Fun with patterns

by rocKStark5 on Jul 1, 2008 11:52 PM EDT

Anybody that's played poker at a winning level (+300K hands) is fully aware of things like variance, regression to the mean, confidence intervals etc. I often giggle at all kinds of reasonings behind numbers just running their course.

I developed a simulator and plugged in 162 games for an imaginary team I created called the St. Louis Donkeys.

We got off to a so so start, 11-9, however, they did have some "momentum" going for them and turned it into a 18-2 stretch. The team had a cold streak here and there and go 13-19 headed into the allstar break at 42-40.

The team had some ups and downs, they're still in this thing. Whoops, 8-19 nose dive. I won't stand for this, I have good talent. My manager isn't getting the most out of his players! 50-59?? This is AT LEAST a .500 team (foreshadowing). I'm gonna fire the manager and put in one that really gives great "win one for the Gipper" speeches. New Skipper goes 12-13, however, at 62-72 and less than 30 games to go, time to wave the white flag. Even though my team is full of professional athletes, I guess they were so demoralized they ended the season on a 2-13 skid (damn morale) and finished the season 71-91, 20 games out. Maybe they should have had a cardboard cutout of Rachel Phelps.

I KNOW they were a .500 team!! How do I know this? My high tech simulator wasn't stratomatic, it wasn't a fancy computer program, or even a video game. It was a coin. Heads = Win, Tails = Loss 50/50.

I guess I'm a terrible coinflipper, however, I have some "interesting" splits:

When Flipping a Quarter I'm 50-48, however, when flipping a nickel I'm 21-43, yikes! To break it down even further, when Flipping quarter left handed I was 25-13. If I flipped the quarter left handed I would have had a 107 win season, right? Anybody scared to take me on flipping left handed quarters? I don't understand where I went wrong when flipping nickels...I made sure I didn't "make it think too much". I know it had .500 potential, must have been the Quarter's ego causing problems in the clubhouse.

Obviously people are not number producing robots, however, most likely if something is out of the ordinary it's most likely probability being probability, not some stupid intangible. When dealing with such small edges you're going to see all kinds of wild patterns and there isn't necessarily some pep-talk behind them.

Do it yourself if you want, that's why I used a coin :-)

0 recs | Comment 15 comments

Story-email Email Printer Print

Comments

I'm not sure if this method has any credibility whatsoever

Maybe I just don’t understand what your method has to do with a team’s winning percentage, or anything in baseball.

The idea that a team has 50/50 chance of winning a game is ludicrous or predicting the future of the game. The first rule of statistics is to never trust an average. For example my Micro-teacher Sophomore year of college once said “You’re going to a party with 6 girls. What’s their average age?”. “21”. “Okay. So we get 5 14 year olds & 1-56 year old”. Basically, not a good party for a 21-year old guy, nor do you ever trust averages again.

Thus, I look at it in a situational context. Or how does a team respond. We believe that a left-handed pitcher has a better chance against a left -handed batter. It’s reading the ability that allows for success. Basically, I would rather go to a party with 22-year olds close to my age than a 6 year old.

Basically, you say that you would flip a coin whenever Albert Pujols goes up to the plate, and that he has the same chance as Randy Johnson to get a hit. That’s ridiculous.

Prospects are good. Especially when they aren't rushed.

by CraveCase on Jul 2, 2008 10:46 AM EDT 0 recs

I think he is saying the opposite.

Stuff happens. Aaron Miles isn’t a great player anymore now then three years ago (for example.) He’s just having a random event and we should still look to replace him. We wont for the rest of the year because he had two good weeks in June…and that’s the point.

by Harknights on Jul 2, 2008 11:25 AM EDT up 0 recs

the trick is being able to distinguish

a random aberration from a non-random one. is kyle lohse simply the beneficiary of a lucky run of coin tosses this year, or has he grown as a pitcher? is chris duncan just getting UN-lucky, or has the league figured out his weaknesses?

usually each player’s (and team’s) results can be explained by a combination of random and non-random factors. the ability to read and weigh those factors accurately is what separates smart teams from dumb ones.

by lboros on Jul 2, 2008 11:34 AM EDT up 0 recs

+1

That’s the hard part, however, most people don’t realize just how big the variance can really be and how long it takes to normalize the numbers. Even I was shocked that my coin flipping team lost that many games.

This is were the Marriage of where Scouts and SABRs come into play. Somebody called it moneyball 2.0 and attribute the Red Sox/Epstein as being the god fathers of it.

This is also why I think predicting pitching performance is SOOO tough because it only takes another plus pitch to make a world of difference.

What I personally love about baseball is there are stats behind the stats.

Ex: ERA+ measures almost pefectly how a pitcher HAS performed, however, FIP, FB to HR ratio, K/BB and other things tell the story behind ERA+ and if it will be maintainable.

"How depressing is it being you? Would you equate it to being a lifelong Cubs fan?"

by rocKStark5 on Jul 2, 2008 12:48 PM EDT up 0 recs

Yeah, random distributions can be killer

If you look at the binomial distribution with n=162, and a 50% probability, you will find that you only get a number of “wins” between 75 and 85 half of the time.

They say that it's never too late, but you don't get any younger...

by Valatan on Jul 2, 2008 3:59 PM EDT up 0 recs

In fact, I'll post the graph as a fanshot

They say that it's never too late, but you don't get any younger...

by Valatan on Jul 2, 2008 4:00 PM EDT up 0 recs

no, that's not what I'm saying at all...or even close to it.

In my example I “programmed” a .500 team. How do I know they are a .500 team? Cause I used a coin.

My point was to illustrate that just because a team is really a .500 team they won’t finish 81-81, they can finish with all kinds of records and during that time you can find all sorts of stupid patterns in RANDOM data.

I’m not saying the cards, the astros the yankees or anybody is a .500 team here, just my team, the donkeys.

"How depressing is it being you? Would you equate it to being a lifelong Cubs fan?"

by rocKStark5 on Jul 2, 2008 11:27 AM EDT up 0 recs

yeah, i think you missed the point of this post...

here’s the way i interpret what rockstark5 is saying: let’s say you have a pretty talented team that, based on a number of projections, should win 90 or so games in a season.

it’d be kind of like having a coin that was slightly weighted on one side, such that heads comes up about 56% of the time. if you flip that slightly weighted coin 162 times, are you going to go 90-72? maybe, maybe not. if you did a whole bunch of 162-flip seasons, the most likely result might indeed be 90-72—but in a given 162-flip season, with the same weighted coin (i.e. with the same talent), you might very well go 81-81. kind of like how the d-backs had such a great W-L record last year, with what seemed to me like a .500-ish level of talent on their roster.

so i don’t really think he’s saying that randy johnson and pujols have the same likelihood of getting a hit in a given plate appearance.

by djones9 on Jul 3, 2008 3:17 PM EDT up 0 recs

But it should be say that there's more to a w-l record than random chance and pythagorean

In particular, a team with a topheavy rotation and a shutdown bullpen will outperform their pythaogrean, on average, because they will win a bunch of close games, and get demolished in their losses.

They say that it's never too late, but you don't get any younger...

by Valatan on Jul 3, 2008 3:27 PM EDT up 0 recs

Exactly

Last year, we had Cavasos, Jimenez, Cate, Falkenborg…we could put them in after Wells, Maroth, and Reyes combusted and wave the white flag. When we stayed in a game we had Springer, Franklin, Izzy to put in. I think pythagorean gets a little more credit than it deserves.

by Merry CRasmus on Jul 3, 2008 3:31 PM EDT up 0 recs

Not exactly

Pyth. shows how well/poorly a team actually played. Good teams don’t get blown out with enough regularity to inflate a pyth dramatically, considering they have enough bats to trim down deficits and make 10-1 games 10-5 games and so on and so forth. Furthermore, they don’t have a ton of shitty pitchers who get bombs dropped on them like crazy. A good team wouldn’t have had to start Reyes, Maroth, Wells, or Mulder, and wouldn’t have had to turn to bums like Falk, Cate, Jimenez, Cavazos, Dove, etc., because those guys wouldn’t be on the team at all on a good team, or at least not in quantity. A far more worthwhile exercise is trying to figure out why a team “beat” or “lost” their Pyth. instead of trying to figure out how the Pyth fell short. As you laid out, in the Cards case it was b/c Franklin, Izzy, and Springer were lights out, and not because the Cardinals were a better team then their Pyth. showed.

"Your Holiness, I'm Joseph Medwick. I, too, used to be a Cardinal."-Joe Medwick, to Pope Pius XII.

by redbirdnation8206 on Jul 5, 2008 4:38 PM EDT up 0 recs

That's circular

If there are rational, repeatable reasons why a team is outperforming it’s Pythagorean record, then it is a better team than it’s Pythagorean would predict. The Pythagorean is a heuristic, after all—it’s a useful tool that helps us understand how win differential relates to w-l records. It is not a definitive definition of how good/bad a team is.

They say that it's never too late, but you don't get any younger...

by Valatan on Jul 5, 2008 7:29 PM EDT up 0 recs

just curious,

Has it been empirically proven that teams that beat their Pythagorean have those qualities? The only thing I remember from last year was the DBacks consistently beating their pythag and everybody was freaking out and didn’t know why. Has it been applied retroactively and proven?

"How depressing is it being you? Would you equate it to being a lifelong Cubs fan?"

by rocKStark5 on Jul 5, 2008 7:37 PM EDT up 0 recs

i definitely agree -- there are certainly other factors at play...

for instance, our pythagorean record took a decent-sized hit due solely to that 20-run drubbing we experienced at the hands of the phillies earlier this year. but should that one game really change the prediction all that much? not really.

by djones9 on Jul 3, 2008 3:34 PM EDT up 0 recs

It's a good point

Exactly when that threshold when a hot or cold streak becomes a glimpse at something substantial is always a tricky science. For example, Mather and Stavinoha’s cup of coffee does nothing to show me what they could project to be one way or another. I defended Duncan for a long time, because there is a little history to show that it all may have just been a slump. However, after the struggles carried on I started to accept the possibility that he may not be what most had projected for him going into the year. Moved to completely on the fence with him and still am now.

There will be a wide amount of variation when you look at a person’s performance over a week, or even a month. As the games, AB’s, innings, etc… build up the pictures become clearer and clearer. The poker analogy is good. A horribly weak player can make money over a week or 2 of play, and a professional caliber player can lose. If you look at those 2 over a few months it will probably straighten out, and if you look over a few years they certainly will.

Personally, in baseball, the struggle is knowing when observations are just reactionary and when they are meaningful long term changes in performance.

by Merry CRasmus on Jul 3, 2008 3:26 PM EDT 0 recs