John t whelan ranking simulator

[JimDahl]

Re: John t whelan ranking simulator

It looks like they match now; must have caught one of us between updates.

http://pwr.reillyhamilton.com/pwr.php will pull/scrape updated results; I never implemented a caching layer and the .html page has fallen wildly out of date.

Thanks. I should have mentioned, I actually fixed a problem on my end in the interim. USCHO incorrectly has Bentley v. Holy Cross listed as a home game instead of neutral, so may calculations for those two were off by a smidge. Easy to find when it's two teams that are off, but a reminder of how painful it's going to be to care about location. As discussed earlier in this thread, I clearly need to write a parser for CHS and begin to treat it as authoritative.

 

[FlagDUDE08]

Re: John t whelan ranking simulator

Thanks. I should have mentioned, I actually fixed a problem on my end in the interim. USCHO incorrectly has Bentley v. Holy Cross listed as a home game instead of neutral, so may calculations for those two were off by a smidge. Easy to find when it's two teams that are off, but a reminder of how painful it's going to be to care about location. As discussed earlier in this thread, I clearly need to write a parser for CHS and begin to treat it as authoritative.

awk FTW

 

[Shirtless Guy]

Re: John t whelan ranking simulator

This seems like the right thread to ask. I just started looking at KRACH today because gfmorris has been using it to predict WCHA results and noticed that USCHO, CHN and SiouxSports all claim to be showing KRACH values for teams but they all have different values?

 

[JimDahl]

Re: John t whelan ranking simulator

This seems like the right thread to ask. I just started looking at KRACH today because gfmorris has been using it to predict WCHA results and noticed that USCHO, CHN and SiouxSports all claim to be showing KRACH values for teams but they all have different values?

Based on my quick spot checks, the rankings are all the same and the ratios between the ratings are all the same. The actual rankings in KRACH are arbitrary, it's the ratio between two rankings that predicts the likelihood of one team beating another.

 

[Shirtless Guy]

Re: John t whelan ranking simulator

Based on my quick spot checks, the rankings are all the same and the ratios between the ratings are all the same. The actual rankings in KRACH are arbitrary, it's the ratio between two rankings that predicts the likelihood of one team beating another.

Clarkson and Yale...chn and siouxsports are insignificantly different but there are multiple teams with a different ranking for uscho compared to chn and siouxsports.

 

[Patman]

Re: John t whelan ranking simulator

Clarkson and Yale...chn and siouxsports are insignificantly different but there are multiple teams with a different ranking for uscho compared to chn and siouxsports.

are there teams with the same value? There are two possibilities... missing game(s). A few substantial differences with some minor but existing deviations. Other possibility, iterations and round-off error... unlikely at this point but if so every team would have small deviations.

Otherwise I'd start to see if they're accidentally putting in weighted games into KRACH.

 

[Numbers]

Re: John t whelan ranking simulator

Clarkson and Yale...chn and siouxsports are insignificantly different but there are multiple teams with a different ranking for uscho compared to chn and siouxsports.

Found this discussion interesting also. I went back and did a check. Unless I can't see what I am missing, uscho and chn show exactly the same values. The ratios between teams in uscho/chn compared to siouxsports are close to with a few parts in 10000.

What I am expecting happened here was that ShirtlessGuy was looking while all 3 sites were processing results from tonight's Beanpot games. One site may have had all the results, the others not yet. It seems now, with all results in, that the numbers for KRACH are the same (up to a constant multiplier, of course. Which multiplier is 1.545)