Problem Stealing Bases! Help!!!

[RonCo]

Here are some plots of the data in #16 above. Note that there can be a wide randomness to the data--the results are not tightly focused on the trendline in most cases. This is a problem for many of us, and is the reason you really can't analyze things in the game with a small sample of players. For example, speed is a driver to how often a runner will run, but you can find high-speed runners who don't run often. On the counter side, however, you cant find many low-speed runners who run a lot.

[RonCo]

I plotted some MLB data and OOTP data to show how the model compares to real life. Overall, it's good, but the very best runners do tend to underperform a little...with one small sample-size caveat...

Data: MLB 1960-2006 (from Lahman) for all runners with at least 10 SBA in a season. OOTP data - one season for all runners with at least 10 SBA.

I laid the OOTP data on top of the MLB data, and you can see that the two match pretty well. I circled the upper region to show where no OOTP player went. While this is only one year's OOTP data, I should note that it's relatively rare for me to find an OOTP runner wh runs a ton and sees 80-85% success...though it can and does happen on occasion.

My take on the algorithm is that it's not "bad" as one poster said...but it's not "perfect." And being "not perfect" it can drive some folks batty, especially those who take the most joy from watching premium stealers peg the scale.

[injury log]

I recently ran a season of a fictional league, with the game's default modern-day MLB style settings (SB frequency set to 'Normal'). I described the league here (there is some discussion there as well about catchers with very poor arms, who apparently can still have a decent CS%):

http://www.ootpdevelopments.com/boar...d.php?t=151243

In effect, all of my players had maximum ratings in Speed and Stealing, on the 1-5 scale, and I ran a lot. I managed the majority of my games, and my SB success % was very high (see Washington in screenshot):

SB: 971
CS: 279
SB%: 77.68%

These results include several botched steals of home, which is, evidently, a low percentage play in-game. While I did Run+Hit at times, most steals were using the 'Steal' button. I never used 'Send Forced'. I didn't pay any attention to pitcher Hold Runners ratings, or to opponent Catcher Arms; I was running in most situations. The sample size here should be large enough that the results are meaningful.

So I haven't seen the problems others are seeing with fast runners and SB%. I wonder if certain league settings are affecting people's results?

[AirmenSmith]

I use to have trouble trying to steal bases and then it happened. I drafted one guy and after 2 years in the minors I promoted him. His first season he was my leadoff hitter and got 122 SB. That is the most I have ever had someone get.

[RonCo]

Originally Posted by injury log

I recently ran a season of a fictional league, with the game's default modern-day MLB style settings (SB frequency set to 'Normal'). I described the league here (there is some discussion there as well about catchers with very poor arms, who apparently can still have a decent CS%):

http://www.ootpdevelopments.com/boar...d.php?t=151243

In effect, all of my players had maximum ratings in Speed and Stealing, on the 1-5 scale, and I ran a lot. I managed the majority of my games, and my SB success % was very high (see Washington in screenshot):

SB: 971
CS: 279
SB%: 77.68%

These results include several botched steals of home, which is, evidently, a low percentage play in-game. While I did Run+Hit at times, most steals were using the 'Steal' button. I never used 'Send Forced'. I didn't pay any attention to pitcher Hold Runners ratings, or to opponent Catcher Arms; I was running in most situations. The sample size here should be large enough that the results are meaningful.

So I haven't seen the problems others are seeing with fast runners and SB%. I wonder if certain league settings are affecting people's results?

The data in my charts says that if you artificially set players to have very high steal and speed rating, then they will run a lot (both ratings influence # attempts). In terms of playing by hand, I suggest this means that they are able to get a jump more often when you send them.

It also says that if you artificially set all your "steals" ratings to very high, you'll get higher success rates. A 77% rate for the most premium baserunners is actually a little low with regard to the modern-day MLB data in the post above--which indicates that players who are star runners whould be able to steal at rates in the mid-low 80s on average. Specific cases should be even higher.

Do you have a listing of SBA and success rates by individual?

[injury log]

Originally Posted by RonCo

The data in my charts says that if you artificially set players to have very high steal and speed rating, then they will run a lot (both ratings influence # attempts). In terms of playing by hand, I suggest this means that they are able to get a jump more often when you send them.

It also says that if you artificially set all your "steals" ratings to very high, you'll get higher success rates. A 77% rate for the most premium baserunners is actually a little low with regard to the modern-day MLB data in the post above--which indicates that players who are star runners whould be able to steal at rates in the mid-low 80s on average. Specific cases should be even higher.

Do you have a listing of SBA and success rates by individual?

I do, and I've checked each player's ratings; details are below. The leader had an 86% success rate in 186 attempts, along with a 235 Stealing rating.

First, just to be clear, I didn't edit any player's ratings in the game; instead I signed or traded for existing players with high Speed and Stealing ratings, in a standard fictional modern-day MLB-style universe. Because that was the only criterion I was using to choose my lineups, I was playing guys who should have been in AA or AAA, which would have made the league somewhat unlike real life. Two players, Cox and Watson, were rated 4/5 in speed/stealing, the rest 5/5.

The individual player stats, along with the speed and stealing ratings (from the Editor) for all basestealers with 10 or more attempts are below (I only checked these at the end of the season, though I don't think they change much over a season). Complete player stats are attached as a screenshot (none of the players missing from the screenshot attempted a steal):

format: SB/CS; SB%. Speed; Stealing.

Tim Williams: 160/26; 86.0%. SPD: 200; STE: 235
Pablo Garcia: 131/29; 81.9%. SPD: 187; STE: 244
Kevin Joseph: 96/39; 71.1%. SPD: 163; STE: 186
Sixto Vargas: 106/35; 75.2%. SPD: 189; STE: 229
David Watson: 90/27; 76.9%. SPD: 125; STE: 179
Zachary LaMoore: 78/25; 75.7%. SPD: 168; STE: 187
Paul Wagner: 91/20; 82.0%. SPD: 201; STE: 228
Anthony Holland: 65/16; 80.2%. SPD: 175; STE: 240
Felipe Hernandez; 26/13; 66.7%. SPD: 177; STE: 179
Joseph Ramey: 48/14; 77.4%. SPD: 202; STE: 225
Jack Cox: 24/6; 80%. SPD: 139, STE: 174
Alfredo Medina: 10/10; 50%. SPD: 167; STE: 180
Anthony Taylor: 12/4; 75%. SPD: 165; STE: 241
Enrique Pinto: 11/4; 73.3%. SPD: 198; STE: 201

[RonCo]

Originally Posted by injury log

format: SB/CS; SB%. Speed; Stealing.

Tim Williams: 160/26; 86.0%. SPD: 200; STE: 235
Pablo Garcia: 131/29; 81.9%. SPD: 187; STE: 244
Kevin Joseph: 96/39; 71.1%. SPD: 163; STE: 186
Sixto Vargas: 106/35; 75.2%. SPD: 189; STE: 229
David Watson: 90/27; 76.9%. SPD: 125; STE: 179
Zachary LaMoore: 78/25; 75.7%. SPD: 168; STE: 187
Paul Wagner: 91/20; 82.0%. SPD: 201; STE: 228
Anthony Holland: 65/16; 80.2%. SPD: 175; STE: 240
Felipe Hernandez; 26/13; 66.7%. SPD: 177; STE: 179
Joseph Ramey: 48/14; 77.4%. SPD: 202; STE: 225
Jack Cox: 24/6; 80%. SPD: 139, STE: 174
Alfredo Medina: 10/10; 50%. SPD: 167; STE: 180
Anthony Taylor: 12/4; 75%. SPD: 165; STE: 241
Enrique Pinto: 11/4; 73.3%. SPD: 198; STE: 201

There you go. A sample of players with very good Speed/Steals who register performance of between 50%-86% success. Sample size is always important. Given these ranges as being the "expected" range of performance, the Lou Brock example upstream is not too concerning. Eyeball plotting these on the charts Iposted above would still maybe be a little low, but not bad.

[bruckm]

[quote=RonCo;2273689]I'm not sure what you mean by "only when I get a good jump." Runners in OOTP either get a jump or don't. There really is no jump quality modeled to the best of my knowledge. It's OOTP lingo for "he actually attempted a steal." Maybe I'm wrong.
quote]

I mean when they do attempt a steal = a good jump. It seems to me that they are not stealing earlier in the count because they are not getting a 'good' jump. I guess it's just my way of understanding this aspect of the game when selecting the steal 2nd option.

Interesting that the game may punish some runners for attempting more steals than they did IRL. Sometimes stealing isn't merely a function of speed and ability but it can also be an issue of strategy. '82 Brewers had Paul Molitor who probably could have stolen 80 bases during the early years, but his managers were more interested in playing for the 3-run homerun than taking chances on the bases. Of course at an 80% succes rate, he was more likely to steal a base than Stormin' Gorman Thomas was to hit a homerun.

Finally, I have been stealing with every runner 10+ to test some of this. I have been attempting steals only against catchers whose arm ratings are 3 or more pts less than the runner's steal rating. Some trends seem to be VERY clear. My best stealers seem to attempt a steal on the 1st or 2nd pitch by a pitcher with a hold rating below 10. In this case the pitch is usually a ball...but not called a pitchout. The <18 stealers take more pitches to 'get a jump' against the same pitchers/catchers but they are more successful %-wise. It seems to me that pitcher hold rating only seems to impact how many pitches it takes for a runner to take off. I can't really tell what holds more influence catcher arm or stealing rating...or if it is equal. It seems that my best base stealers have a 50-50 chance when stealing 2nd against a crappy catcher. Again, the others can seem to get thrown out.

[RonCo]

I don't think it's a matter of punishing players for stealing more than they did in real life. Or at least not too much so. I'm purely guessing that it's Markus's way of making sure he doesn't get 100 Rickey Hendersons in a hundred year's time.

Some of your commentary about what makes a stolen base successful is at the root of why this kind of algorithm is really hard to model to its full depth. We can talk about situational elements and surprise and know what we mean, but when you need to write it down and convert it into hard-code binary that makes decisions it gets really tricky. I mean, how do you derate the three-way game theory between the pitcher, catcher, and runner. Or better yet, the influence of the guy with the bat in his hands...or maybe even the quick glove of the second baseman? How do you factor in the fact that the first baseman is a lefty rather than a rightie? Do you? How do you weight the influence of experience vs. the natural tendency to lose speed by the time a player is in his mid-20s? What about the pitch selection process? Hw much influence does a curveball have over a fastball? I mean, really? Exactly how much advantage do you give a runner who runs on a curveball vs. one who runs on a fastball--especially when I don't think the game's fundamental results engine works on a pitch-by-pitch basis? How do you determine surprise? What situations? There are a billion...and the state engine to capture all these things is highly complex.

The running game is perhaps the most difficult element to model of all of baseball. To do it much better would most lilely require a really detailed model, and one that historical simmers would hate...because at the end of the day, all Markus has to work with is the Lahman database. All he can do right now is take the numbers that real life players put up and do his best to extrapolate context--and that is really an almost impossible task. Well...not almost. It's really flat-out impossible to do with any certainty.

In the big picture, Markus's code for the running game seems to model real life at the results level pretty well...but it's off just a little in the details. The challenge for a beta-tester...or anyone who really wants to see how it works...is to find a big enough scope and do a big enough set of tests to determine what is real and what is just random luck.

I'm not tryng to be a Markus apologist here. I think the running game can and should be done better. But I acknowledge that it's a tough, tough area to model, and I tip my cap to Markus for getting the overall environment to be one that matches historical output really well in the overall scheme.

[bruckm]

Originally Posted by RonCo

I don't think it's a matter of punishing players for stealing more than they did in real life. Or at least not too much so. I'm purely guessing that it's Markus's way of making sure he doesn't get 100 Rickey Hendersons in a hundred year's time.

Apparently even Rickey is 'regulated' by this invisible hand. Istarted a league with the '80 A's and he's having the same issues as Molitor, Griffey, and Morgan were having. Interestingly again, my sub-15s are able to steal at a higher %. This is especially true against pitcher with poor hold and catcher with poor arm. Apparently every ball is a pitchout when Rickey runs and not when Dwayne Murphy runs against what should be a gimme for Rickey.

Interesting game.

[RonCo]

If you can put a csv dump of your career batting stats file and the players file on a place where I can get it, I'll do a quick plot of you league's overall SB performance and we'll see what things look like from the big-picture perspective.

[bruckm]

OK....everything seems to work perfectly in sim mode. Catcher who should only be throwing out 20% of base stealers do just that.

In Play-by-play mode, that same catcher (<10 arm rating) throws my runners out in excess of 33%. Again, the 17's+ get thrown out at a 40% rate and the 10-15's at a <20% rate. This is not counting announced pitchouts.

In play-by-play mode base stealing is either randomly generated, meaning the stealer's and catcher's skills have no relationship to the actual success of each attempt, or the engine used is badly flawed and must have a glitch of some kind.

[spider_monkey_b]

Just to add a little to the discussion here, though not as in depth as some of the replies. I've been playing my games out (though in one-pitch mode to save time), and I am able to use pitch outs to my advantage. One opposing team has 4 players with 10 speed and steal (10 scale), and if I pitch out I throw them out fairly frequently, though not all the time.

The only aspect of stealing that really gets under my skin is when I tell a runner to steal 2nd on 4 consecutive pitches and he never gets a good enough jump. That seems to happen to me all the time.

[statfreak]

this has come up a lot in the past several months. On a league wide basis, ie just looking at league totals, stolen bases and SB% look okay. It is when we examine individual players, especially those that should be the elite stealers that we see the game does not give what most of us consider expected/realistic results.

To "fix" the stolen base issue to my satisfaction, I resulted to adjusting the OOTP engine ini file (C:\Documents and Settings\user\Application Data\Sports Interactive\OOTP Baseball 2007\data\config\engine). I know others, including online leagues, have done the same. I bumped my GAME_STEALING_SUCCESS up to 120 and have found the results to be more realistic and closer to my expectations (when looking at individual players).

I would suggest tinkering with that. I know you shouldn't have to and in every other version of the game you didn't have to because stolen bases were about the only thing that felt right without having to adjust this or that, but for whatever reason a change was made in 2007 that, as far as I am concerned, ruined one aspect of the game that had been working properly.

[bruckm]

I know I'm a one-trick pony here, but this is ridiculous. A 5 arm catcher is 7 for 17 in throwing out base stealers. ALL 7 are against me. It's early in the season still, but in the simmed games against computer teams he is 0 for 10. Against me he is 7 for 7. No runner has been less than 14 steal. None have been 'pitchouts.'

sheesh

[Left-handed Badger]

Originally Posted by spider_monkey_b

Just to add a little to the discussion here, though not as in depth as some of the replies. I've been playing my games out (though in one-pitch mode to save time), and I am able to use pitch outs to my advantage. One opposing team has 4 players with 10 speed and steal (10 scale), and if I pitch out I throw them out fairly frequently, though not all the time.

The only aspect of stealing that really gets under my skin is when I tell a runner to steal 2nd on 4 consecutive pitches and he never gets a good enough jump. That seems to happen to me all the time.

Yes, I have same problem with getting that good jump.


But, to look at it another way though. OOTP is the only baseball games I have ever played that actually allows steals. I still High Heat where Mike PIazza would throw 80% of basestealers.

[injury log]

Originally Posted by spider_monkey_b

The only aspect of stealing that really gets under my skin is when I tell a runner to steal 2nd on 4 consecutive pitches and he never gets a good enough jump. That seems to happen to me all the time.

I think the game needs to do something to prevent the human player from running wild on the basepaths, for the sake of realism. Even though it is difficult to get a jump, I was able to quadruple the real-life MLB team SB record, just by stacking my team with fast guys and running as often as possible. I never used 'Send Forced', so who knows how many steals I could have had that way...

[RonCo]

Originally Posted by injury log

I think the game needs to do something to prevent the human player from running wild on the basepaths, for the sake of realism. Even though it is difficult to get a jump, I was able to quadruple the real-life MLB team SB record, just by stacking my team with fast guys and running as often as possible. I never used 'Send Forced', so who knows how many steals I could have had that way...

I'm not really sure this is the right way to look at it. If a real life human team/manager wanted to triple the SB record I would guess it would be possible specifically by doing exactly what you did--value incredible speed at all positions and hire a manger who is told to pull the trigger are every single opportunity. The reason is doesn't happen is that that's a risky way to creat run value. Power and OBP is much less risky. In addition, speed is the skill that degrades the quickest...so loading up with guys who can really run means you've likely created a team for the (very short) moment.

I am a huge proponent of realistic stat levels. However, I think the game should let a human in-game manager do what he wants. By definition, the design is already allowing the in-game manager to affect realism by choosing who to pitch to and who to walk (for example) and in the counter example of giving them the ability to create a team that _never runs_. I can of course chose to never send a runner and end up with zero SB. Not realistic, but in that case it's exactly how I want to run my team. Guys who are playing out every game are by definition not out for overall statistical realism, but instead are looking for realism in the instant--realism on the probabilities of any specfic action happening on demand. If they decide to play the game in such a fashion that they wind up with guys with 250 SB, so be it.

[injury log]

I think we're talking about slightly different things. To me, the 'Steal' button is an instruction to the runner to go if he has a good jump. So, I'm imagining, for example, with circa 1982-era Rickey Henderson, his manager was pressing that Steal button every pitch when Rickey was on base- the guy had the green light, and was running like mad. And despite a .398 OBP, he still only attempted 172 steals. Using only the steal button, my .316 OBP guy managed 186 attempts- and this with me too impatient to play out more than two thirds of my games. By all means, the human manager should be able to do whatever he likes, but the 'Send Forced' button is there to send a guy even if he doesn't have a good jump- the human manager can still run his players at will. But if I'm using the Steal button, my guys shouldn't choose to run much more often than real-life players who always have the green light.

And, of course, as you mention above, surprise is an important element in SB success. A real life team that ran almost every pitch would get caught more often than normal, I'd think- teams would start stronger-armed catchers, and pitch out a lot. OOTP's AI isn't smart enough to analyze opposing team strategy in this way; AI teams only pitch out on specific counts, for example, and I'd be (pleasantly) surprised if it were true that AI teams start better defensive catchers against fast teams.

[Chief Wahoo]

Originally Posted by statfreak

To "fix" the stolen base issue to my satisfaction, I resulted to adjusting the OOTP engine ini file (C:\Documents and Settings\user\Application Data\Sports Interactive\OOTP Baseball 2007\data\config\engine). I know others, including online leagues, have done the same. I bumped my GAME_STEALING_SUCCESS up to 120 and have found the results to be more realistic and closer to my expectations (when looking at individual players).

I have seen this fix posted in the past but unfortunately can't find the folder "Application Data". I have played seasons in different eras and am in a league now that I really enjoy but it is in the year 1902 and it seems the stolen base % is even more off than what some of you guys are reporting. In 1901 the 1st year of the league my stolen base leader had 74 stolen bases and 76 caught stealings.