[DEV] ISFL Forums
*TPE Allocation Correlation for On Field Success - Printable Version

+- [DEV] ISFL Forums (http://dev.sim-football.com/forums)
+-- Forum: Community (http://dev.sim-football.com/forums/forumdisplay.php?fid=5)
+--- Forum: Media (http://dev.sim-football.com/forums/forumdisplay.php?fid=37)
+---- Forum: Graded Statistical Analysis (http://dev.sim-football.com/forums/forumdisplay.php?fid=153)
+---- Thread: *TPE Allocation Correlation for On Field Success (/showthread.php?tid=9189)

Pages: 1 2

*TPE Allocation Correlation for On Field Success - nunccoepi - 06-19-2018

Alright, so a while ago I posted a statistical analysis of tpe allocation in terms of on field success by calculating the correlation coefficient for each position's attribute totals and their statistical output. In lay man's terms, this means that I took a very simple look at how each of the positions' skill areas (speed, tackling, endurance, etc.) related to their stats (tackles, sacks, interceptions, etc). I did this for all the offensive positions. Here, now that real life work has slowed down a lot and I've caught up with other league responsibilities, I've finally followed up with the same analysis for the defensive positions.

Like before, I used a season's worth of data, only this time it was Season 7. I also didn't use DSFL data as a player with lower tpe will do "better" due to the lower level of competition and was thus throwing off the results. The good part about looking at the defense is that all the stat areas are basically the same (instead of neglecting to measure a QB's rushing ability or a RB's receiving ability, for example) and most players play in their own position. Also, all the players I measured played in all 14 games. The downside is that some of the most important and game-changing stats for defense don't happen all that often--defensive touchdowns, interceptions, safeties, and even sacks. Unfortunately, that means that this correlational measure won't be able to accurately demonstrate what we are looking for in those areas. So, in most cases, I left them out.

So, without further ado and a timely reminder that correlation does not imply causation, let's jump into it.

Safety first.

[Image: LA4Hj3e.jpg]

Remember, the higher the number (the closer it gets to 1), the stronger the correlation between the two measures. Also, if you read the last one I wrote, you'll notice that I included averages beneath each attribute to assist in measuring the importance of updating each attribute.

Basically, for safeties, it would seem that Speed, Endurance, and Hands are their most important attributes. Looking at the sacks and tackles numbers, however, I wonder if the averages are brought down by the other areas. Agility and strength would be important if a safety were to be used for blitzes, it seems.

Here's another more visually appealing way to look at the information.

[Image: C3l5y3S.jpg]

Next up, let's look at Cornerbacks.

[Image: 0bRo4EM.jpg]

Intelligence is clearly a very important attribute. Did you know that it would be that important for CBs? I didn't. It looks like the most important attribute for almost all the stat areas for CBs although if you are chasing interceptions--and who isn't--there are several other areas almost just as important--notably, speed.

For the visual learners:

[Image: LJahY4Z.jpg]

Now, we have the golden boys of the defense, the Linebackers.

[Image: VZeQ0jW.jpg]

Here we see a lot of green, similar to the safeties. If I had to guess, one of the reasons that we see a lot of green (meaning low or even negative correlation) on defense might point to those stat areas being controlled more by offensive skill sets, defensive scheme, or offensive playbooks. Of course, the green may be an error caused by just simply too small of a sample size for those stat areas (rarely would players force more than 2 fumbles, for example, and many forced none). Nevertheless, there is still information to be gleaned from this one as well. Linebackers are kings of the tackles--and the sacks. A lot of positive correlations for those stats, particularly in the attributes of strength and agility. Once again, endurance is important for a valuable stat: sacks. That may be because teams will give up more sacks towards the end of the game when their endurance is lower, therefore linebackers with higher endurance will have an easier time. That may be giving the sim too much credit, however. A final thought on this does have to do with a negative correlation that may have some merit. The negative correlation between the tackling attribute and forced fumbles is pretty strong. That means that the better a player is at tackling, the fewer fumbles he will cause. I wonder if this was intentional or not. Someone needs to teach these linebackers the meaning of a strip tackle.

[Image: yD1a1iS.jpg]

Alright, on to the Defensive Ends, my personal favorite position to play.

[Image: ygimL5P.jpg]

For a sim that so heavily favors speed, I was surprised to see that dark green column. In fact, I even went back and double checked that I had my formulas right (I did). Basically, I think that the reason that the green was so pronounced is that almost all the DEs have maxed speed (either 80 or 85). There just isnt as much variability there to show a dramatic change in stats. So, in the long run, speed is probably more important than represented here, but its hard to say exactly how much. Otherwise, tackling, intelligence, and probably hands are the three next most important attributes to upgrade if you are at the Defensive End position, especially if you are looking to be a turnover-causing machine.

[Image: AhJYcwO.jpg]

And finally, here are the Defensive Tackles.

[Image: WSeJax8.jpg]

My formulas are correct here again despite that green row in tackles for a loss. For some reason, last season, poorer DTs just had a lot more tackles for a loss than better ones. Consider this: the two worst DTs in terms of overall rating (E. Medved at 76 overall and B. Wojcik at 72 overall) combined for 12 tackles for a loss on the season (in fact they were two of the top three in this category). However the top FOUR DTs in overall rating (Kamaka at 100, Miller at 99, Clegane at 98, Askins at 94) COMBINED for only half--6 tackles for a loss. That one is just confusing to me, to be perfectly honest. I am inclined to think that is more than dumb luck or a small sample size. I wonder if there could be something there, especially since you see a similar (but weaker) trend in the DEs--who knows with our sim. The rest of the stats give a clearer picture, however with Strength, Intelligence, Endurance, and Tackling as the more important for overall tackles and sacks. If you want to cause more fumblerooskis, go for Speed and Hands as is consistent with what other positions have shown us.

[Image: GZ6He4E.jpg]

Hope this was illuminating for you. I certainly enjoyed doing it. Of course, there is a lot of room for error and inconsistencies. I think I'd noted most of them either in this article or the previous one. If you have any questions or suggestions, let me know!

*TPE Allocation Correlation for On Field Success - timeconsumer - 06-19-2018

I have done some of my own analysis on the side except using 40 years worth of CPU generated sims as my dataset and we're getting a lot of similar stuff, but we split in other ways as well. One thing I think that is rather important to point out is with linebacker sacks. Look at some of the top LBs in sacks for S7 and you see Winchester and Tweed, both who were maxed in every stat, and Sharpei who was very high in them. That alone is going to skew the data about endurance, but it is an important stat. The other really difficult thing with linebackers and sacks has a lot to do with their scheme and what spot they played in, which largely in S7 for top sacking LBs was SLB in nickel, it's where the stats came from.

Now here's a really fun stat. Using my system of 40 years of CPU data I wanted to see what effect tackling had on missed tackles. So take tackles/(tackles + missed tackles) and you get the rate of a successful tackle, make sense? Here's where it got weird, the higher a player's tackling stat, the more likely they were to miss a tackle. WTF right? We're talking maybe 96% success vs 98% so not a ton, but still head scratchingly weird.

*TPE Allocation Correlation for On Field Success - nunccoepi - 06-19-2018

(06-19-2018, 02:48 PM)timeconsumer Wrote:I have done some of my own analysis on the side except using 40 years worth of CPU generated sims as my dataset and we're getting a lot of similar stuff, but we split in other ways as well. One thing I think that is rather important to point out is with linebacker sacks. Look at some of the top LBs in sacks for S7 and you see Winchester and Tweed, both who were maxed in every stat, and Sharpei who was very high in them. That alone is going to skew the data about endurance, but it is an important stat. The other really difficult thing with linebackers and sacks has a lot to do with their scheme and what spot they played in, which largely in S7 for top sacking LBs was SLB in nickel, it's where the stats came from.

Right. Top LBs will probably end up playing SLB where there are more sacks. Whats more, endurance tends to be one of the last attributes people will pay attention to or at least try to max. It may just be that the ones who have allocated a lot in their endurance have also allocated a lot elsewhere (because its the last of their stats to max) so the stats have a higher correlation. So, not necessarily because endurance is high, but because people who have high endurance are also strong in other attributes.


(06-19-2018, 02:48 PM)timeconsumer Wrote:Now here's a really fun stat. Using my system of 40 years of CPU data I wanted to see what effect tackling had on missed tackles. So take tackles/(tackles + missed tackles) and you get the rate of a successful tackle, make sense? Here's where it got weird, the higher a player's tackling stat, the more likely they were to miss a tackle. WTF right? We're talking maybe 96% success vs 98% so not a ton, but still head scratchingly weird.

That is weird. Any thoughts on why those with better tackling tend to recover fewer fumbles?


*TPE Allocation Correlation for On Field Success - timeconsumer - 06-19-2018

(06-19-2018, 03:59 PM)nunccoepi Wrote:That is weird. Any thoughts on why those with better tackling tend to recover fewer fumbles?

Nothing I've really looked in to. If I had to guess I'd just say it's noise.

*TPE Allocation Correlation for On Field Success - kckolbe - 06-19-2018

(06-19-2018, 12:48 PM)timeconsumer Wrote:Here's where it got weird, the higher a player's tackling stat, the more likely they were to miss a tackle. WTF right? We're talking maybe 96% success vs 98% so not a ton, but still head scratchingly weird.

Was this just LBs or all defense? Because, in the case of DTs, who have high tackling, max agility is painfully low. If this was a defense-wide revelation, I would look at them first. In the case of CBs (especially m2m), strength is a low attribute.

*TPE Allocation Correlation for On Field Success - timeconsumer - 06-19-2018

(06-19-2018, 04:08 PM)kckolbe Wrote:Was this just LBs or all defense?  Because, in the case of DTs, who have high tackling, max agility is painfully low.  If this was a defense-wide revelation, I would look at them first.  In the case of CBs (especially m2m), strength is a low attribute.

Both. And this wasn't based upon our archetypes, this was the sim randomly generating it's own players and simming 32 teams for 40 years.

*TPE Allocation Correlation for On Field Success - kckolbe - 06-19-2018

(06-19-2018, 01:09 PM)timeconsumer Wrote:Both. And this wasn't based upon our archetypes, this was the sim randomly generating it's own players and simming 32 teams for 40 years.

Oh wow. That's a lot of potential variables then. Out of curiosity, did higher tackling result in greater TFL numbers? If so, that would indicate that higher tackling attribute meant more aggressive tackling, which could explain the higher missed tackle percentage.

*TPE Allocation Correlation for On Field Success - Beaver - 06-19-2018

Interesting stuff. What I was going to say about endurance has already been said but this pretty much tracks with what I've heard/seen building my player.

*TPE Allocation Correlation for On Field Success - timeconsumer - 06-19-2018

(06-19-2018, 04:57 PM)kckolbe Wrote:Oh wow.  That's a lot of potential variables then.  Out of curiosity, did higher tackling result in greater TFL numbers?  If so, that would indicate that higher tackling attribute meant more aggressive tackling, which could explain the higher missed tackle percentage.

Logical assumption but I really don't think the sim is that intelligent to factor things like that in. My guess is that higher tackling is associated with higher stats in general, and higher stats is more correlated with higher snap counts, and higher snap counts means more of an opportunity for fatigue to lessen performance.

*TPE Allocation Correlation for On Field Success - kckolbe - 06-19-2018

(06-19-2018, 03:43 PM)timeconsumer Wrote:Logical assumption but I really don't think the sim is that intelligent to factor things like that in. My guess is that higher tackling is associated with higher stats in general, and higher stats is more correlated with higher snap counts, and higher snap counts means more of an opportunity for fatigue to lessen performance.

I fear you're right. It's also possible opportunity cost is a factor? I have no clue how players are auto-generated, but higher tackling could mean lower values in str/agi/end, but that is on an assumption that auto-gen players with a higher att have something lower to compensate.