The GAA Battle

[Beeeej]

I just sent a brief note to USCHO's Stats department.  I'm probably not only wasting my time, but obsessing a bit much.  :-)

I understand that GAA is typically only *represented* in stats to the second decimal place.  But the fact that David McKee's GAA *rounds up* to 1.43 (1.427363078875) and Matti Kaltiainen's GAA *rounds down* to 1.43 (1.43060319585) tells me something's wrong with the way you're listing them on the stats page.  It's one thing to *list* them both has having a 1.43 GAA, that I understand.  I simply dispute whether it's accurate to list them as being *tied* by virtue of those "matching" 1.43 GAAs.

Beeeej

[Greg Berge]

Yes.  You are obsessing too much.  And this coming from me.

[Beeeej]

I'm a black kettle!!

Thanks, pot.  ;-)

Beeeej

[DeltaOne81]

To get an idea, McKee's GAA would be exactly equal (to several decimal points) to Kaltiainen's had McKee played 2:51 less, or had Kaltianianen played 1:49 seconds more. So we're talking a margin of error of less than what a goalie could be pulled for delayed penalties and extra man attempts. While the numbers are big enough (i.e. not a few seconds) to argue that McKee may indeed have played a tiny, tiny bit better, I also can't blame them for calling it a tie.

[Beeeej]

I can't, either - in fact it's probably an algorithm rather than any conscious choice on USCHO's part.  I don't have to like it, though.  :-{)}

Beeeej

[andyw2100]

[Q]So we're talking a margin of error of less than what a goalie could be pulled for delayed penalties and extra man attempts. [/Q]

Does the time the goalie is off the ice for delayed penalties really come out of his time played stats? (I'm sure you're right that it does...I just never realized that it did.)

Thanks.
                                  Andy W.

[jtwcornell91]

[Q]andyw2100 Wrote:

Does the time the goalie is off the ice for delayed penalties really come out of his time played stats? (I'm sure you're right that it does...I just never realized that it did.)[/q]

I could have sworn it didn't, but I couldn't find anything about GAA in the NCAA rules.

[French Rage]

Hasnt Matti only played half the total time McKee has?

[Avash]

[Q]French Rage Wrote:

 Hasnt Matti only played half the total time McKee has?[/q]

David has played 1261:04, and Matti (who has been sharing goaltending duties with Cory Schneider, who happens to be 4th in the country in GAA) has played 796:52

The Kaltiainen/Schneider duo is having the same type of success that, recently, Howard/Doyle and Underhill/LeNeveu had.

[billhoward]

Yes, it's obsessive.  But does one more decimal place really make a difference, on GAA or save pecentage? As others noted, for GAA the rounding difference amounts to to perhaps 5 minutes played more or less if you carry the season out to three deccimal places.

Remember the term JND, or just noticeable difference? I think if you carry it out one more decimal place, you've gotten beyond what's noticeable.

With save percentage, a third decimal place implies precision that isn't matched in reality because the stat-keepers are going to miss the odd save here or there, or give credit to a shot gloved just outside the cage. McKee has 432 saves on 462 shots allegedly on goal, or .935. (actually .9351 so it rounds legitimately to .94). But if the scorekeeper misses one save per game, which wouldn't be hard, it's 412/442=.932, and if the scorekeeper was overly optimistic and counted one shot each game as a save when it was outside the cage, it would be be 452/482=.938, still rounding to .94. So there's a danger in thinking that because you can move the decimal places out, you should.

I'd say if two goalies finish the season with a .95 save average or a 1.40 GAA average, it's a tie, regardless of whether they're really .951 and .952 on saves, or 1.399 and 1.401 on goals. It's a tie.

Maybe we're on this thread because game clocks now count the final minute in tenths. Really, it should only be the final ten seconds in tenths, which divides the final ten seconds alone into 100 components. On the court for the players, coaches, and fans, it makes sense to go to tenths at 1:00 remaining and not :10 remaining because if you glance up and see a 1 and a 5, you might not know if it's 15 or 1.5 seconds remaining.

But are sports that granular as the final minute in 600 segments? I think there should be a journalist's code of ethics (oxymoron, right, like legal ethics?) that says you only cite the tenths in the last five seconds. The fact that morons on TV do it doesn't mean the print media with hopefully more brains if not such whitened teeth should follow suit. Bill Russell was great expounding on whether the clock had a long or short two seconds left in the fourth quarter. Other than that, there isn't a JND.

Let's get back to something more useful, like Cornell's chances of getting a #1 seed.

[DeltaOne81]

wrong thread

[jeh25]

[Q]billhoward Wrote:

 
Remember the term JND, or just noticeable difference? I think if you carry it out one more decimal place, you've gotten beyond what's noticeable.

[/q]

Back to intro psych for you, Bill. :p

I may not play one on TV, but I am in fact a psychophysicist in real life, and you've utterly abused a technically precise piece of jargon here. Weber is rolling in his grave as we speak.  

The concept I think you are looking for here is one of significant digits, not simulus intensity.

Pedantically yours.

jh

[jkahn]

[Q]billhoward wrote:
I'd say if two goalies finish the season with a .95 save average or a 1.40 GAA average, it's a tie, regardless of whether they're really .951 and .952 on saves, or 1.399 and 1.401 on goals. It's a tie.[/Q]
So, .9451 and .9549 would be a tie (though separated by .0098), but .9449 and .9451 would produce a clear winner?  This makes no sense at all.  Whoever has the best save percentage (or GAA), no matter how many decimal places it takes, is the person who has the best percentage.

[billhoward]

One can carry the math out to four, five, six decimal places to produce a goaltending number that's just slightly different from another ... but the margin of error is higher (for save percentage certainly) because of the vagaries of stat-keeping.

That's why three decimal places of precision for GAA is enough (1.xx) and two for save percentage (0.xx) where it's possible to be off by several shots a game on shots that may or may not have been saves.

Ditto for, say, points and assists per game. Does the ref always get that second assist right?

[cornelldavy]

This reminds me of the American League batting title race of 1910 (who could forget it?) between Ty Cobb and Nap Lajoie. The winner of the batting crown was to receive a new car, and because Cobb was so hated, when Lajoie played in St. Louis on the last day of the season, the Browns' manager ordered his third baseman to play on the outfield grass to allow Lajoie to bunt for hits all day long in order to pass Cobb. Lajoie went 8 for 9, but even with the help, he lost the race to Cobb with batting averages of .384944 to .384084. What's the point of all this? Even though at the time, it seemed like Cobb won by going to the fourth (and generally unseen) decimal place, historical research ultimately has shown that Lajoie actually won the race, .384 to .383. Cobb still got the car, and the research was done after Lajoie's death, so I'm sure it's hardly consolation.