Week 18: That ‘Stupid’ Net Wins Stat

 

Michael presented at an industry conference last week in New Orleans on the role of the Next Generation in the Nuclear Industry. Yes, he’s that old that he now talks about the next generation.

2022 Year to Date Power Ranks
TOTAL			HITTING			PITCHING
1	Arthur	3.93	1	Cory	3.25	1	Dean	3.00
2	Dean	4.00	2	Arthur	3.50	2	Michael	3.50
3	Cory	4.29	3	Carl	4.50	3	Brian/Josh	3.83
4	Michael	4.43	4	Dean	4.75	4	Carl	4.50
5	Carl	4.50	5	Michael	5.13	5	Arthur	4.50
6	Brian/Josh	5.57	6	Keith	5.13	6	Dave	5.50
7	Dave	5.71	7	Dave	5.88	7	Cory	5.67
8	Keith	5.93	8	Brian/Josh	6.88	8	Paul	6.83
9	Paul	7.43	9	Max	7.63	9	Keith	7.00
10	Max	8.43	10	Paul	7.88	10	Max	9.50

 

The power ranks continue to tighten at the top but a new name is now in third place, year to date: Cory. Cory is coming off an 18 home run, 105 R+RBI outburst…while still somehow coming up short against Dave in their matchup. The other big power ranks mover right now is Keith. Keith has gone from a 7.00 rank to this 5.93 rank in 5 weeks. It’s a big deal. The blog tried to write him off a few weeks ago and the hill to climb is steep, but after the events of this week, he is very much alive.

About last week, ho boy. The Yahoo Standings that were already tighter than normal, were turned on their head with three blowouts by teams lower in the Yahoo Standings and the top 3 teams all losing. For perspective on how tight things are, I looked back at every Floored end of year standings since 2007. Right now, Dean’s 7^th place, 0.526 winning percentage would normally be good enough for 4^th place. With the average 1^st place winning percentage is 0.595, this would lead to 4^th place being 20 games back of first after 18 weeks. Well, right now, 7^th place is a mere 13 games back of 1^st instead of 30 like we’d see in a normal year. Yes, this year is insane.

----

This week we’re going to put some data to a frustration I’ve been voicing intermittently throughout the year: the Net Win stat. Never one to not have data behind my opinions, if possible, and certainly one to keep an open mind to new ideas…I was kinda hoping to learn something.

Before I get into it, the disclaimers. I know how we got here. I know we didn’t think this was going to be a perfect situation when we made these rule changes this season. In many ways, the rule changes have been successful: streaming has been far lessened in value (though lately some teams have been able to do so successfully), the 6 AM early birds have had far less of an advantage, and innings pitched leaders without good ratios haven’t had the advantage they’ve had in previous years.

…but can we do better? Let’s talk about it.

The best way I thought to analyze this was to examine if the net win stat meant anything in terms of correlations with other successful pitching metrics. Think about it, if a player has a good home run total, he’s going to have a good slugging percentage or high totals of runs or RBI totals with enough At Bats. Yet, does a pitchers net win total mean anything? Cy Young voters recognized about a decade ago that the pitcher with the best win-loss record didn’t necessarily mean they were the best pitcher that year, with the tide shifting most notably when Felix Hernandez won the Cy Young in 2010 with a 13-12 record but a 2.27 ERA. David Price that year had a 19-6 record with a 2.72 ERA.

I ran data from all starting pitchers with at least 10 Innings Pitched (to get rid of ridiculous 0.00 or 15.00 ERAs over 4 innings that don’t help analysis), this got us 228 data points, note that I pulled the data August 10^th just because that’s when I ran the data (it takes me that long to write a good blog these days). The statistic I’m relying on here is R_squared. R_squared, as past blog readers will remember, in layman’s terms, is a measure of how close a set of data’s points are to the data set’s trendline. If there are a lot of outliers away from a trendline, that shows that the correlation isn’t strong and the R_squared value is low. A strong correlation looks like this. Think about ERA vs WHIP:

You see the R_squared to be 0.7643 or 76.4%. (for those wondering, that 19.85 ERA and 3.09 WHIP is Reiver Sanmartin for the Reds…he had a bad 11 innings this year)

I’ve heard arguments for all of the following options being indicative of a good way to find net wins. “Just pick the player on the good team” (note for players that were traded at the deadline like Frankie Montas, I used their previous team’s win% for the correlation since that’s where 90% of their stats were accrued); “just find the best pitchers.” Each of these makes you nod your head, “yeah, makes sense,’ but wait…

I ran a number of other correlations to the Net Win Stat and here is how they correlated:

Correlation to NW
	ERA	FIP	IP	WHIP	Team Win%	Team Bullpen ERA
R_squared	22.60%	21.10%	13.50%	24.60%	32.80%	12.20%

 

32%? 12%? Please

For perspective, I ran a few other correlations that felt weak in my head to see what the R_squared was. For the top 100 stolen base leaders I checked their Run totals: 8%. For the top 100 hitters in plate appears I checked their wOBA: 4%. I also checked one correlation on the hitting side that I felt was stronger, the top 100 hitters in Plate Appearances and their Run totals: 29%.

Net Wins when correlated to other notable pitching statistics is a weak correlation, but not zero. Want to win the net win stat? Get lucky…and maybe have your pitchers be on good teams. Anecdotally, look at this week’s output. Arthur had a 5.00 ERA and for 3 NW while Cory had a 3.84 ERA and -4 NW, and BJ had a 2.94 ERA (over 113 innings!) and got 2 NW.

The net win category isn’t TOTALLY random, it’s just not as well-correlated with pitching performance as I’d want it to be.

So, what do we do about it? idk

For leagues that have weekly lineup setting, the Innings Pitched stat makes sense. Everyone only gets 9 roster spots in those leagues each week to fill. If your 9 pitchers can throw 60 innings while someone else’s throws 50, that’s probably a good indication of pitcher performance. For reference, the R_squared of ERA vs IP per game started was 25%...so still not great. Strikeouts per 9 innings as an idea for a new category would lead to some redundancy with the strikeouts category. And on and on and on.

Ideally, we could find some other independent pitching category with at least some reasonable correlation to positive performances. We’ll keep looking. Until then, as I’ve said before, let’s just hope it doesn’t determine our league champion…because it is definitely going to have a role in who makes the playoffs.