We all know that OOTP is weird. Sometimes a rating has a greater or lesser effect than we expect, and often the relationship between ratings and stats can be completely counterintuitive. I decided to analyze the three OOTP "running" ratings - speed, stealing, and baserunning - and see what effect they had on stolen bases.
To begin with, my data only included the 2129 season. Since ratings can change, I thought it would make the data less reliable to include previous seasons. I also only included players with at least 6 steal attempts. A consequence of this is that most players on the list had good running ratings, but I don't think this matters, as nobody would want to steal a base with a 3/3/3.
There were a total of 104 players on this list. Their average ratings were 7.5/8.6/7.9 (speed/stealing/baserunning) and their median ratings were 8/9/8. I analyzed each individual running rating in isolation and compared each one to SBs (total number of bases stolen) and SB% (steals divided by steal attempts). I'll begin with the stealing rating.
The slope of the trendline for the stealing vs SBs data was 4.75 - in other words, for every increase of 1 in stealing, we could expect an increase of 4.75 steals, all else being equal. So a 7/9/7 would be expected to have 9.5 more steals than a 7/7/7 (over the period analyzed, which is one full season minus three weeks). I personally don't think SBs are particularly indicative, since team strategy and player strategy settings have a huge effect and are difficult to account for, but SB% is a better indicator. The slope of the trendline for stealing vs SB% was 3.17 - an increase of 1 stealing rating point led to, on average, an increase of 3.17% success rate. This is less than I expected, and indeed, the R^2 coefficient (a statistical metric used to determine strength of correlation) was just 0.151 (out of a maximum of 1), suggesting a relatively weak correlation. In other words, statistically speaking, stealing rating has a tangible but fairly small effect on stealing. Incedentally, the R^2 for stealing rating vs SBs (not SB%) was 0.243, indicating that the stealing rating is a stronger indication of the number of steals than the steal percentage.
On to the baserunning rating, which might be the least used rating in the game. The slope of the SBs vs BSR trendline was 2.19, and for SB% vs BSR, it was 2.33. In other words, an increase of 1 BSR rating led to, on average, an increase of 2.19 stolen bases and a 2.33% increase in success rate. The R^2 coefficients were 0.089 and 0.141, suggesting that baserunning has much less of an impact on stealing bases than the stealing rating.
Finally, on to (in my personal opinion) the most interesting data: the speed rating. The slope of the trendline of speed vs SBs was 3.28, and the R^2 coefficient was 0.172, suggesting a correlation between speed and SBs. However, the trendline of speed vs SB% had a slope of essentially 0 (-0.0948), and the R^2 was
0 (the software I used rounds R^2 to three decimal places, so technically I should say the R^2 was less than 0.0005 and therefore rounded to 0). In other words, speed has essentially no effect on stolen base percentage. The correlation between speed and SBs could be because teams set their fastest players to steal more often, or it could be programmed that way within the engine.
An interesting example of this is Northwest's
Phillip Moore, who had the highest SB% of my entire list - 13 steals and only one CS, for a rate of 92.9% - with a speed of just 5. His STE and RUN ratings are 9 and 8, respectively. Oslo's
Marcio Stallone is also an interesting case. He currently has 20 SBs and a speed rating of 10 - but his low STE (only 5) means his actual success rate is barely over 50%.
Finally, a clarification for those of you who know a bit about statistics. In this article, I used linear trendlines simply because they're easiest to explain. However, I did check all the data with exponential, quadratic, cubic, logarithmic, and power series trendlines, and in every case, the R^2 stayed very similar to the linear trendline. This actually surprised me, as I had believed all OOTP ratings follow a normal distribution rather than a uniform one, and therefore believed that exponential trendlines would fit the data better - but this did not occur. A potential explanation of this is that running ratings do not follow the same normal distribution that batting and pitching ratings do, which actually is supported by (at least my) anecdotal experience - I notice many more 9s and 10s the running ratings than in batting ratings.
I did make graphs for these data, but am not including them here out of a desire for brevity. I'll post them on Slack if people are interested.
Guests: 77
Hidden: 0
Users: 0