TL;DR

I wanted to make a good way to rank player value in roto leagues, and I came up with four methods, each with its own successes and flaws, so you can decide for yourself which was the best:

​

​

I have been in a standard 5x5 roto league for about 10 years now, and every year there is one thing that bothers me about roto. It is nearly impossible to rank or compare players, especially those who have extremely different skill sets. For example, comparing the value of a starting pitcher to a closer is incredibly hard to do given that they contribute two very different aspects to your team. The same can be said when comparing SB/BA guys like Whit Merrifield or Starling Marte to HR/RBI guys like Luke Voit or Pete Alonso. In points leagues this comparison is simple, just by looking at the points they have scored, ppg, etc. But in roto leagues this is much more difficult, so that has been my project this offseason: accurately rank roto players by comparing how much they contribute to your team.

I began by simply taking the percentile each player had in each of the 5 stats (for hitters: BA, HR, RBI, R, SB; for pitchers: W, ERA, K, WHIP, S) and multiplying each by the value they hold in our league (BA, HR, W, and ERA are worth 3 points, all others are worth 2). Therefore, a perfect score in this scenario would be 12 points, if a player led the league in all 5 categories. This method resulted in the top 5 hitters of 2020 being Trea Turner, Jose Ramirez, Manny Machado, Mookie Betts, and Juan Soto, and the top 5 pitchers were Shane Bieber, Trevor Bauer, Yu Darvish, Kenta Maeda, Clayton Kershaw. Although those are arguably solid top 5 lists, there were also a few questionable rankings (Xander Bogaerts as a top 10 hitter, but Jose Abreu at 20 and Marcell Ozuna 21). My thought process in where this went wrong boils down to one small flaw in this system: leading a league in a category by 1 is the same value as leading the league by 100. And let's say there is a huge drop off in a stat after a certain number of players. That drop off is valued exactly the same as the increments of the players who are all bunched together. So I remade the model.

Now, instead of looking at percentiles, I looked at the percent of the total number of those stats that were contributed by a player. Using this method did not change the rankings of the rate-based stats of BA, ERA, and WHIP since you can't determine what percent of a rate was contributed by a player, so I just used the same method as before for these. For all others, I divided the player's number for each stat by the total amount of that stat in the season (for example, Luke Voit hit 22 HRs, and there were 2304 HRs hit in 2020, so his score would be .0095). These scores all resulted in extremely low numbers, so in order to standardize it, I multiplied each score by the inverse of the league leader's score in that stat. So using Luke Voit as my example again, since he led the league in HRs with a score of .0095, I multiplied every players' HR score by 1/.0095, which was about 104.72. This gave Luke Voit a HR score of 1 (which is what I wanted) and each HR less than him resulted in a drop in your score. Using this valuation, the top 5 hitters were Jose Abreu, Jose Ramirez, Fernando Tatis, Trea Turner, and Marcell Ozuna, and the top 5 pitchers were Shane Bieber, Yu Darvish, Trevor Bauer, Liam Hendriks (WHAT!?!?!), and Gerrit Cole. Overall I agreed with these rankings much more, but the one that surprised me was Hendriks as number 4. This trend continued, as there were 4 RPs in the top 15 pitchers. It makes sense that the model would do this since guys like Hendriks, Brad Hand, and Alex Colome are obviously getting so many more points in saves, and will also be near the top in ERA and WHIP, and they even are slightly competitive in SO, since the huge SO guys like Bieber, Bauer, and DeGrom were not as far away in strikeouts this season as they normally are. But amazingly, using the same scoring method for 2019, there were 5 RPs in the top 13 and 12 in the top 30, including Josh Hader at number 3, above guys like DeGrom, Bieber, Strasburg. This surprised me, but it also made me consider the idea that RPs are slightly undervalued in roto leagues (a little hypocritical since I am a firm believer of throwing saves and instead using SPaRPs to gain a lead in wins and strikeouts).

I also made a third scoring system that is based on the mean for a stat and standard deviation a player is away from that stat's mean to score a player's value. The thought process here was to value players consistently solid at 4 or 5 stats over players who are really good at 1 or 2 (I only made this so I could feel as though my college education was being put to work, fuckin stats majors am I right). This method resulted in nearly identical rankings to the shares method above, as the only difference for hitters top 5 was Freddie Freeman replace Trea Turner at 4, and for pitchers it just swapped Hendriks and Cole at 4/5. The one thing I noticed is that this method valued RPs significantly less than in the previous scoring method, and I personally think this method values RPs the most accurately. On the hitting side, the rankings are almost identical all the way down; DJ LeMahieu was the only player in the top 30 who's rank moved by more than 5 spots (he moved from 20 to 14, and I agree that he was a top 15 hitter). Overall I think I like this scoring method the best, even though it also has its fair share of issues (the means for the counting stats are a bit odd since, unlike the rate stats, they do not follow somewhat normal distribution curves, so the mean number of stolen bases is like 3 or something, and any player with 0 stolen bases got a negative score in this method. I am not sure if that is bad, since overall the rankings were pretty much the same as before, just interesting to see where they differ.

Because I am still unsure which method is more accurate, I made a fourth ranking that takes the average score of the shares method and z-score method and ranks players based on that, and I think that is what I will be using going forward. Since I don't know which of the two methods is better, please let me know what you think regarding those two methods. They are basically the same in hitting, and pitching is generally similar besides the discrepancy in RP ranks.

In conclusion, this has been an extremely interesting project for me to work on, and I will definitely use these scoring methods in the future when trying to examine and compare different players and their value to my team. These three methods are all in a spreadsheet I have created that, once completed, will be able to rank very player (both overall and by position) based on both my own projections and other peoples' projections, and will also have different uses during the draft process, such as the other teams' rosters and positional needs by team to help me decide who I should pick. I am still learning all the ins and outs of excel, but man am I having fun learning how to use it. If you have any questions or comments about scoring and ranking players in roto leagues or want to learn more about the excel file let me know in the comments. Thanks for reading!