New evaluation metrics

[acnunnally]

In an effort to apply some advanced analytics to SimD, at least for purposes of HOF voting, GM33 had the idea of using FanGraphs points used in Ottoneu FLB, which are an adaptation, on the pitching side, of SABR scoring. Ottoneu scoring is based on the following:

Hitting

AB: -1.0
H: +5.6
2B: +2.9
3B: +5.7
HR: +9.4
BB: +3.0
HBP: +3.0
SB: +1.9
CS: -2.8

Pitching

IP: +7.4
K: +2.0
H: -2.6
BB: -3.0
HBP: -3.0
HR: -12.3
Save: +5.0
Hold: +4.0

In its unmodified state, as applied to a career, this scoring system will over-reward longevity (based on accumulation of more points over time) and punish players who get less PT. After going back and forth a bit, GM33 and I settled on the idea that looking at both a cumulative metric, as well as a per 162 score for positional players and a per 33 games started score for SP would highlight absolute quality of performance as opposed to aggregation. Accordingly, we give you...FanGraphs Points (FGP), FGP/162G, and FGP/33GS. Unfortunately, for pitchers, these calcs require HRA (available only for the 2000 season and later), as well as holds for relievers. Since SimD doesn't record holds, we are sort of SOL on relievers. But it's easy enough to rely on "regular" stats to evaluate the few relievers who get HOF consideration. One other note on relief pitching...The FGP/33GS metric assumes a pitcher is a STARTING pitcher. In the rare case of someone like HOFer Rawley Eastwick who spent significant time as both a starter (338 GS) and a reliever (585 appearances in relief, 191 saves), the FGP/33GS falls apart, but cumulative FGP scoring still sort of works (except for missing holds).

For all but the more recently elected HOF SPs, we can't calculate these FGP-based metrics because we don't have their HRA. I have, however, done two things. First, for those HOFers who straddled pre/post-2000, I measured their HRA/GS for peak and non-peak years, weighting those to project for full peak and non-peak years to come up with a career projected HRA/GS. From this set plus those HOFers for whom full data was available, I searched for "best fit" pitcher profiles close to the HOF average ERA (3.32) and WHIP (1.21) and averaged their HRA/GS to create a coefficient for the pre-2000 HOFers. (This turned out to be 0.615 HRA/GS.) I used this coefficient multiplied by GS to provide the career HRA input for their cumulative FGP calculation. If you follow that, you are as big a nerd as I am.

Anyway...when you roll all of this up, we find that:

• The average HOF positional player has a FGP/162G of 1,015 and a cumulative FGP of 15,198

• The average HOF SP has a FGP/33GS of 1,090 and a cumulative FGP of 17,427

• There are HOFers with substantially higher and lower scores

• In some cases, players with relatively lower FGP/162G or FGP/33GS were inducted where they enjoyed relatively higher cumulative FGP, and, on the other hand, there are some cases in which players had relatively higher FGP/162G or FGP/33GS but relatively lower cumulative FGP. This reflects your accumulators (e.g., Ed Walker, Rick Brennan, Ed Martinez) and your short career or lower health bright burners (e.g., Lowe Morgan, John Nash, Dave Leonard).

We think application of these metrics will be helpful for HOF considerations and help us to resist the temptation to avoid overweight stats that are largely beyond players' control such as W and RBI. Of course, I'm sure there will be plenty of healthy debate around these sorts of points, as well as on how far "below average" a player can be on these metrics to be considered Hall worthy.

Let the debate begin!

[Jonathan8]

Outstanding idea. I appreciate the additional time and effort necessary.

[acnunnally]

Thanks, Jonathan. And thanks again to GM33 for the thought partnership on this.

What interesting about all of this is that roughly, an average CCL HOF career seems to hover around 1,000 FGP/162G and 15,000 cumulative FGP for positional players and a bit more that that for SP. Given that basically no one plays 162 G, you can think of this as a player playing in 146 games per season at a 1,000 FGP/162G level, which translates to 900 FGP per season of 146 G. In 17 seasons, such a hypothetical player would accumulate 15,300 career FGP. But 17 seasons is quite a lot (e.g., age 23 to age 39), so a more typical HOF career might have the player cranking out 1,000 FGP per season of 146 games played, which corresponds to a 1,111 FGP/162G level, over 15 years to hit 15,000 career FGP.

[GM33]

Brilliant! Great work expanding on the idea acnunnally. I am looking forward to incorporating this and expect to have my votes and any additional thoughts submitted in short time.

[GM33]

To calculate estimated HRA for pitchers who played prior to 2000, does it make sense to use a coefficient based on IP rather than GS? I wonder if that will help to compare relief pitchers vs other relief pitchers and also level the playing field a bit for pitchers who both started and relieved in their careers. I tried a coefficient of 0.0785 for HRA/IP in my own calculations, which is about 20 HRA per 255 IP.

[acnunnally]

I did consider this. One problem is that we are never going to be able to get relievers in line with an FGP-based scoring scale on par with SP given that we are missing holds as a stat (4 FGP per hold). There are so few RP worthy of HOF consideration, it is easy enough considering them on a case-by-case basis, so I'm not too worried about that. I will note that for any SP/RP hybrid careers (e.g., Bubba Mason, Rawley Eastwick), we can't apply FGP/33GS scoring because they will start getting punished for HR given up in relief that don't have a matching GS denominator contribution.

The reason I like HR/GS for the coefficient for pre-2000 career performance that rolls up to FGP/33GS is that it keeps a consistent GS basis. If we used a coefficient based on HR/IP, we'd have to make some assumption about how many innings should, on average, correspond to a GS. We could take the IP/GS for each pitcher and multiply the HR/IP coefficient to arrive at HR/GS, but that would be a round trip wash anyway. As long as we're not worrying about relievers (for which there is no feasible solution anyway), I don't think one is better than the other.