you ever used Whelan's tool, you know that the great benefit of it is that for the conference tourneys, you can enter any game combination you want. In other words, you could play the conf tourneys multiple times with your own predicted results, a full week or 2 ahead of the last weekend. This allowed you to find things like, "If this series goes 3 games, then Mankato loses the compare with No Mich...." or things like that. Do you have capability for this in your program? Another good piece would be the ability to see just which compares are being won. As you know, some get very tight at the end of the year, and some schools' chances depend on a particular set of results, which can be found if you know where to look.

Thanks for the work. Those of us who use it appreciate it.

Numbers

It's a .csv file so it would mean learning the string used to create a game and plugging it in (same as we did with the tool at slack.net)

[Numbers]

Re: John t whelan ranking simulator

Thanks FlagDude,
I got it to download. Looks like a good tool. And, that you are going to update with other games is good, too. If you ever used Whelan's tool, you know that the great benefit of it is that for the conference tourneys, you can enter any game combination you want. In other words, you could play the conf tourneys multiple times with your own predicted results, a full week or 2 ahead of the last weekend. This allowed you to find things like, "If this series goes 3 games, then Mankato loses the compare with No Mich...." or things like that. Do you have capability for this in your program? Another good piece would be the ability to see just which compares are being won. As you know, some get very tight at the end of the year, and some schools' chances depend on a particular set of results, which can be found if you know where to look.

Thanks for the work. Those of us who use it appreciate it.

Numbers

[FlagDUDE08]

Re: John t whelan ranking simulator

Thanks flag dude. I haven't yet downloaded, and probably won't until January. No need until then - Priceless starts his yearly reports of who has fallen in or out of the field in past years about that time, and until then the different pieces of the PWR are too volatile.

My question for you: If I download this Java and the associated file with the results, does the file automatically sync as you update, or do I need to re-download every weekend?

And, I am running XP Home and Chrome for a browser, and I can't find a way to download the data file. Look at it, yes. Download, no. Any help?

Thanks.

Thank you for the download; I really do appreciate it.

I do not have automatic sync built in as of yet. For now, you will need to either re-download the file or fill in the games as you go. If you select a game in the games list, you have the ability to put in there who won the game, and you may also add, modify, or delete games as necessary (something I will do for tournaments). So long as the game is listed as "No Contest", it will not be included in the calculation. I will, however, look into programming a feature that will allow you to download the file from my website.

As for downloading the data file, is your browser perhaps opening the file internally? If so, you should be able to right click anywhere on the page and select "Save As".

Once again, thank you for the download. Without all of you on the USCHO Forums, I wouldn't have a reason to put this together.

[Patman]

Re: John t whelan ranking simulator

Other than checking each other's programs, the pre-season is completely worthless. This is why I prefer a predictive approach, no matter how weak the results are.

[FlagDUDE08]

Re: John t whelan ranking simulator

Something seems to be amiss with someone's calculation, whether it be mine, USCHO's, or someone else, and I think I know where it is. I'm going to take the Northeastern example given earlier.

Northeastern has won four games: ALH, ALH, HLC, HLC. Three were home, one (HLC) was away, meaning out of a possible 3.6 points, they have gained 3.6. Divide into it for 1.00, and multiply by the factor of 25 for 25.00.

Let's now take a look at the teams they have played:
Alabama Huntsville (ALH) has not yet had any other opponents, therefore the Opponent's Winning Percentage (OWP) part from that team is 0 out of a possible 0, meaning it is indeterminate because of the existence of 0/0. Remember that the OWP parts do NOT count the two particular teams in question, i.e. any games between ALH and NE are thrown out.
Holy Cross (HLC) has had one opponent in Boston University, and BU won. Therefore, it is 0 out of a possible 1, or 0.00. Remember, weights do not apply to OWP or OOWP.
I add all of this up for indeterminate + indeterminate + 0.00, divide by the number of games which is 3, and multiply by the factor of 21.

To take a team's Opponents of Opponents Winning Percentage (OOWP), for each of the team's games, you must at each part take your opponent's OWP.
For ALH, it's two parts Northeastern. With Northeastern, they have played Holy Cross twice and won, making 2 out of a possible 2. Therefore, ALH's OWP is 1.000.
For HLC, it's one part BU, two parts Northeastern. With BU, they have had three other games at home, and have won two (beat UMA and WIS, lost to RPI) for 0.667. As we saw earlier with Northeastern, take the games with Alabama Huntsville, 2 out of 2. I add this all up for 0.667 + 1 + 1, divide by the number of games which is 3, for 0.889.
Now, let's put the OOWP together: ALH 1.000 + ALH 1.000 + HLC 0.889 + HLC 0.889 = 3.778, divide by four games for a total of 0.9445, multiply by a factor of 54 for 51.00.

Therefore, BEFORE the quality wins bonus, the RatingsPI of Northeastern I have at 76.00. I don't see why they would remove any of the games from Northeastern's consideration. MAYBE remove Holy Cross games because of the dip in OOWP. I'll check the formula I use and adjust it accordingly. If USCHO did that, then USCHO probably had the divide by zero issue I spoke of earlier, and so they removed the ALH games as well. We'll have a better calculation come November 1st.