Best stealers underperforming historical success rates

[Brad K]

In my current game the most prolific base stealers are far under performing their historical success rates. I have run this game 1951 through 1980.

My settings are full history not imported, 3 year recalc with no weighting, development on. I have not altered league stats totals.

Top players for SB in game shown below.

Here is their game vs historical success rate.

Name - Game - Historical

Joe Morgan 679 810
Bert Campaneris 689 765
Lou Brock 637 753
Luis Aparicio 655 788
Cesar Cedeno 699 754
Maury Wills 623 734
Bill North 622 709
Willie Davis 632 752
Willie Mays 617 766
Bobby Bonds 632 732

[xgeoff]

I believe this has been going on for a while. In OOTP20 I had a team with a lot of players with high base stealing ratings (80+). Their stealing %'s were terrible, and one of them (Martin Dihigo) got caught stealing over 50% of the time.

I wondered if it was just bad luck in terms of running into catchers with good arms and/or pitchers with good 'hold runner' ratings. So I went and reviewed a number of games (not exhaustive, but a sample of about 5 games) where he was caught stealing and in every occasion the catcher's arm and pitcher's rating was marginal at best.

Very frustrating. Not as thorough analysis like yours, but some anecdotal evidence to go along with what you are saying.

[Brad K]

Originally Posted by xgeoff

I wondered if it was just bad luck in terms of running into catchers with good arms and/or pitchers with good 'hold runner' ratings. So I went and reviewed a number of games (not exhaustive, but a sample of about 5 games) where he was caught stealing and in every occasion the catcher's arm and pitcher's rating was marginal at best.

As you say, a super small sample, but my assumption was success rate was low because players didn't decrease the number of attempts when facing good pickoff pitchers and good arm catchers.

[Brad K]

The players previously listed are attempting more steals per single in the game than they did historically. (Yes, singles are a rough estimate of steal opportunities but its the one used in OOTP for historical accuracy checks.)

Steal Attempts Per Single

Name - Game - Historical

Joe Morgan .628 .499
Bert Campaneris .554 .479
Lou Brock .590 .554
Luis Aparicio .381 .305
Cesar Cedeno .655 .524
Maury Wills .590 .426
Bill North .650 .659
Willie Davis .590 .426
Willie Mays .311 .255
Bobby Bonds .622 .531

We have an outlier, Bill North, but otherwise all of the players attempts many more steal attempts in the Game than in real life. I'm wondering if perhaps the vicinity if .650 is a maximum that can be achieved in either the game of RL. That would be a possible explanation for North's performance. But still that would not explain his lower success rate. Interestingly, he has the second lowest success rate in the game and the lowest RL. Don't know how to interpret that.

If North can be explained then the rest of the data supports a theory.

1. In the game players are more likely to run against lower quality opposition (catchers with weaker arm rating, pitchers with lower hold rating).

2. If, in the game, if a player is asked to run more, he's used up the low quality opposition and so is exposed to higher quality opposition and his success rate drops.

3. In the game, AI managers are asking good stealers to attempt too many steals (all the players in the game played exclusively for AI controlled teams).

-------------

Comments requested. I haven't posted this to the bugs forum because I'd like some input concerning the thoroughness of the documentation and whether conclusions are logical.

[ExeR]

Originally Posted by Brad K

The players previously listed are attempting more steals per single in the game than they did historically. (Yes, singles are a rough estimate of steal opportunities but its the one used in OOTP for historical accuracy checks.)
1. In the game players are more likely to run against lower quality opposition (catchers with weaker arm rating, pitchers with lower hold rating).
2. If, in the game, if a player is asked to run more, he's used up the low quality opposition and so is exposed to higher quality opposition and his success rate drops.
3. In the game, AI managers are asking good stealers to attempt too many steals (all the players in the game played exclusively for AI controlled teams).

Better formula for ratio would be: 1b+2b+walks(all kinds, incl. hit by pitch).
As for stealing in general, I've always paid attention and tested it. In 21 there's a change towards unicorns with 200+ speed. Here's the test I've ran:



Results help to understand some of misconception that make you guess between options.
Speed is primary attribute, Stealing is modifier.
Player with max Speed will attempt more than player with good Speed, even if Stealing is maxed for both. If you're accustomed to different "stats correlations" types, this is one of the examples:
a) vs poor Hold pitcher & poor Arm Catcher speedster will be effective even if he has stealing 10 of 250.
b) vs good ones, he will get caught max, but Speed will still give him good volume of SB
c) W/o speed, even 250 Stealing player won't get anything. High Floor on Speed attribute, but low Speed will automatically make him attempt less, so he will only go for it in some very good spots (like a)
d) Max Speed, max steal guy is now unstoppable: 60% attempt ratio & 90% success. So if you get elite OBP guy with max SPE&STE, he will get you insane amount of extra runs. So powerless CF prospects can be "steals" in late 1st round, but max speed is very very rare.
Now all that's left is your tactical instructions. As you can see from Stealing bases 1 (yellow background, row 11&12), you can shut down even most eager players. So using a, b, c, d you can exclude all the options you listed.
As was my experience, balancing of historical players was done through both SPE&STE. Now that SPE is much more prolific, historical players with skew towards STE are struggling. As you can see from row 19&20, you can get even very good guys (175 of 250) into negative BsR by running recklessly. Meanwhile 250 of 250 guy running recklessly will get you 1/3 of elite bat production from both stealing & baserunning. 9 WAR with 0 2b, 0 3b, 0 HR. So the middle ground is lacking balance, that's why you see difference in outcomes.
I'm not stating there are no other factors affecting this. Maybe HOLD or Catcher Arm is OP. Maybe stealing max amount is too much and there's "diminishing returns" principle in engine. But I'm pretty sure if you transfer STE rating into SPE rating, things will improve at same AI settings. Right now there are pretty much 3 tactical settings for stealing(see last rows):
10 (max stealing bases) for elite guys to max their positive actions
5 (average stealing bases) for above average or close to elite guys to carefully pick their spots and prioritize success above volume
1 (min stealing bases) for everybody else. They will still get some, but they don't accumulate any volume, so minimizing outs is pretty much the best thing to do. If there's close game & you're down 1-2 runs, manager will probably still push them to steal if it's 3 instead of 1.
As you can see from BP column (lime background), sensible Baserunning/Stealing Bases values to match player abilities allows all kind of players to produce positive results from Baserunning. But stealing only starts being positive at 100 of 250 SPE & elite (150+) STE. And even then it's minuscule @ "best possible situations" frequency. So if STE is factor for coach to increase frequency and for "old" SB output formula, it can mess up 150 SPE, 200 STE results. Which most of them are?

[Brad K]

"Better formula for ratio would be: 1b+2b+walks(all kinds, incl. hit by pitch)"

Maybe. Maybe not. I don't know. But none of the events always result in an opportunity to steal, which is what we're trying to determine, so all are inaccurate. Anyway, I choose SB attempts / 1B because that's what the game uses in its simulation accuracy analysis. It seemed important to match the game's method in the event of a future comparison with overall results.

You have provided very interesting information and have put a lot of work into this. The conclusion speed is the attribute and stealing ability the modifier is enlightening.

But, as you say, its a bit hard to understand. Does it provide any clues about the success rate in my data under performing historical?

[polydamas]

(quoting myself from another thread)

I'm not sure if this will be of any help or not tbh

Originally Posted by polydamas

You can fix this easily in the settings. I've attached what my settings are.

Notice how Barry Larkin is actually a better basestealer than he was IRL. His irl sb% was about 83% while in my game it was about 86%

Henry's is 87.5%. Wise's 87%. I have a lot more examples from this league. These were just the first 3 I looked at.


You should be able to fix it by bumping up the stolen base success % (mine is 1.050)


Edit: I found a guy with almost 90% success. I attached it (Reggie Smith).

You will have to go to

https://forums.ootpdevelopments.com/...49#post4673749

if you want to see all the pictures (it is the first response)

Edit: Here is my most proficient base-stealer (as far as I am aware). He turned out significantly better than IRL as he was 137-86. I would test 1.025 first though, because 1.050 seems a little high.

Ed.2: had the wrong link

[David Watts]

Is there a way using OOTP to tell how many of a players caught stealing come while attempting to steal 3rd base?

[Questdog]

Originally Posted by David Watts

Is there a way using OOTP to tell how many of a players caught stealing come while attempting to steal 3rd base?

No.

[ExeR]

Originally Posted by Brad K

But, as you say, its a bit hard to understand. Does it provide any clues about the success rate in my data under performing historical?

Opportunities to steal come only from On-Base (and not 3b, because you can't steal it). So I'm pretty sure it's correct to take all On-Bases at 1st and 2nd. Maybe there are modifiers: 3rd is not as easy to steal from 2nd as 2nd is from 1st. But I don't think there's a way to find it out from available stats.
As for underperforming, I suggested two possibilities:
1) As you can see from my data, min "Stealing Bases" tactic from manager stops even GOAT. So AI tactics will have a significant impact. You can also see results different combinations of tactics & player ability. So if AI algorithm isn't perfect/optimized for 21 version changes, it can lead to suboptimal or unexpected results. Best example of it is 175 of 250 guy being negative contributor in max "Stealing Bases"&"Base-running" tactic.
2) Continuation of first. If balance was based on Stealing attribute having more impact it can mess things up. Historical players may have more stealing & less speed (in regards to 21 norms). And AI managers can base their "Stealing Bases" tactic on Stealing attribute.
21 version changes were probably made for modern reasons. And historical stuff can always be fixed by stolen base success %.

[Brad K]

Regarding how much on base data to include to determine opportunities, c'mon. Being on first after a single is more likely than being on second after a double, an attempt of third is less likely than an attempt at second, and a single is more likely to result in there not being a runner on second than a walk. So adding the other stuff may result in less accuracy. Besides, THE GAME uses SB/single in its data and so a standard is established.

[Brad K]

Originally Posted by ExeR

21 version changes were probably made for modern reasons. And historical stuff can always be fixed by stolen base success %.

Not if its only the top players who are affected. That's yet to be determined.

[Brad K]

OK, I did a quick check 66 - 82 (66 is the first year SB/single reached as high as .08, and it remained there or higher (up to .13) except for one year. SB percentage variation from historical is -1.1% the way its presented appears to mean - 1.1% of the success percentage.

This is not weighted by the number of attempts. I just quickly ran down the simulation historical accuracy chart.

[Argonaut]

ExeR, have you noticed any effect from the Baserunning rating in your testing? Having a quick look at your Excel file, it seems like you held RUN pretty constant without isolating its effect. I'm wondering if RUN has an effect on stealing results.

[Brad K]

"21 version changes were probably made for modern reasons. And historical stuff can always be fixed by stolen base success %."

"You should be able to fix it by bumping up the stolen base success % (mine is 1.050)"

No, because the overall numbers are good. Doing this to fix the good stealers would make lesser players too good at it.

See below which requires some explanation.

Estimated SB and Estimated CS are because I don't have the raw data from the game, just the success percentage. So I estimated the number of SB and CS in the game by applying the percentage to the number of real life attempts.

The percentages listed on the Totals line are based on real and estimated attempts.

The percentages listed on the Average line is the average of the percentages and is thus not weighted for attempts. But it is interesting that within each method the results are very close.

I have concluded that there is a problem with prolific stealers on AI teams attempting too many steals and encountering diminishing returns from their efforts.

[ExeR]

Originally Posted by Argonaut

ExeR, have you noticed any effect from the Baserunning rating in your testing? Having a quick look at your Excel file, it seems like you held RUN pretty constant without isolating its effect. I'm wondering if RUN has an effect on stealing results.

As you can deduct it wasn't just stealing test. I just wanted to understand scope of "production" from "goat" speedster player, and how related tactical instructions impacted everything. F.e. "stupid" AI can ruin productive speedster with minimal instructions(and vice versa), and I would have to fix that manually.
As for results, on one hand it shouldn't impact. Baserunning is measured by itself, if I'm not mistaken. But it happens after hits. Stealing is before hit. So when we get number of On-Base @ 1st&2nd we have something to compare to (SB+CS).
On other hand, what you are talking about has two components:
1) Tactical strategies like Run&Hit can affect Stealing/Baserunning numbers. If hitter fails to hit, it can be "Steal",but if he hits it's "Baserun". Maybe not, but it should be so.
2) Amount of opportunities to steal per On-Base varies. Previous point about Run&Hit can be example here. But following batters & opposition pitcher can also play a role.
Problem is that these precise tests are tough to execute. You don't control all the moving parts. And the margins are not worth it. In ideal world it would be (SB+CS)/(On-Base 1st+On-Base 2nd)*(amount of pitches per On-Base). Every pitch is opportunity for stealer to deliver.
But if we use common sense:
a) Tests showed how good/unstoppable speedster is. You should maximize his impact through maxing "Stealing bases" & "Baserunning" and...
b) Players in lineup behind him should be OBP specialists, not HR. Then we can maximize amount of pitches & amount of hits. Pitch to Steal, hits to Baserun.
c) Opposition pitcher is last part, but it's just adjusting vs his Hold & Catcher's Arm. If he is pitch around guy, or their team uses that to the max, it's good for speedster. If he got strong Hold & Arm we should Steal less. And vice versa.
So ideal test would be to make all pitchers same and have all AI teams same in "pitch around" tactic. But it will never be the case in real leagues, so it's just engine testing.
That's why I just enjoyed knowing basics about elite speedsters w/o digging deeper or getting pure results.

[fredbeene]

Originally Posted by Brad K

The players previously listed are attempting more steals per single in the game than they did historically. (Yes, singles are a rough estimate of steal opportunities but its the one used in OOTP for historical accuracy checks.)

Steal Attempts Per Single

Name - Game - Historical

Joe Morgan .628 .499
Bert Campaneris .554 .479
Lou Brock .590 .554
Luis Aparicio .381 .305
Cesar Cedeno .655 .524
Maury Wills .590 .426
Bill North .650 .659
Willie Davis .590 .426
Willie Mays .311 .255
Bobby Bonds .622 .531

We have an outlier, Bill North, but otherwise all of the players attempts many more steal attempts in the Game than in real life. I'm wondering if perhaps the vicinity if .650 is a maximum that can be achieved in either the game of RL. That would be a possible explanation for North's performance. But still that would not explain his lower success rate. Interestingly, he has the second lowest success rate in the game and the lowest RL. Don't know how to interpret that.

If North can be explained then the rest of the data supports a theory.

1. In the game players are more likely to run against lower quality opposition (catchers with weaker arm rating, pitchers with lower hold rating).

2. If, in the game, if a player is asked to run more, he's used up the low quality opposition and so is exposed to higher quality opposition and his success rate drops.

3. In the game, AI managers are asking good stealers to attempt too many steals (all the players in the game played exclusively for AI controlled teams).

-------------

Comments requested. I haven't posted this to the bugs forum because I'd like some input concerning the thoroughness of the documentation and whether conclusions are logical.

manually go through and up their ratings.
manually go through and lower catcher ratings

[polydamas]

Originally Posted by fredbeene

manually go through and up their ratings.
manually go through and lower catcher ratings

Why not change one setting instead of re-rating every player in the league?

[fredbeene]

Originally Posted by Brad K

In my current game the most prolific base stealers are far under performing their historical success rates. I have run this game 1951 through 1980.

My settings are full history not imported, 3 year recalc with no weighting, development on. I have not altered league stats totals.

Top players for SB in game shown below.

Here is their game vs historical success rate.

Name - Game - Historical

Joe Morgan 679 810
Bert Campaneris 689 765
Lou Brock 637 753
Luis Aparicio 655 788
Cesar Cedeno 699 754
Maury Wills 623 734
Bill North 622 709
Willie Davis 632 752
Willie Mays 617 766
Bobby Bonds 632 732

My experience has been catchers with little experience get imported with inflated catching ability. We are talking 95 and over ratings
This has had an effect on my stealing rates.

I am imported players with 3 year fielding

[fredbeene]

Originally Posted by polydamas

Why not change one setting instead of re-rating every player in the league?

Sure that would be great.
What is the setting.

Not sure setting the sucess rate is equivalent to the actual ratings of the players.
Rating would barely matter???