Substitutions
 

I posted a thread about a slight flaw I noticed regarding substitutions on the General Discussions board. Now I'm pretty sure the whole logic process is flawed.

In the example I highlighted, there were 11 substitutions in the span of 1 inning, leaving only 4 pitchers able to enter the game in case of an injury. On top of that, they had a non catcher playing that catcher because they had pinch hit their backup catcher and subsequently replaced him in the field.

I had another game which did this, with one fewer substitution, but the logic was much worse. If other people are experiencing this I can highlight the exact situation, but by the 10th inning, this team had an outfielder at 3B, and outfielder at C, and the backup catcher playing 1B. They did have one emergency sub available, however. The thing I can't understand is why they would leave an outfielder at the C position when the backup C was still in the game!! I'm using strategy settings of 1901, and it seems there are way too many subs, especially favoring the L/R split. Another game I was playing had a team pinch hit for their best hitter, and one of the best in the league, just because he was left handed. The PH coming in was rated much less, and ended up striking out IIRC. It would be like pinch hitting Nick Punto in for Joe Mauer just because the pitcher is left handed. No sober manager would make that move.

Don't get me wrong, substitutions are a good thing. I just think that with the settings at "very rarely/rarely" for substitutions, I should NEVER see double digit subs in the span of one inning. I know programs can't "think", or "plan ahead", but it seems like the AI inserts subs simply because they can without even looking at the obvious best move. Yes some things are subjective, and there are multiple good choices to make, but when theres an obvious best move such as "play your back up catcher at the C position instead of your backup CF", anything else is obviously a poor move. I recall seeing double switches in 2007/8 (something unheard of in 9 so far as I can remember), and although the volume of subs was much less, way more of them made perfect/really good sense, and the others you could probably make an argument for.

Thanks for reading!

I'm doing some data logging right now to see how often the AI will sub in. I'll post updates the farther along I get. The league is based on a 1901 league, no injuries, all substitutions are set to VERY RARELY, including pitchers which I'm not counting.

Here's the stats so far, after 217 games

• 8.5 Innings
  • 80 games, average of 5.038 subs, standard deviation of 2.91
• 9 innings
  • 112 games, avg 7.018 subs, stdev of 3.904
• 10 innings
  • 10 games, 13.6 average
• x>11 innings
  • 15 games, 16.4 average

What does this mean? There's an overall average of 7.24 subs/game regardless of the number of innings. Based on this small sample set the biggest problem is jumping into extra innings. As long as the game stays in regulation, the number of subs seems okay for a normal setting. Still, IMO "very rarely" should mean just that, and 3-4 subs per team per game doesn't seem very rare to me.

I probably read too fast, but two things initially I'd like to ask.

Your totals are for both teams, right? So a 7.24 subs/game means ~3 subs per team average?

How are you removing relief pitchers from your calculations?

Those are the totals for the teams combined. When you read on the box score, above where the detailed BATTING stats are, it lists the subs as:

a. - J. Doe
b. - J. Doe
c. - J. Doe
d.- J. Doe

etc etc


So I "count" the letters for each team and type those quick into my excel document. Actually adding in the relief pitchers would add a ton of time to do this since I'd have to scroll down the page to count those.

A lot of clicking and typing, I know, but I'm okay with that!

Okay, through 400 total games:

• 8.5 Innings
  • 5.201 Average
  • 2.753 Stdev
  • 149 Total
• 9 Innings
  • 6.722 Average
  • 3.697 StDev
  • 212 Total
• 10 Innings
  • 13.647 Average
  • 17 Total Games
• X > 11 Innings
  • 15.091 Average
  • 22 Total Games
  • Average of 12.1 Innings/game

Overall Average of 6.910 Subs/ game at this point. Using some quick rough estimation I would expect the Overall Average for regulation games to approach 6.1 over a very large sample set. This looks very good, and better than I expected. Still maybe a little high for "VERY RARELY" substitute, but its very acceptable IMO. I'm going to focus on extra inning games for a little while.

On a side note, I could get even better results if I were to use the DH rule, that would totally eliminate PH and PR for pitchers. However, there isn't always a PH/PR for pitchers, I've noticed, so this should be okay for now. Also, now that I think of it, one of the sub. settings is PH for pitchers, so thats okay too.

Okay, these last games skewed my 9 inning game results because I used the next 100 biggest files in the box scores folder, so ignore the 9 inning games.

• 8.5 Inning
  • 5.329 Average
  • 2.893 STDEV
  • 152 games
• 9 Inning
  • 7.137 Average
  • 6.198 STDEV
  • 227 Games
• 10 Inning
  • 14.818 Average
  • 2.8 STDEV
  • 33 games
• X>11 Inning
  • 17.5 Average
  • 4.778 STDEV
  • 88 total games
  • 12.5 Inning/ game average

Based on these results, I see a 8.918 sub/game average, but again these are skewed since normally you won't see this high a proportion of extra inning games.

The number that scares me here is 14 average in 10 inning games. It has a low STDEV, but only half are from a random set. If thats accurate, it equates to approximately an extra 4 substitutions per team as soon as extra innings take effect.

Interestingly, there was one incredibly weird game that went 14 innings with 9 total subs. However, all 9 of the subs were on the visiting team. The home team made 2 pitching changes in 14 innings, and those were the only moves they made. Hrmm.....

Last update for tonight, I can probably finish all my box scores either tomorrow or Wednesday.

• 8.5 Innings
  • 5.391 Average
  • 3.09 STDEV
  • 284 Games
• 9 Innings
  • 7.398 Average
  • 3.86 STDEV
  • 405 Games
• 10 Innings
  • 13.782 Average
  • 2.95 STDEV
  • 55 Games
• X>11
  • 17.261 Average
  • 4.82 STDEV
  • 92 Games
  • 12.43 Innings/game Average

Overal 8.221 subs/game average. As soon as the game goes into extra innings it becomes 15.959 subs/game average.

As stated previously, I skewed my results for 9 inning games until I can get more of the "low sub" games into the data, so I'm very pleased with the results in regulation length games. The problem as I see it is that the AI gets a little trigger happy as the game goes on, and I'm projecting approximately double the amount of subs in a game that goes from just the 9th to the 10th inning. I see this as a tad unrealistic, or at the very least too many for the setting of "VERY RARELY". If this is actually realistic, then "VERY RARELY" should be exactly what it says, conversely "VERY OFTEN" (or whatever the converse setting is) should be even higher than usual.

Overall this has been a reassuring experiment so far, I'm seeing the subs in regulation games to be a very acceptable number. At the same time, its not just the number of substitutions that bother me, rather the fact that by making all the substitutions the AI basically self destructs itself in the late game, where mistakes are even more costly than in the beginning of the game. Having an outfielder play catcher/3B, or being stuck with very poor hitters because you earlier subbed your last players in for a L/R match up can cripple a team and lose a game. I can see possibly making these moves if the risk/reward is great enough, but often the subs come with 2 outs and nobody on, and then they're immediately replaced the next half inning in the field, sucking up two reserve players for no apparent reason other than "simply because you can".

EDIT I think I want to add that these teams probably only have 6 or 7 pitchers max. 1901 settings give a 4 man rotation, and they by default take advantage (or would I be taking advantage of them only having 2-3 relief pitchers?). This doesn't change my point, however. If they're so aggressive they're capable of burning through 9-11 players in 1-2 innings of play (leaving NO subs except the bullpen), whats to stop them from doing the same thing in the modern day carrying only 5-6 subs?? It's possible its just an AI flaw when the roster is set like this, but I suppose it's something to check out.

So do we need to log anything here?

As the OP hasn't specifically flagged anything up as an issue and there has been no response in the last 7 days, this thread will be closed.