Like many of us, I'm very, very familiar with the stats of 1908 Addie Joss.  So familiar that when I noticed 2014 Clayton Kershaw's ERC# -- 1.61 -- I immediately thought, "Hey, same ERC# as Joss!"  

I compared their other stats (Joss listed 1st, Kershaw 2nd):

OVA# -- .210 vs. .203
WHIP# -- 0.90 vs. 0.90
HR/9# -- 0.10 vs. 0.35
K/9# -- 3.82 vs. 6.69
BB/9# -- 0.99 vs. 1.50

I'm not sure who's better based on the above -- I guess Joss if you were facing HR hitters in a HR park, Kershaw if you wanted strikeouts in a singles park.  Either way, it's close.  Yet Joss is nearly $5,000 an inning cheaper ($37,363 vs. $42,354 per IP).

If you play mostly Open Leagues and your primary concern is winning, why would you ever take Kershaw over Joss?  

You wouldn't in OLs, because HRs kill.

You take Kershaw in theme leagues when you know your defense is going to suck and you need to suppress balls in play...

Posted by crazystengel on 10/5/2014 3:23:00 PM (view original):
Like many of us, I'm very, very familiar with the stats of 1908 Addie Joss.  So familiar that when I noticed 2014 Clayton Kershaw's ERC# -- 1.61 -- I immediately thought, "Hey, same ERC# as Joss!"  

I compared their other stats (Joss listed 1st, Kershaw 2nd):

OVA# -- .210 vs. .203
WHIP# -- 0.90 vs. 0.90
HR/9# -- 0.10 vs. 0.35
K/9# -- 3.82 vs. 6.69
BB/9# -- 0.99 vs. 1.50

I'm not sure who's better based on the above -- I guess Joss if you were facing HR hitters in a HR park, Kershaw if you wanted strikeouts in a singles park.  Either way, it's close.  Yet Joss is nearly $5,000 an inning cheaper ($37,363 vs. $42,354 per IP).

If you play mostly Open Leagues and your primary concern is winning, why would you ever take Kershaw over Joss?  

Precisely why Joss is a cookie. He's under priced.

Someone with better data analysis skills could surely reverse-engineer the primary components of the pitcher pricing formula, but it seems fairly obvious that K's cost more in that formula and in this case are probably the main reason for the discrepancy in $/IP for Joss and Kershaw (and just for deadball pitchers vs. modern pitchers in general).  We also know that the + stats are used in determining AB outcomes, and Kershaw has the higher OAV+ and HR/9+, among others, than Joss.  Assuming the relevant + stats would also carry weight in that pricing structure, that would be further explanation for the discrepancy.

I think it's better to think of modern pitchers as being overpriced than deadball pitchers being underpriced.  The raw effect of K/9 on $/IP seems out of line with the relative weights of other pitching stats in the pricing formula.  This is all just speculation though; I'd love to see the results of a deeper analysis. 

I think the # stats are useful as a general reference, but they compare the stat to the historical average.  The actual individual AB outcomes use the + stats of the hitter and pitcher in combination with the raw stats to determine hit or out, so how the hitter or pitcher fared against his own peers from that same season is what carries the weight in the actual PA outcome, and I suspect is weighted accordingly in the salary formula. 

I tried searching for the Event Tree thread where the Power Point slides could be found, but they no longer appear to be available.  The thread is there, just the pictures can't be seen anymore.  Hopefully, someone has saved a screenshot and still has them available.  But in one of the slides we saw a calculation of Pedro versus Ruth and what stats were used.  In determining hit versus out, Pedro's OAV was adjusted using his OAV+, and Ruth's AVG was adjusted using his AVG+, and then the result of the calculation determined hit or out.  There was a handedness adjustment for Ruth (LHB vs. RHP, I think it was 4.5%), but no # stats were to be found.  Just raw OAV/AVG and OAV+/AVG+.  

If I remember the slide correctly, the calculation was 2000 Pedro (.167 raw OAV) and 1923 Ruth (.393 raw AVG).  That Ruth's batting average against that Pedro was determined to be .261.  

Just comparing Joss to Kershaw in HR/9, Joss is far superior in suppressing HR.  The raw HR/9 is still the biggest determinant of the outcome, it's just adjusted in the calculation by the HR/9+.  Joss' HR/9+ is still much better than 100 (neutral), so his HR suppressing ability is much greater than Kershaw's.  Kershaw is just considered to be better relative to his peers than Joss, which only means that Kershaw would be better suppressing HR of 2014 hitters than Joss would be at suppressing HR of 1908 hitters.  But in terms of suppressing HR from all seasons, Joss wins by a lot.

I just suspect that the + stat is used in the pricing formula, so that even though Joss is better overall at suppressing HR across all eras, because Kershaw is better relative to his own peers, that single component of the pricing formula adds more $/IP to Kershaw than it does to Joss.  The amount may be small though, I really don't know.  Of course, I would hope that the salary formula weighs the raw HR/9 more heavily, and that edge certainly goes to Joss.  

Since K/9 affects pitch count, it is not in itself unreasonable that you pay for high K/9 pitchers. The problem is that K/9 should be normalized era-wise in the first place, and so should HR/9. In other words, if Joss gave up 1 HR per season and the average was 5, he was excellent on HR/9 compared with the league average, but not as much as a pitcher from say the 1990s who gave up 0.3 HR/9 when the league average may have been 30 HRs a season. That is where it seems the normalization effect on pricing has broken down, leading to the over-pricing of modern pitchers (the debate about this reminds of the old Marxist economics question about over-production or under-consumption being the causes of recessions ). 

So while charging by the number of pitches/9 makes a certain amount of sense, if that is not also being normalized, so that pitchers who pitched in low pitch count (because low K/9 and HR/9 numbers were typical in the deadball era) are not given a bonus for the era they lived in any more than steroid-fortified batters from the late 90s should be (or are apparently) then we get an over-pricing of Kershaw compared with Joss. 

One wonders about something else as well: what if pricing were based not on RL baseball results (though performance of course should be, properly normalized) but rather on LISTEN UP MY LIBERTARIAN FRIENDS - on supply and demand here at WIS?  That is,the more Joss gets used the more he costs, as would every other player. This seems like a simple, elegant solution. Either that or we elect boogerlips to be Big Brother state bureaucracy and he can arbitrarily set the prices for each player and each outcome as he sees fit. (where is he by the way? he hasn't answered any of my provocations for weeks now). 

Market-based pricing has been discussed often on this site, but ultimately the conclusion is usually that it probably won't work because pricing needs to be consistent to be able to run leagues with an reliable expected quality by cap.

On the one hand, you could limit market-pricing to Open Leagues.  On the other, you could have updates more than once every freakin' millennium ...

The # stats allow you to compare players without having to go through the work of using the raw and + stats.

Joss is underpriced, but there are two things that make this a little closer than one might think.

First, Kershaw costs less per pitch than Joss. He will give you 3191 WIS pitches at $2,632/pitch, while Joss gives you 4801 pitches at $2,661/pitch. Of course, since Kershaw K's more, those pitches will be slightly less efficient, but it's something.

Second, the difference in fielding normalisation between 1908 and 2014 is pretty massive. If you have bad fielders to start with, the bad normalisation from Joss (plus the fact that Joss will make them work more) could tip the balance.

That said, I agree with crazystengal. I think the only circumstances when I'd take Kershaw over Joss are

(a) when having so many IP tied up in one pitcher was bad for team balance somehow, or
(b) when the opposing hitters were so high contact that Kershaw wasn't going to burn himself out K-ing them, and my fielders were so bad that they needed all the help they could get from normalisation.

But those are rare situations.

But can Joss beat the Cards in the playoffs?

Posted by pinotfan on 10/6/2014 1:59:00 PM (view original):
On the one hand, you could limit market-pricing to Open Leagues.  On the other, you could have updates more than once every freakin' millennium ...

...another six months have ticked by...

Again, the simplest solution to all of these player pricing issues on this site, and the lack of an update is (for once, yes I admit) a simple market solution: the prices of players should change daily as they are used in OLs. Period. This in one fell swoop would make the whole site more exciting and interesting, because from one team you build to the next you would not know how much Joss, Kershaw or Babe Ruth for that matter would cost - Ruth being too costly would lead to him not being used and then briefly to being somewhat more affordable and then costly again. It would keep us on our toes, and not really require any further updates for a long while. 

boogerlips, want to chime in on this one?

The ability of WIS to accurately price a player can be graphed as a bell curve. WIS generally has done an outstanding job of pricing players, the majority of which would fall in the tall, central part of the graph indicating an accurate price. There are a few players though that inevitably through human error will be a bit overpriced or underpriced. These represent the much shorter outskirts of the bell graph.

Instead of relying solely on complex math to solve the remaining inaccuracies in pricing, it makes good sense to listen to the market also and let it do much of the work for you. If certain players are being used an extraordinarily high percentage of the time, the odds are very high that that player is indeed underpriced. If certain player are never used, then odds are also high that those players are underpriced (although here you would have to differentiate between underpriced and under-useful).

The degree to which price should increase should correlate with the bell curve. I believe the variables needed for input could be calculated based on usage. Someone more familiar with pre-calculus would have to head that project. The idea is to turn the bell shaped usage into a thin, parabola shaped usage chart.

Good sample size would be needed for future adjustments. Once a quarter might be a nice target for the pricing changes.

The goal should be to achieve accuracy. If the price changes too often and/or is changed randomly or at least not based on anything concrete, then I think that makes the game worse. There is an inherent trust that SLB users have in the players being priced fairly. This then allows the user to focus on strategy rather than bargain hunting.

Bargain hunting is certainly what has killed the charm of open leagues. I might have an idea on how to score runs more efficiently using a compbination of certain skill types, but if those certain skill types aren't labeled "Dilone" or "Weiss" or "Joss" then my creativity is at a disadvantage. It's like we're battling to see who found the best coupon instead of the best concept. BORING.