Like many fans, I've long been interested in inventing statistics. I've spent a good deal of time playing with correlations, trying to devise interesting combinations of stats that might define a player or team in some way other than those we already use. To accurately measure performance, though, we must consider a few things.

 

Winning is a team's ultimate goal

Runs correlate extremely well with wins

Players should be creating runs so that their teams can win

 

So it appears Bill James was right. Creating runs is a player's most important duty. So James' stat, "runs created" is certainly one that very accurately measures player performance. In fact, when measured by team, it correlates with offensive runs with a correlation coefficient of about .87 since 1998, a very strong positive correlation.

 

Interestingly, though, OPS, or on base percentage plus slugging average, correlates about equally as well with offensive runs. And it has the added bonus of producing totals that are on average about 758, while teams tend to score about 770 runs. So we can tell pretty closely how well a player is doing if we know his OPS and understand how many runs it might take for his team to win so many games.

 

A third statistic, though, also correlates to an even slightly higher degree with run production. A few years ago, a couple of serious statisticians devised linear weights, a system of defining how many runs a player "creates" in a very specific and accurate way. According to those statisticians, the following run totals are accumulated:

 

Walk = 0.33

Single = 0.47

Double = 0.78

Triple = 1.09

Home run = 1.4

 

Obviously, as we can see, linear weights are not easy to mentally calculate, despite their accuracy. One of my goals has been to devise a statistic I could sum at the ballpark while looking at the scoreboard. Linear weights gave me that clue.

 

Something very quickly jumps out about linear weights: the difference between each type of hit is 0.31, or roughly the same as the value for a walk. In other words, these weights can be broken down like this:

 

Walk = 0.33

Single = 0.33 + 0.14

Double =  0.33 + 0.14 + 0.31

Triple = 0.33 + 0.14 + 0.31 + 0.31

Home run = 0.33 + 0.14 + 0.31 + 0.31 + 0.31

 

And if we define 1 to equal a rough value of about 0.31, we get:

 

Walk = 1

Single = 1.5

Double = 2.5

Triple = 3.5

Home run = 4.5

 

It appears that we can get values similar to linear weights if we just measure the number of bases the player accumulated and give him another 0.5 credits for getting the hit. In other words, we might be able to measure player performance by adding his total bases plus walks plus half his hit total. Or, in abbreviated form, TB + BB + H/2. This idea can also roughly be derived from the OPS formula previously described.

 

I suppose it is debatable as to whether this statistic is in fact easier to calculate than OPS, but it does offer a few additional advantages. It can be used equally well to measure pitchers, and might be called something like "base values" since it in essence measures the value of the bases the player is accumulating, which in turn correlate well with runs scored.

 

But perhaps the biggest advantage of this statistic is its ease of use in simulations and its durability: if we know what percentage of the time a player bats to which types of events (single, walk, etc.) we can judge his ability per a certain number of plate appearances. For example, if a player walks 10% of the time, singles 15% of the time, doubles and hits homers 5% of the time each, it is simple to add (10 x 1) + (15 x 1.5) + (5 x 3.5) + (5 x 4.5). Adding up those numbers, we see that per 100 plate appearances, this player is getting approximately 72 of these base values. It's not a measure of how often he gets on base or simply how many bases he records, but is instead a good measure of just how valuable each plate appearance of his is based on what types of hits he is getting how often.

 

Therefore, because of its ease of calculation and accuracy, I use it often as a benchmark statistic.