Stats stats stats... any geeks want to share tricks?

[YountFan]

No, Parkers need to get to 300 by the time he gets 600 ab. Or 176 H in his next 560 ab or a .314 average....very doable.

The issue with tracking deltas is you can factor out the streaks. I know from my data that most players have (on average) 2 27 games segments when they do bad, 2 average and 2 monster. Depending on when these fall in the season you can make bad selections. If they happen at the start you may drop a great season, or keep a bad one. If the monster is at the start the stats could still look good at game 54, maybe even 81 depending onthe start, but he'd have two bad deltas for you to think about.

Ball-park dependent guys are even more streaky than others...think Strawberry here.

I've had guys stink for 81 games and then be an all-star the last 81, or visa versa. You can never tell if it is the card, the competition or the dice. In the end it comes down to the dice.

Roll the dice..play the game

This is Will Clark and Horner in a current league. That first split ifor Clark s pretty bad.
[code:1:24302ecf9c]Clark
Spilt BA OBA SLG OPS
s1: .245 .322 .330 .652
s2: .353 .375 .655 1.030
s3: .289 .349 .474 .823
s4: .392 .461 .627 1.088

Horner
s1: .276 .323 .506 .828
s2: .192 .234 .315 .549
s3: .266 .318 .481 .799
s4: .326 .373 .632 1.004

Works for Pitchers too: Dave Stieb
ERA WHIP
s1: 4.50 1.54
s2: 4.10 1.34
s3: 3.02 1.18
s4: 2.40 1.17

[/code:1:24302ecf9c]

[yak1407]

Are your stats for your deltas just the 27 games, or are those the numbers after 27, after 54, after 81?
I appreciate your point about Parker over the long course of the season.
The point I was making, however, is most of us, myself included, make short-term decisions. What is he doing after 100 ABs, in his last 10 games etc. With Parker, after 100 ABs, I might be considering letting him go if, for example, he was batting .250 even though over his last 60 games he was batting .300.
Deltas are a great idea.

[YountFan]

The deltas are for 27 games only 1-27, 28-54, 55-81, 82-108, etc. SOM already tracks the cumulative stats. As you can see in the deltas that Clark was droppable after the first split, but obviously he a keeper. It's the hot or cold start that fools people. A cold start they get dropped and a hot start they keep a non-keeper. Sometimes I track 9-game deltsa for the first 27. BTW 27 games makes month. And if you track enough monthly stats you know that 10 homers a month is a lot, and so are 5 wins and 30 hits and 25 RBI's, and so on.

The other key is to normalize the hitters production to league averages. Adjusted production is a easy way to do it. If the league OPS is .800, then 830 is not a good as if the league OPS is 750. etc. The same can also be done for ERA.

As for Parker example. If he was at .250 and had hit .300 ove the last 60 games (over 200 ab's) he a keeper because he must have been really stink-o before.

All this stuff is waht makes it fun

[Outta Leftfield]

If you're as lazy as I am--or as unmethodical--you could simply take screenshots of your team at various points in the season.

That would give a point of reference for the team and how different players are doing without the effort of entering all the data into a spreadsheet.

[seanreflex]

[quote:e42b21b457]If you're as lazy as I am--or as unmethodical--you could simply take screenshots of your team at various points in the season. [/quote:e42b21b457]

I use the same high-tech method as you, Outta ... which probably explains why I have the worst winning pct of all the active managers playing the game :(

[dkoulomzin]

I've thought about how much the pitcher's cards can skew stats... its something that's not easy to fix without a bunch of work. I'd have to somehow adjust each year of stats with the "average" performance of pitchers in the league. This would be a monumentally difficult task.

I haven't done too much research on how the cards attempt to recreate season stats, but I assume that if a guy with a .300 BA faces a pitcher who allowed a .250 BA, he will hit .275 on average (assuming all other things equal). Also, I assume that each card assumes that the opponent (hitter or pitcher) is the league average for that year. So I could estimate the true probability for a hit by calculating the batter's card's chances for a hit (as a BA) and averaging it with the league average BA. The same can be done for pitchers. This should make the bayesian prediction a bit more accurate.

Of course, to do that, I'd probably want to write some software to pull stats from the TSN site. Which would be some work. Of course, then I'd be undefeatable! *Cough*

-Dan

[dkoulomzin]

[quote:9c7b6ac542="yak1407"]
I appreciate your point about Parker over the long course of the season.
The point I was making, however, is most of us, myself included, make short-term decisions. What is he doing after 100 ABs, in his last 10 games etc. With Parker, after 100 ABs, I might be considering letting him go if, for example, he was batting .250 even though over his last 60 games he was batting .300.
[/quote:9c7b6ac542]

That's precisely why I like my Bayesian method. Rather than just guess what season a guy is in, it guesses the probability for EACH season. Taking Parker (4 for his first 40), we get:

78: 3.9%
79: 10.5%
83: 33.8%
85: 9.7%
86: 41.7%

(I've rounded, which is why it doesn't add to exactly 100%)

So yeah, '86 is most probable, but its not even half odds. '83 (his worst year) is about 1/3 chance. With more info (like BB, K, etc.), more educated guesses can be made... and remember that this is based off his real season stats, so it doesn't factor in the fact that he's facing (probably) better pitching than he did in real life.

Note that if he went 8/80, he'd be 94% likely to be in 83 or 86... so the guesses become more sure with more samples. By the time you're at 40/400, he's in '83 10% of the time, and '86 90% of the time (the other seasons are so statistically unlikely as to be negligible). Of course, in this case, BP effects and pitchers could really be skewing stats, since its an edge case. But you'd have definitely dropped him after he bats .100 for 400 ABs. :)

[honestiago1]

How much do ballparks skew numbers? What if the league is more of a hitter's or pitcher's park? What if Big Dave eats three cheese balls before every at bat? (okay disregard the last question, pse).

[YountFan]

1978 Rice in Riverfront in 632 AB

BA SLG OBP OPS
.256 .502 .304 .806

78 Card
315 .600 .369 .969

Explain that!

[yak1407]

Finally pulled the plug on Parker after he went 11 for 62 (.177) following his 4 for 38 (.105) start. What really made the decision was the fact he had 1 extra base hit, a double, in 100 abs.
Considered Yount's delta approach, but didn't want to take a chance that I would have a second bad 27 games. My league is tight and I didn't want to fall off while I waited for Parker to produce.
This is actually the second time he's been a dud for me.
But I'm curious as to what the Bayesian method makes of this:
This is 114 games into the season:

Cincinnati #11 (R) vs. L 104 12 1 0 2 5 10 11 .115 .183 .214
SS vs. R vs. R 227 50 9 1 3 22 15 29 .220 .308 .275
TOTAL 331 62 10 1 5 27 25 40 .187 .269 .256

In his worst season, he should be hitting LHPs slightly better than RHPs. However, Larkin has nothing that would suggest this kind of performance, and it has been consistently bad all year. The only reason I've kept him is that there is nothing else out there and my team, playing in Fenway, got off to a bad start and even with tweaking has only become a .500 team and will likely end up with around 70 wins.
In the lone season where he supposedly hit RHPs better than LHPs, he batted over .300.
Any theories, ideas?????