DMN Blog: Cowboys visiting with Igor Olshansky

[AdamJT13]

But the argument here wasn't about tackle stats - the issue was that I noted that solo tackle stats are probably a more consistent measure than are overall tackles (i.e., those that include assists) because when a guy is credited with a solo it more often than not means he was clearly involved in the tackle.

If you look only at the "solo" tackles, you're punishing players who play in games where the statisticians award two assists on one play instead of awarding one solo and one assist. That's not to mention the mere randomness of some solo tackles being credited (such as in pileups) or the inherent margin of error.

 

[AbeBeta]

It's useless as a direct comparison between players on different teams.

Consider a scenario in which Player A on Team X is credited with 50 "solo" tackles and 25 assists, and Player B on Team Y is credited with 40 "solo" tackles and 25 assists. Even if we assume that they had the same number of opportunities, can you accurately say that Player A was more involved in tackles than Player B? No, you can't, because the stadium statistician in Player A's home games and some road games might have been ones who credit a "solo tackle" and "assist" on the same plays, and might be much more likely to credit multiple players on the same plays. The statisticians in the majority of Player B's games might be much more likely to credit two assists instead of a solo and an assist or much more likely to credit only one player on a play, even if multiple players were involved.

Even if neither player was involved in any plays when the wrong guy was credited, the mere differences in how tackles are recorded could result in something like this being actually what happened --

Player A
30 times made the tackle by himself
20 times was credited with a "solo" tackle when he got there first but had help on the tackle
5 times was credited with an assist when he got there first but had help on the tackle
20 times was credited with an assist when he was the second tackler involved
5 times was not credited with an assist when he was the second tackler involved


Player B
35 times made the tackle by himself
5 times was credited with a "solo" tackle when he got there first but had help on the tackle
17 times was credited with an assist when he got there first but had help on the tackle
8 times was credited with an assist when he was the second tackler involved
20 times was not credited with an assist when he was the second tackler involved


As you can see, Player B actually was more involved, but the method the various statisticians used for crediting tackles makes it appear that Player A was more involved.

First, the issue I was discussion was SOLO tackles -- focus on that.
The scenarios are irrelevant to the issue - my point was that unless the inconsistencies you present -- RELEVANT TO SOLO TACKLES -- occur so often that they make it impossible to predict what sort of scoring decision will be made most of the time with regard to whether a player gets credited with a solo tackle or not, then the statistic retains value.

I think one issue that you and several others have here is the belief that a) measures are or should be error free and b) how to make evaluations using less than perfect measures.

 

[AbeBeta]

If you look only at the "solo" tackles, you're punishing players who play in games where the statisticians award two assists on one play instead of awarding one solo and one assist. That's not to mention the mere randomness of some solo tackles being credited (such as in pileups) or the inherent margin of error.

Again, my point was if someone was credited with a solo tackle, you can be pretty sure that he clearly made a play -- there are of course errors that you get like the pileup solo tackle but what you suggest here are errors that go in both directions. And over the course of a long season where a guy participates in #s of plays, you'll likely see the plus and minus even out a good bit on those.

Also, see my earlier comment about how one might use these values to more validly compare across players (e.g., % the d-line's of solo tackles by team). You start to standardize measures like that and you get a far better picture that accounts for much of the variation across teams.

 

[gmoney112]

This discussion has gone way too far.

 

[AdamJT13]

First, the issue I was discussion was SOLO tackles -- focus on that.

You're missing the fact that what is credited as a "solo" tackle by one statistician is credited as an "assist" by another statistician.

The scenarios are irrelevant to the issue - my point was that unless the inconsistencies you present -- RELEVANT TO SOLO TACKLES -- occur so often that they make it impossible to predict what sort of scoring decision will be made most of the time with regard to whether a player gets credited with a solo tackle or not, then the statistic retains value.

So, a margin of error that is less than "most of the time" is OK with you?

I think one issue that you and several others have here is the belief that a) measures are or should be error free

Nobody ever said that. The question is what is an acceptable margin of error. And there's a difference between a margin of error and inconsistency in the methods used.

Given the small numbers involved, the margin of error alone makes tackles a dubious statistic, but the differences in methods used clearly makes them unreliable. The NFL itself has stated this.

 

[AbeBeta]

You're missing the fact that what is credited as a "solo" tackle by one statistician is credited as an "assist" by another statistician.

I understand that -- but you are missing the fact that we need to know how many are doing that. And again, the argument was that if there is a solo tackle credited it almost always means the player had significant involvement.

So, a margin of error that is less than "most of the time" is OK with you?

I was dumbing it down

Nobody ever said that. The question is what is an acceptable margin of error. And there's a difference between a margin of error and inconsistency in the methods used.

Given the small numbers involved, the margin of error alone makes tackles a dubious statistic, but the differences in methods used clearly makes them unreliable. The NFL itself has stated this.

But again, there are ways to address that error and make something out of those raw data rather than just dismissing the values as too error-ridden to be of use.

 

[FuzzyLumpkins]

Abe you go ahead and find out how many do each paradigm and try and show that the disparity doesn't create too much of a variance. Until then the stat is worthless.

 

[AbeBeta]

Abe you go ahead and find out how many do each paradigm and try and show that the disparity doesn't create too much of a variance. Until then the stat is worthless.

I'm not the only claiming that the stat is inaccurate because of that issue and I never called for using the stat to DIRECTLY compare players, I suggested ways that you would reduce considerably any measurement error that came into play and noted what the solo tackle stat DOES tell you.

Of course, I wouldn't expect you to understand anything that complex or nuanced.

 

[Avenging Hayseed]

WE DONT CARE !! All we care about is if were gonna SIGN HIM or not! Hope we do.

 

[AdamJT13]

Again, my point was if someone was credited with a solo tackle, you can be pretty sure that he clearly made a play

That's all well and good, but the question was whether solo tackles are useful for comparing players on different teams. If one player is credited with 40 solo tackles, he most likely (given the normal margin of error) did something on those 40 plays. If another player is credited with 35 solo tackles, he most likely did something on those 35 plays. But can you compare the two and say the first guy made more plays? Absolutely not, because as I showed in the scenario you ignored, the second guy might have had many more times when he was credited with an "assist" instead of a "solo" tackle for doing the EXACT SAME THING that the first guy did to be credited for many of his "solo" tackles.

Also, see my earlier comment about how one might use these values to more validly compare across players (e.g., % the d-line's of solo tackles by team). You start to standardize measures like that and you get a far better picture that accounts for much of the variation across teams.

But it doesn't account for the difference in the quality of a player's teammates. Just because a player is better than his teammates, it doesn't make him better than a player on a different team who isn't better than his own teammates. If Player A has 30 solo tackles on a defensive line that has 100 solo tackles, and Player B has 35 solo tackles on a defensive line that has 125 solo tackles, who was more involved in tackles?

 

[jobberone]

Is their a screaming in anguish emoticon?

 

[AbeBeta]

That's all well and good, but the question was whether solo tackles are useful for comparing players on different teams. If one player is credited with 40 solo tackles, he most likely (given the normal margin of error) did something on those 40 plays. If another player is credited with 35 solo tackles, he most likely did something on those 35 plays. But can you compare the two and say the first guy made more plays? Absolutely not, because as I showed in the scenario you ignored, the second guy might have had many more times when he was credited with an "assist" instead of a "solo" tackle for doing the EXACT SAME THING that the first guy did to be credited for many of his "solo" tackles.

Again, the solo tackle statistic itself is not useful for direct comparison -- you would have to ADJUST the value in some way to account for differences among teams and ADJUST your interpretation to account for a reasonable amount of error in measurement. For example 40 vs. 35 solo tackles shouldn't tell us a whole hell of a lot about differences between the players. But 40 vs. 20 would likely gives us some information about who tends to be involved.

But it doesn't account for the difference in the quality of a player's teammates. Just because a player is better than his teammates, it doesn't make him better than a player on a different team who isn't better than his own teammates. If Player A has 30 solo tackles on a defensive line that has 100 solo tackles, and Player B has 35 solo tackles on a defensive line that has 125 solo tackles, who was more involved in tackles?

What stats do you know of that meaningfully account for differences in the quality of player's teammates? Seems like any statistic needs to be interpreted in that context - regardless of what you are measuring. For example is an RB who gets 5.0 YPC running behind a great OL a better RB than the guy who get 4.5 running behind a terrible OL?

And to answer your question about those two players -- they are too similar in productivity to determine one was any more productive than the other. However comparing the % of awarded solo tackles does address some of the issues in crediting, showing that player A had 30% and B had 28%. Given the small #s, most reasonable folks would note those guys were involved at about the same level.

 

[AdamJT13]

But again, there are ways to address that error and make something out of those raw data rather than just dismissing the values as too error-ridden to be of use.

The way to do that would be to standardize the method used to record tackles and create at least four different categories of tackles. Better yet, they could record the "official" tackle stats only after a review of the game film.

For previous seasons, you'd have to go through the play-by-play while watching each play in order to determine how to correctly credit each type of tackle, but I don't think that would be worth the effort. It wouldn't even be worth the effort to look just at the game-by-game tendencies of the statisticians and adjust for those.

Tackle stats are fairly valid and useful for comparing players on the same teams in the same games, but not very valid or useful for comparing players on different teams in different games. Until the system changes, that'll always be the case.

 

[AbeBeta]

The way to do that would be to standardize the method used to record tackles and create at least four different categories of tackles. Better yet, they could record the "official" tackle stats only after a review of the game film.

For previous seasons, you'd have to go through the play-by-play while watching each play in order to determine how to correctly credit each type of tackle, but I don't think that would be worth the effort. It wouldn't even be worth the effort to look just at the game-by-game tendencies of the statisticians and adjust for those.

Tackle stats are fairly valid and useful for comparing players on the same teams in the same games, but not very valid or useful for comparing players on different teams in different games. Until the system changes, that'll always be the case.

Again, I think we differ in terms of how accurate the statistic has to be to be valid -- in many fields, you'll find measures that produce considerable variations around a true score which is clearly the case here. Many fields use far less accurate metrics to make decisions, knowing that any decisions must be tempered by an appropriate understanding of the limitations of the measures.

Of course much of the error in measurement discussed can be accounted for as I explained (or even more finely tuned by doing it on a game by game basis) but obviously not enough to make the metric close enough to satisfy you.

 

[Avenging Hayseed]

:bang2::horse:That horse been dead wayyyyyyyyyy long ago already!

 

[AbeBeta]

:bang2::horse:That horse been dead wayyyyyyyyyy long ago already!

Actually Adam and I are having what I believe is a productive discussion. Clearly though, it was a grade level or twenty higher than you are comfortable.

 

[Manwiththeplan]

I think you guys may be overrating him just a bit if you think he's a huge improvement over Canty.

6 in one hand, half a dozen in another if you ask me. I'd even go as far as saying Canty is more talented, but I have no reason to believe Igor can't match his production or even exceed it.

 

[Avenging Hayseed]

Actually Adam and I are having what I believe is a productive discussion. Clearly though, it was a grade level or twenty higher than you are comfortable.

Keep your high heals on and dont get yourself all in a twitter. Im only poking sticks atcha out of boredom waiting to see if were gonna sign this guy or not.

 

[AdamJT13]

Again, the solo tackle statistic itself is not useful for direct comparison -- you would have to ADJUST the value in some way to account for differences among teams and ADJUST your interpretation to account for a reasonable amount of error in measurement. For example 40 vs. 35 solo tackles shouldn't tell us a whole **** of a lot about differences between the players. But 40 vs. 20 would likely gives us some information about who tends to be involved.

Yes, you probably can glean some information when there is a huge discrepancy in the numbers.

And yes, you would have to adjust the numbers to account for the different methods used to record them.

What stats do you know of that meaningfully account for differences in the quality of player's teammates? Seems like any statistic needs to be interpreted in that context - regardless of what you are measuring. For example is an RB who gets 5.0 YPC running behind a great OL a better RB than the guy who get 4.5 running behind a terrible OL?

That's much different from what you proposed -- judging players on different teams based on comparing them to their teammates. Would you judge receivers based on their percentage of their teams' wide receiver corps' receiving yards? Larry Fitzgerald had 29 percent of his team's receiving yards. Braylon Edwards had 34 percent of his. Does that mean Edwards (873 yards) was somehow superior, more productive or more involved in the passing game than Fitzgerald (1,431 yards)?

Or would you judge a defensive back based on his percentage of his secondary's interceptions? Terence Newman had 57 percent of his secondary's interceptions, and Ed Reed had 50 percent of his. Does that make Newman (four INTs) more of a ballhawk than Reed (nine)?

And to answer your question about those two players -- they are too similar in productivity to determine one was any more productive than the other. However comparing the % of awarded solo tackles does address some of the issues in crediting, showing that player A had 30% and B had 28%. Given the small #s, most reasonable folks would note those guys were involved at about the same level.

OK, so we've established that a 17 percent difference in the raw tackle numbers could be entirely irrelevant, without even taking into account the flaws of the numbers themselves.

How about a 25 percent difference? If Player A had 40 "solo" tackles on a defensive line with 135, and Player B had 30 "solo" tackles on a defensive line with 100, who was more involved in tackles? Or were they "about the same"?

Shall we keep going? Player A had 28 "solo" tackles on a defensive line with 125. Player B had 20 solo tackles on a defensive line with 90. Who was more involved in tackles? Still "about the same"?

 

[Woods]

Keep your high heals on and dont get yourself all in a twitter. Im only poking sticks atcha out of boredom waiting to see if were gonna sign this guy or not.

I really think we should sign this guy.

Reasons:

1. Wade knows him very well. His strengths and weaknesses.

2. We have a need on the DL after losing Canty and parting ways with Tank.

3. We brought him in for a 2 day visit. That's a lot of time to spend on a potential recruit. Clearly, the team is very interested.

4. We DO have enough money under the cap, even taking into account Ware, to pick Igor up for $5-6mm per year.

There, I feel more confident now that we'll sign him. I've convinced myself.