Comparing stats across eras

[forgetful]

I wanted to discuss these elements further in a non-player specific manner.  Modern analysts rely heavily on statistics when examining players and their place historically.  However, there are immense flaws in using many of these statistics pre-1986.  I use that as a reference as the first year 3-pt shooting percentages rose above 30% and they show a marked variation in PER.

For example.  From 1986-2016 the average PER for the league leader was 30.  During that time period 50% of the years (15 times) the leader had a PER over 30.

From 1956-1986 the average PER for the league was 27.  If we use a 95% confidence level and throw out the Wilt outliers in 62-64, the average drops to 26.6.  Wilt Chamberlin was the only person (from 62-64) that had a PER over 30 during those 30 years. 

There is a difference in PER of 3.4 units following 1986, because of how the game changed. 

Statistically, the data sets cannot be compared.  The 3.4 unit value is 3 standard deviations away from the 1986-2016 mean. 

PER is the worst of the advanced stats in comparing across eras.  Others, also suffer severely due to changes in the game.  When comparing players across leagues, different competitions, different eras, one has to look at performance relative to the era. 

I'm curious for the statniks out there? 

Which stats are most comparable across eras?

[brandx]

Which stats are most comparable across eras?

None.

When statheads really want to compare players across eras, they compare a player's stats with the league average of when that guy played. Of course there are variables/adjustments built in to account for the breadth of talent at the time (8 or 9 teams back in the day compared to 30 now).

Therefore, they can measure how much better Wilt was and how much better Jordan was using similar measurements.

I am a big baseball stat guy and this method will show that Ruth and Bonds are clearly the greatest hitters of all time. Of course Ruth was drunk and Bonds was on PEDs much of the time, but that is completely different argument.

[Tugg Speedman]

Time series analysis seems to work best in baseball  for two main reasons.

The game's rules are little changed in 100 years.  This includes equipment changes.  Baseball has not changed the bats (wooden) and the construction of the ball in a century.

Lots of obversations with a 162 game season.

Almost no other sport can say the same two reasons apply (track & field is the only other example I can think of).

To head off a potential responds, some sports like soccer have not changed their rules in decades.  But they don't have lots of observations in a season.

Other sports like golf and tennis have literally thousands of observations in a season but have tremendous changes (equipment) that make time series analysis hard.

[muwarrior69]

Time series analysis seems to work best in baseball  for two main reasons.

The game's rules are little changed in 100 years.  This includes equipment changes.  Baseball has not changed the bats (wooden) and the construction of the ball in a century.

Lots of obversations with a 162 game season.

Almost no other sport can say the same two reasons apply (track & field is the only other example I can think of).

To head off a potential responds, some sports like soccer have not changed their rules in decades.  But they don't have lots of observations in a season.

Other sports like golf and tennis have literally thousands of observations in a season but have tremendous changes (equipment) that make time series analysis hard.

The only caveat to that is during the '90s the players were juiced (steroids), whereas if you look at some of the players of previous generation they were also juiced (booze, not a performance enhancing drug) but still broke many records.

[Jay Bee]

PER doesn't exist in my world. Have never given it an ounce of attention.

[forgetful]

PER doesn't exist in my world. Have never given it an ounce of attention.

Which stats do you find most important?  I know you are a big statistics guy so am curious.

[forgetful]

None.

When statheads really want to compare players across eras, they compare a player's stats with the league average of when that guy played. Of course there are variables/adjustments built in to account for the breadth of talent at the time (8 or 9 teams back in the day compared to 30 now).

Therefore, they can measure how much better Wilt was and how much better Jordan was using similar measurements.

I am a big baseball stat guy and this method will show that Ruth and Bonds are clearly the greatest hitters of all time. Of course Ruth was drunk and Bonds was on PEDs much of the time, but that is completely different argument.

Personally, I believe that statistics are not as important in basketball.  But I can see how they would be very important for baseball...a lot more situational experiences where statistics can be very reliable.

[MU82]

Oftentimes, if we just use common sense and try to filter out our biases, we don't need stats -- especially not complex stats -- to make comparisons.

Take Ruth, for example. He changed the game. He was hitting home runs when nobody else was, and it took several years for the others to catch on. In 1927, Ruth famously hit 60 home runs. Gehrig hit 47. Nobody else hit more than 18. Think about that! It would be as if, in 1998 McGwire hit 70, Sosa hit 66 and nobody else hit 25.

Add in the fact that Ruth also was one of the best pitchers of his time and it's such a slam-dunk on him being the greatest ever that whoever your No. 2 is -- Mays, Bonds, Cobb, Gehrig whoever -- is MILES behind.

Whenever somebody says something like, "Bonds actually has surpassed Ruth as a ballplayer," I'll respond: "Yeah, how many games did Bonds win as a pitcher? What was his ERA? How many home runs did he hit relative to the rest of the players of his time?"

I mean, it's not CLOSE!

The fun LeBron-Bird debate we had in another thread ... I closely followed both of their careers and I use the eye test to say LeBron is a better player than Bird was; the stats simply support that IMHO.

Now, others use the "made teammates better test" and the really arbitrary "toughness test" to declare Bird better. That's OK. This one is closer than Ruth-vs-Anybody. Besides, it's just opinions, a friendly (I hope) debate. Stats are almost irrelevant.