This article really doesn't have a lot to do with the Cubs, but a lot to do with the uncertainty of wins. No, this is not another article about how poor the wins statistic is for pitchers. This is an article about how we really shouldn't put a lot of stock in wins for a team - much less a pitcher. Even furthermore, this article is not about predicting future success for a team either. We're not looking at pythagorean wins or anything like that. This article is only going to point out something anyone who has taken basic statistics would know, but for whatever reason has not noticed the implications of it. When a baseball game is played there are basically two outcomes: you win or you lose. This a simple case of what statistics books call a binomial event. There are two outcomes and (generally) we can assume that two games back to back are unrelated. Furthermore in statistics they have what is called a normal distribution. The binomial and normal distributions are not the same thing, but when you have a large sample they converge so we can use normal distributions to approximate them.

Let's look at a simple example of two identical teams playing each other. Each team has a 50/50 chance of winning the game. We want to know the chances of team one team winning at least one game in a two game series. That probability is 75%. Now expand that over the course of 162 games. Using the binomial distribution we find that there is about a 9 percent chance of that average true talent team winning 90 games! That's huge!

Now let's step back a bit. There is also what is called a standard deviation. In this case a standard deviation measures the dispersion of the distribution. Basically, in a normal distribution a certain percentage of the data lies within one standard deviation of the mean, a certain percentage within two, and so on. The exact numbers: 68% lie within one, 95% within two and 99.7% are within three.

The standard deviation for a binomial is simple to find. It is sqrt(n*p*(1-p)). In this case n is the number of games played and p is the probability of winning. Looking at our average team, we get sqrt (162 * .5 * .5) = 6.36 wins. So that means that there is a 68% chance that out average team wins between 75 and 87 games. Standard deviation is dependent on how good the team is and .5 maximizes that. The worst and best teams have the smallest standard deviation, but all baseball teams will have a standard deviation in the low sixes.

Now let's look at the implications of this. There are 30 teams in baseball. In any given season we would expect ten of them to have their final win total be at least six wins away from their true talent win total. That's one-third of the league! Remember that 5% of the data is two standard deviations away. We expect that one or two teams would be alteast TWELVE wins away from their true talent win level. Finally we would expect that about once every eleven years a team would be NINETEEN wins away from their true talent win level.

One thing to note is we're not talking about under and over performances, we're basically talking about flipping a coin. If you flipped a coin 162 times you would expect on average to get 81 heads and 81 tails. If you did that 162 "flip season" 30 times. You would have expected at least one "flip season" to have heads or tails 92 or 93 times!

Now let's think about how close division races are at the end of the season. Usually a team is within six games of not making the playoffs, most of the time its less than three or so in the National League it seems. Now whenever someone asks you "why" the team finished ahead of the other team by one or two games you can answer the only correct reason: they got lucky.