Skill or Chance?

[cummings2]

Thnx for the bump J. 8)

Back to skill or chance,

What I've been thinking of today is along the lines of seeing results not from leagues or multiple teams by a manger through several seasons but rather into the smaller data of how runs are produced/prevented.

By this I mean, is the pythagorian record completely off from actual records of teams? My guesss is that it isn't. If the regular indicators that would lead to evaluating and projecting wins/losses are still, to a great extent in effect then perhaps it would be valuble looking at specific "players" performances in different environments.

To illustrate my point, because I don't think I'm doing too good. Fatty posted a team of his in petco in the strategy forum, the point was about solid defense and stud pitching with David Ortiz as the main offensive weapon. Well his David Ortiz in petco hit about 80 points higher than a current David Ortiz of mine in Ameriquest. I'll post links and stats later (I can't at the mo.) That would probably tip the scale more towards the chance argument.

Does this make sense?

[rgimbel]

one trend I have noticed(although of course I could be wrong) is that the longball is much more important in 06. I just finished with a team that was 14 games out of first. I had terrific pitching defense and great ob in dolphins stadium. My team hit .277 with 541 walks and the oba was .255 with 447 walks. the 1st place team that demolished me hit .275 and his oba was .274 but he outhomered me by a large margin. he had crappy def no pitching but managed to grab a-rod pujols and ortiz in 05 he would of had a terrible team but in 06 it is the guys like me with everett hudson and edmunds who suffer. my biggest mistake when I started like most people is that I went for power and disregarded def and ballparks a typical newbie mistake. In 06 you can win building your team like a newbie and that is why the vets are not doing as well.

or it could just be chance

[J-Pav]

[b:7fe03f9d85]C2:[/b:7fe03f9d85]

The bump was because some goofball named [b:7fe03f9d85]Jewish Prophet[/b:7fe03f9d85] posted like ten threads about nipples. The moderators cleaned 'em out about five minutes after I bumped it.

About your post...

Across all the years there have always been wide swings in individual player performances. This year is no different. So yes, on my teams I notice large differences as well, and I accept it as part of the game.

Here's the thing. Any individual game is chance. At a certain point, after a long series of games, it is mathematically no longer chance. If [b:7fe03f9d85]Marcus' [/b:7fe03f9d85]link is correct, and statistically speaking it's okay to use .05 as the baseline p value, then the Skill or Chance Code has been solved, at least as far as I can tell (with a few questions waiting for answers from our much smarter brothers).

I never thought the game was chance. I just didn't know why I didn't know it wasn't.

[b:7fe03f9d85]rg:[/b:7fe03f9d85]

I hear what you're saying. I don't know that I entirely agree with the conclusion, but I do think that defense is priced in more efficiently with the offense, so 3s aren't the liability they were last year (even 4s now, I guess).

I'm watching for balance, nowadays, trying to have avg, obp, slg, some speed, some gloves, solid pitching, while avoiding the extreme teams. As I watch and learn, maybe a theme will jump out at me. For now, my best teams are plain solid teams, and my struggling teams are the ones where I'm trying to be too clever.

[MARCPELLETIER]

[quote:b4548445c6]
Marcus:

This is a pure stats question, I think, and maybe the Fisher thing doesn't answer this question.

If we use .05 as our baseline, applied to ALL participants, what number of managers do you think we'll find from the total number of managers in the pool (1800, say) under .05? Then, when we have this number, how do we prove it isn't a number you'd see anyway in a regular bell curve distribution?[/quote:b4548445c6]

J-Pav,

I'll give a pure stats answer. You can't pool different players. Fisher tests (actually, all stats tests) work only if each event (for example each dice roll) is mathematically independent from each other. If you pool different players together, and if these players have played one against the other, then you broke the rule because you'll pool together two records based on dice readings---broking the assumption of independency.

Cummings,

I wasn't sure if you had a question for me. Please reformulate if so.

[MARCPELLETIER]

BTW, I disagree about this season giving an edge to power teams. Just like last year, I still think it is easier to win in a pitcher park, [i:2c1d503f63]all things being equal.[/i:2c1d503f63]

But things aren't equal, thankfully. There is a bulk of people going for a pitcher park, so it is perhaps the case that you make your life much easier in finding the players you need in a hitter park then in a pitcher park.

I had two homerun parks, and I had no problem findings all the outfielders I needed after the draft­.

[J-Pav]

[b:801ad0b039]Marcus:[/b:801ad0b039]

Maybe I phrased the question poorly. I'm not talking about pooling the players, I'm talking about testing each individual team and looking at the totality of the results. If you graph the total results of all individual p values, do you just end up with another bell curve? Or can you end up with 85% of the teams showing sub .05 p values, showing that 85% of the pool is operating outside the realm of chance?

[J-Pav]

In 2006, I've found very few teams under .05 in spite of what I thought would be an ample supply of games played, but admittedly I just breezed through the top 20 or so.

[MARCPELLETIER]

I have phrased my own answer quite poorly as well. I should have said that you can't pool managers, at least those managers who have played one against the other. So you can't pool different managers of a same league.

But you can indeed very well pool each entry of one participant (of course, we assume that the participant doesn't play against himself).

When you pool different entries of one individual manager, it doesn't matter whether you pool different seasons or not, because you are testing at the level of games. You are testing the probability of having this record given that number of games. Whether the number of games is 162, 81 or 40, is arbitrary. So yes, you could start to draw curve around the p-values of 162 games, but you could also do it for p-values of 40 games. The bell curves that you will create in each case will all be centered around the exact p-value of all games pooled together.

[J-Pav]

[b:c4502c4971]HAL[/b:c4502c4971] is gonna [i:c4502c4971]have[/i:c4502c4971] to love me now. If we've proven that SOM is a game of skill, but you need to play A LOT (and I mean, [b:c4502c4971]A LOT[/b:c4502c4971]) to demonstrate said skill by generating a sub .05 p value in a Fisher test, then this is an endorsement to play again and again to prove your skills to the world.

I'll anxiously await tonite's favorable sweeps, especially against [b:c4502c4971]Brother C2[/b:c4502c4971] in Tour League 3, when HAL will shower me with his omnipotent benevolence. :D :D :D

[MARCPELLETIER]

But as I said, J-Pav, taking .05 is playing the devil's advocate (in stats terms, a strategy to avoid a type-I error). But in fact, as soon as you get under .30, you can suspect (without betting your wife) that there is some skill under play.

Or you can bet your wife quite lightly as well; all depends....