NERP and Catcher Defense

It's difficult to assign a specific SB percentage that's works universally. Stolen bases come with some risk (CST) but it's a good strategy under certain circumstances:

It's very valuable to steal second base if you are tied in the 9th inning with a singles hitter like Willie Keeler batting next. Maybe 50% is OK.

Even at 85% it's probably not worth the risk when you are down by two runs and Barry Bonds (walk or HR) is up next.

Poor arm catchers can sometimes trick HAL into stealing when the percentages are good even if the situation doesn't warrant it.

In Marc P's earlier post he said that holding a runner changes the middle infielder from a SS-1 to a SS-4.
I thought the effect of holding was an increase from SS-1 to SS-2. Can anyone confirm if Marc is correct?


"And the impact of not holding a runner is huge: holding a runner has an impact on the defensive charts, a ss-1 (or a 2b-1) becomes basically a ss-4 (or a 2nd-4) but with still double-plays capabilities. Without holding the runner, the defensive player gets restored to his ss-1. "

The infielder's range goes from 1 to 2 but 4 groundouts turn into singles as well, so I think he's accounting for those and saying a 1 responsible for a hold is effectively the same as a 4 with no hold, which is correct.

In Marc P's earlier post he said that holding a runner changes the middle infielder from a SS-1 to a SS-4.
I thought the effect of holding was an increase from SS-1 to SS-2. Can anyone confirm if Marc is correct?


"And the impact of not holding a runner is huge: holding a runner has an impact on the defensive charts, a ss-1 (or a 2b-1) becomes basically a ss-4 (or a 2nd-4) but with still double-plays capabilities. Without holding the runner, the defensive player gets restored to his ss-1. "

The infielder's range goes from 1 to 2 but 4 groundouts turn into singles as well, so I think he's accounting for those and saying a 1 responsible for a hold is effectively the same as a 4 with no hold, which is correct.

hmmmm I'm not sure about that. I'm looking at the super advanced fielding chart , and I'm just not seeing it.

I too am looking at the SADV infield chart and there are 4 groundouts marked with a # for each of the 5 range ratings. Those are the groundouts that turn into singles if the infielder is in or holding a runner. Add those four to the two natural singles for a 2 rating and you get 6 singles, which is what a 4 gives up when playing back.

For some reason I thought Pelletier on catcher defense had already been linked but if not: https://forum-365.strat-o-matic.com/com ... b1c8243ace

If you use his arm formula you don't have to worry about adjusting for the extra outs as it's all included, as is the effect on holds.

I realize I should've added, if you did want to do it manually rather than cribbing from Marc, basically the formula is 4.85 r/g, divided by 27 outs = 0.18 runs per out above average in a neutral environment. You're comparing the catchers' absolute CS totals against each other in terms of the 0.18 runs prevented by the additional out. And it goes for all types of outs. This is value additional to the inning state transition, it comes from making the game shorter or longer, whereas linear weights measure events in terms of within-inning expectations.

The effect on holds, OTOH, I haven't the slightest idea how to value, which is why I'm happy to crib from Marc.

Trying to add a few more data points to the discussion:

I used the Windows game to perform simulations of many seasons under specific circumstances. The players were from the 2021 set. Teams were based on a specific 365 league that I was in last year. In Windows, you can edit a player’s ratings, as well as set an entire league’s manager strategy. I used J.T. Realmuto to test the impact of arm strength and used manager strategy to test SB strategy. All other variables (teams, players, strategies, etc.) were held constant. I tested a +5 arm and -5 arm as well as SB strategy of normal and SB strategy of Extra Conservative. Statistics recorded for each season were Realmuto’s SB allowed, CS, and the team’s runs allowed. 20 seasons were simulated for each combination of arm/strategy. The following summarizes my results:

Arm	SB Strategy	Avg. SB	Avg. CS	Avg. SB%	Avg. Runs Allow.
-5	Normal	33.4	14.1	70.4%	885
+5	Normal	120.5	33.8	78.1%	878
-5	Ex. Conserv.	22.0	7.3	75.0%	876
+5	Ex. Conserv.	84.9	17.9	82.6%	854

Differences between SB% by Arm and by strategy are statistically significant
Most run differences are not statistically significant, however the difference between normal sb strategy and Ex. Cons. Sb strategy at a +5 arm is statistically significant

Significance measured at p=5%

Please note that saying a difference is statistically significant just means that it is likely that the means of those distributions are different. It does not mean that the difference is exactly as measured in these simulations.

Great stuff. Obviously far more stealing going on in ATG than (presumably) 20xx but that's to be expected.

Trying to add a few more data points to the discussion:

Arm	SB Strategy	Avg. SB	Avg. CS	Avg. SB%	Avg. Runs Allow.
-5	Normal	33.4	14.1	70.4%	885
+5	Normal	120.5	33.8	78.1%	878
-5	Ex. Conserv.	22.0	7.3	75.0%	876
+5	Ex. Conserv.	84.9	17.9	82.6%	854

Differences between SB% by Arm and by strategy are statistically significant
Most run differences are not statistically significant, however the difference between normal sb strategy and Ex. Cons. Sb strategy at a +5 arm is statistically significant
.

Great analysis- thank you. It's surprising to see the +5 arm led to the fewest average runs scored.
Maybe the Windows game doesn't hold runners using the formula earlier predicted?

Hello John Henry!!

Oh I didn't even notice that result lab pointed out. Seems bad from a game design standpoint...HAL's SB logic is clearly borked if weaker arms are preventing more runs than stronger arms.

Oh I didn't even notice that result lab pointed out. Seems bad from a game design standpoint...HAL's SB logic is clearly borked if weaker arms are preventing more runs than stronger arms.

You can't conclude that from this data. The differences are not statistically significant which means that there is some chance that they are samples from a distribution with the same mean. The difficulty in measuring the impact of a small factor (like catcher's arm) on runs is that the standard deviation of runs is around 40. I'm skeptical of the results for runs allowed by arm strength with Extra Conservative SB strategy. Other analysis I've done of run values suggests that success rates this high should produce positive value for SBs. I may run more simulations, but this is a tedious process.