Game Score is a data point I have used for several years. It is a starting pitcher’s performance report. Ideally, it tells us how effective the pitcher was in every start he has made. While it will not tell us exactly how he will pitch today, it does have predictive qualities when we apply sequencing and patterns to it. It also can be used in a couple of very specific situations such as the start of a season and when the pitcher is off a very poor result. I will get into all of this but let’s first explain Game Score (GSc), why it is effective, and how it is calculated.

Bill James was the creator of GSc. As we have come to know, a pitcher’s individual win and loss records are quite meaningless in the context of quality of performance. Wins and losses many times are not in the control of the pitcher as the offense, the defense, and the bullpens tend to factor into these wins and losses more often creating a false narrative of quality based on wins credited to the pitcher. Game Score is an effective tool and has been revamped from the original creation to even be more precise. Game Score v2 is the better choice to utilize. As you can see in the graphic above, GSc is a calculation based on performance and length of time on the mound. The graphic is the traditional or first version of GSc.

Game Score v2 is very similar in layout with a couple of variances. I will go over the calculations which then will also further explain the original calculation shown in the graphic. GSc only applies to a starting pitcher. GSc v2 begins with the pitcher at 40 points (GSc was 50). Here is the rest:

40

+2 outs

+1 K

-2 walks

-2 hits

-3 runs

-6 HR

*(Note: The K is double-counted, 2 points for the out, and 1 extra. The HR is double-counted, 2 for the hit, and the 6 extra.)*

The differences are in the starting point of 40, walks are more costly from (-1) to (-2), innings over 4 are removed, and the Home Run is added separately. This provides a more accurate assessment of the quality of performance the pitcher gave. This version is quite linear in that the result matches a percentage of the time the performance would win a game. If a score was 55, then it can be assumed that 55% of the time, that performance would win the game. Let’s use it for a random game. This box score is the pitching box score from the Marlins @ Braves 09/10/2021 where the Braves won 6-2.

MIA Trevor Rogers threw 4.1 Innings allowing only 3 hits but 4 runs. He did strike out 6, but what kind of performance did he turn in via GSc v.2? He produced a GSc52. The average MLB GSc is 50, so he just slightly better than average. On the season, Rogers averaged a GSc59 which is elite. ATL Ian Anderson threw 5 innings allowing 5 hits and 2 runs. He struck out 9 but allowed an HR. His GSc was 53. His season average is GSc53.

The purpose was to show how the score is calculated and how it reflects the quality a pitcher threw. In this case, Anderson threw slightly better than Rogers but could have had a much better outcome had not allowed the HR.

Now we can discuss how to use Game Score in predictive ways. First, let’s talk about the start of the season. As we are speaking in numbers, we need to understand averages and median averages. Players are not robots and will have good and bad days, thus their performances will fuxuate. The better (or higher GSc) a pitcher produces, the less fluctuation should be expected. Each pitcher must establish the bar of their performance or their “median” average at the beginning of the season. This is where we can gauge performance to be expected via sequencing. It takes roughly 4 starts to establish this. The median is the middle score of a distribution. Let’s use Ian Anderson from the above game as our model player.

We are looking at his game logs for 2021. It is chronological in order from bottom to top. His top line is his season average. GSv2 is the column on the far right. As you can see, there is a value for every game he started and his season average is 53. At the beginning of the season, his GSc results for these first 4 games were 56, 42, 47, and 68. His mean average or his “bar” was established at 53.25 for those games. Now we have a regression or progression process to follow, but before that, we need to discuss these four games! Virtually every pitcher will go thru high, low, and average performance in their first 4 starts thus establishing their quality “bar”. We can check them off kinda like clockwork using patterns to forecast what type of outing we can expect. Remember we have boxes to check off in these first 4 starts (HIGH, AVG, and LOW). Anderson started with 56 which at the time we know is above the MLB average 50, so we can assume it is going to be within the AVG box. It is not over 60 which is a very good start and we are looking for one of those. Going into his next start, we can assume he will either throw a HIGH or LOW performance game because his AVG box was checked off. Seeing how it was nearer to the HIGH side, it can be assumed a lower result should be expected on this start (regression to the mean). Remember, we are establishing a median average where there will be variances so we will expect them and make plans to predict the performance based on that. His 2nd start was a GSc42 which will check off the LOW box. His current GSc for the season is 49 and barely below the average pitcher. Now we can assume he will throw AVG or better in his next two starts because he has already checked off the LOW and yet to check off the HIGH. His 3rd start was a GSc47, which again means he did not check off that HIGH box. While assuming he would’ve thrown better prior to this game or at least a high avg game, this one sets us up for what we are looking for. He has yet to hit that HIGH box and his 4th start is on the way! We can only assume this will be the one. We can assume this again because he has yet to have a HIGH score and needs to progress the mean average as he is below the bar right now. His 4th start was a 68! There we go. It means he threw an excellent game and we anticipated it. This comes in very handy when trying to make wagers! If you “know” he will throw either a good or bad game, we have an edge that we can capitalize on!

Another scenario where we can certainly use GSc to gain an edge is where a pitcher is off a terrible start. The average start in GSc50. When a pitcher has thrown something under 40, it is a terrible day. What tends to happen after that start? The law of averages kicks in(progression back to the mean). The below 40 starts have lowered the season average below the median average, so we should expect an adjustment up to correct that. It can happen in the very next start or it may take two great starts in a row, but we can expect above-the-bar performances off of below-the-bar performances. Let’s look at Anderson to see if this comes through for us. On 5-29-21 he faced the NYM and threw a Gsc30. We will expect his next start to be above average or better. He threw a 52 which is slightly higher than average and not the offset to the low score we need so in this case he will likely use two starts to get there. His next is the jackpot! He threw a GSc76! Other starts below 40 for the season were 07-11 GSc18, where his next start was a GSc62, and 09-04 GSc18, where his next start was a GSc53, and lastly on 09-17 GSc37, where his next start was a GSc81. We can see the pattern here. Expect good to great performance off of poor ones. It is also true to expect poor or AVG starts of great ones. It is just the law of averages (regression/progression) but if used properly, we can predict to a degree of what type of performance we should expect.

Games Score can be a valuable tool if utilized in the proper ways. Standing alone, it is a data point that accurately grades performance. When applied to patterns and sequencing, we can see predictive tendencies that will generate a profit for us.