What were the least likely go ahead or game tying HRs in the 9th inning or later in postseason history?

I used the Log5 formula from Bill James and got the data from this Stathead search All: 347 Home Runs in 1903-2024 Postseason – during 9th Inning or during Extra Innings (I had to go through and find the ones that were go ahead or game tying HRs-I hope to create a link with all 152 at some point so if I missed any please let me know).

Log5 is used to predict what batting average or HR% a batter will get off of a given pitcher with adjustments being made for the league average. Here is the formula:

where

p_B = the probability that the batter hits a HR (or HR%)

p_P =  the probability that the pitchers allows a HR

p_L =  the league probability that a HR is hit

I used the %'s from the regular season for the year each event happened. For World Series games I used the MLB average for the league average. For all others I used the appropriate league %.

This table has the least likely cases:

Year	Series	Gm#	Batter	Pitcher	Prob.
2005	WS	2	SPodsednik	BLidge	0.00000
1915	WS	5	HHooper	ERixey	0.00194
1988	NLCS	4	MScioscia	DGooden	0.00300
2016	NLWC	1	CGillaspie	JFamilia	0.00320
1976	ALCS	5	CChambliss	MLittell	0.00387
2006	NLCS	2	STaguchi	BWagner	0.00457
2005	NLDS	4	BAusmus	KFarnsworth	0.00464
1973	NLCS	4	PRose	HParker	0.00541
1988	NLCS	4	KGibson	RMcDowell	0.00619
2006	NLCS	7	YMolina	AHeilman	0.00638

Podsednik hit no HRs in 2005. So his chance is 0. The way the formula is set up, if either the pitcher or hitter had 0 HRs, Log5 predicts a 0 HR%. If he had hit 1 HR in 2005 his chance would still be just .00111 and still the lowest (he was the only player or pitcher with 0 HRs in the year in question).

To see how this works, let's look at Scioscia/Gooden (I started thinking about this again since the Dodgers & Mets faced each other again this year in the NLCS). In 1988 Scioscia hit 3 HRs in 440 PAs (I took out his 12 intentional walks and did so for all calculations). That is 0.00682 or a 0.682%  HR%. Gooden allowed just 8 HRs in 1020 PAs for a 0.00781 chance. The league average was 0.01764. Once all those numbers get plugged into the Log5 formula, we get 0.00300 or a 0.30% HR%.

Scioscia hit a 2-run HR off of Gooden in the top of the 9th to tie the score. The Dodgers won game 4 to tie the series at 2-2. Without this HR, the Mets probably would have gone up 3 games to 1. I noticed at this time that both Scioscia & Gooden each had low HR%'s, making this a very unlikely event but I don't think I knew about the Log5 formula at that time. I recall even mentioning to a Dodger fan that it was incredible to see this.

Gooden threw 133 pitches in that game. That could have been a factor. But Gooden had a high that year of 138 pitches in a game (and did not give up a HR in that game). He did not give up a HR in any of his top 5 pitch count games that year and those all had at least 126. He had 22 starts with 100+ pitches but gave up a total of only 5 HRs in those games.

One more thing about that game: The HR Kirk Gibson hit of Roger McDowell was in the top of the 12 inning and put the Dodgers ahead 5-4. It is the 9th least likely HR here. McDowell gave up just 1 HR in 1988 in 371 PAs.

If anyone is wondering about any particular HR, like Mazeroski or Carter, just ask me in the comments or email me.

One factor here is who has the platoon advantage. So far I have not taken that into account. But if I did, there would be more cases of a pitcher or hitter who had 0 HRs during the regular season, automatically giving a 0 predicted chance. That is because there were a number of guys with just 1 HR during the regular season.

Chambliss hit his HR off of Mark Littell. Littell gave up only 1 HR in the regular season in 1976 in 428 PAs. But he gave up none to lefties in 164 PAs. So if I took platoon splits into account, the probability would be 0. McDowell also did not give up a HR to a lefty in 1988, so Gibson's HR off of him would get a 0 chance as well.

Cortes only gave up 2 HRs in 153 PAs to lefties in 2024. So the Freeman HR chance (which is 0.03942 right now) would probably go down. (Update: It would be 0.01759 if we only used how they each did against lefties and the LHB vs. LHP for all of MLB in 2024 to find the league average).

But the platoon splits are definitely worth looking into. Just eyeballing the complete list of 152 cases my guess is that the vast majority of these HRs will be with the batter having the advantage. So some o the 10 listed here might have been a little more likely than at first glance. But probably some more of the guys who gave up just 1 HR or hit just 1 HR would see their chance become 0.  

 

Update Dec. 1, 2024: Batters had the platoon advantage in 77 of the 152 cases. So the difference was not that great.

There may be other relevant factors like park effects, weather, day/night, how many pitches the guy has thrown, how many times the batter has faced the pitcher. Also, instead of using just that year's data, it might be better to take a 3 year average for these %'s or use Marcel the Monkey Forecasting System which takes into account a player's 3 most recent seasons, among other factors.

Update Dec. 3, 2024:

What if we used the 3 year average for the Scott Podsednik/Brad Lidge in Game 1 of the 2005 World Series? For that I used total PAs and HRs over the years 2004-2006.

Scott Podsednik would still have a very low HR chance, 0.00802 (up from 0). Lidge would be slightly higher at 0.02300 (up from 0.01718). The league average would be about the same, 0.02837 (originally 0.02711).

What does that get us? Another low probability, 0.00648. That would be the 11th lowest, if no other probabilities were recalculated.

My guess is that some of the others in the top 10 would rise if they were also re-figured using the 3 year average and that they could go higher than this adjusted chance for Scott Podsednik/Brad Lidge.

The reason I think so is because an unusually low HR% for the batters and pitchers involved in the 10 lowest cases is what put them there in the first place. There is a good chance that their surrounding years were higher.

Second update Dec. 3, 2024:

There were only 5 cases of a LHB vs. a LHP. The one with the lowest predicted HR% was Hooper/Rixey. Baseball Reference does not have complete platoon splits for that year but Rixey allowed 2 HRs in 1915 and his splits page does show that he gave up 2 to RHBs. So that means none to LHBs and, therefore, Log5 would predict a 0% HR%. Baseball Reference has Hooper with 2 HRs that year. His splits page has 1 vs. RHPs and 0 vs. LFPs.

The next lowest predicted % in the case of a LHB vs. a LHP was the Freeman/Cortes HR. We have already seen how that one was altered. The altered chance would rank as 45th least likely.

Update Dec. 5, 2024:

I calculated the probability for the Scioscia/Gooden HR using the 3 year data (1987-1989). Scioscia's chance rises from  0.00682 to 0.01346. Gooden's rises from 0.00781 to 0.01244. The NL HR% rose from 0.01764 to 0.02038. The Log5 prediction rose from 0.00300 to 0.00820. That would still be the 13th least likely.

I also looked at the HR that Log5 said was the most likely and re-calculated it using 3 year data. That was the Yordan Alvarez/Robbie Ray HR from Game 1 of an ALDS in 2022.

Here are their %'s based just on 2022 along with the AL% and what Log5 predicted

Alvarez  0.06703
Ray  0.04129
AL  0.02822
Log5  0.096285

Now the same thing using 3 year data

Alvarez  0.061662
Ray  0.04129
AL 0.03130
Log5 0.08088

Although it falls to 0.08088, it would still be the 4th most likely.

 

Update Dec. 6, 2024:

I looked at the platoon splits for the Alvarez/Ray HR (which was LHB vs. LHP). In this case the probability predicted by Log5 is 0.04989. It would fall from 152nd (where it was the most likely HR) to 129th. So still more likely than 128 of the 152 cases. It would be in the upper quartile.  I only used one year data for this (from 2022).

Second update Dec. 6, 2024:

I used platoon stats for the Scioscia/Gooden HR. Based on the data for the NL in 1988, this probability would be 0.00423. It would still be the 5th least likely HR.