October 10, 20187 yr I put together some rush stats in our games vs USC, TCU, and OU. Baseball has all sorts of useful stats that are published for each player, but football has comparatively few. For running backs, one of the most commonly given stats is YPC. But to determine how consistent an RB is, YPC is not very useful. For example, if an RB made fourteen 1 yard runs and then broke a 40 yard run, his YPC would be 3.6. However, the confidence in his ability to pick up 3 yards on a run would not be very high. Really, want you want in a run game is the ability to consistently gain yards and YPC is not a very useful measure for evaluating a reliable run game. I've come up with another measure to to determine how reliable a ground game and given RB is at picking up yards. I refer to is as RYPC and it is calculated as follows: RYPC = YPC - [sum_1 to n (if(carry k -YPC < 0, then YPC - carry k, if not then 0)]/n, where n is an RbBs number of carries. RYPC is a measure of how close an RB will come to a given number of yards or better in a given carry. In the above example, if the RB made fourteen 1 yard runs and then broke a 40 yard run, even though his YPC would jump from 1 to to 3.6, his RYPC would only move from 1 to 1.2. Players and offenses with high RYPC will be perceived as having a reliable, punishing ground attack capable of long, sustained drives. If you want to consistently get about 3 yards or better on 1st down, you wouldn't want to go with a back that has a 1.2 RYPC. Here are the rush attempts and calculated RYPC for Watson and Ingram for the USC, TCU, and OU games: Watson: 3.8 YPC, 2.6 RYPC Ingram: 5.9 YPC, 3.8 RYPC
October 10, 20187 yr I've thought of something similar in the past for individual stats, but my initial thought was to calculate the mean and standard deviation of a back's runs and then calculate the probability that they would gain 10 yards or more on a series of 3 carries. Since you've got the data set, go ahead and let me know. 😄
October 10, 20187 yr 3 minutes ago, Huckleberry said: I've thought of something similar in the past for individual stats, but my initial thought was to calculate the mean and standard deviation of a back's runs and then calculate the probability that they would gain 10 yards or more on a series of 3 carries. Since you've got the data set, go ahead and let me know. 😄 I sent you the data. Let me know if you can copy it in that format. I originally went down the same path you outlined and calculated deviations and probabilities. I think that is also useful data. I came up with the RYPC because it is based on yards and not deviations and probabilities and so I figured it would be more tangible. Also, large runs, although good, tend to skew the deviations to an excessive degree for these purposes. With RYPC, the large runs do not impact it beyond adding a single count to the average.
October 10, 20187 yr I got rounded values of 75% probability for Ingram and 60% probability for Watson of three runs gaining 10 yards or more (74.86 and 60.34 percent, respectively, out to two decimal places).
October 10, 20187 yr Wouldn’t MEDIAN yards per carry kind of do what you want without having to invent a new statistic?
October 10, 20187 yr How hard would it be for one of you guys to expand this to all backs nationally against all FBS-only opponents? Same for QBs, also. Also also, how hard it would be to do something similar for WRs? Seems like this could aid in projecting outcomes for teams or fantasy players in the future.
October 10, 20187 yr 23 minutes ago, XYZ said: Wouldn’t MEDIAN yards per carry kind of do what you want without having to invent a new statistic? Median isn't a very accurate measure for this purpose. For example, and RB with the line 1,1,1,1,1,1,5,5,5,5,5 has a median run of 1 yard, while an RB with the line 1,1,1,1,1,5,5,5,5,5,5 has a median run of 5 yards. You can play around with some other lines and see that it isn't very useful. 1 minute ago, closetojumping said: How hard would it be for one of you guys to expand this to all backs nationally against all FBS-only opponents? Same for QBs, also. Also also, how hard it would be to do something similar for WRs? Seems like this could aid in projecting outcomes for teams or fantasy players in the future. The hard part is just getting the data in copyable table form. Otherwise, you have to go to the play by play for each game and input the data one at a time. For the entire league it would take many hours. Maybe some stats guys know where this type of data can be found. I've got other statistical analysis and indicators that I think are much better than those that are commonly given. I'd post a lot more analysis if I had accessible data. I'm assuming the same is true for Huckleberry.
October 10, 20187 yr Good metric. Aleph: Wouldn't it be better to use RYPC as a pivot instead of YPC? RYPC = YPC - [sum_1 to n (if(carry k -RYPC < 0, then RYPC - carry k, if not then 0)]/n. Not sorry for the iteration. It just makes sense that a run over your RYPC would always increase your RYPC. Edited October 10, 20187 yr by JBJ
October 10, 20187 yr http://www.cfbstats.com/ sales the data in bulk (I used it some time ago to learn SAS/SQL when it was free) but you can get individual game play by play results from http://sports.snoozle.net/search/fbs/index.jsp
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.