Do we know what formula the engine uses to decide the result of a stolen base attempt?
I've been trying to figure this out but can't really come up with it.

Let's say an excellent catcher, like A++ arm, I will say 50% of throwing out base stealers against an excellent base stealer like 90% successful attempts. What is the % of outcome of this match up?

Compared to a D- arm, like 20% vs that same 90% base stealer?

Or the match up between an average C, let's say 30%, and an average base stealer let's say 65% what should we expect?

Anybody knows the exact formula?

I don't know the exact formula (unless it changed since this chat it involved different weightings divided across three eras), but after the weighting, I think it's just a dice roll on the output of the CS%*SB%, so 0.50*0.90=0.450. the dice is rolled and if it is 450 or lower it's a SB, otherwise it's a CS.

Posted by just4me on 4/30/2025 1:50:00 PM (view original):
I don't know the exact formula (unless it changed since this chat it involved different weightings divided across three eras), but after the weighting, I think it's just a dice roll on the output of the CS%*SB%, so 0.50*0.90=0.450. the dice is rolled and if it is 450 or lower it's a SB, otherwise it's a CS.

Not sure if I understood what you wrote correctly but if it's 0.50 x 0.90 = .0450 it wouldn't make sense when you'd compare, say a base runner who has a 70% success rate vs a catcher who throws out only 20% of runners, 0.20 x 0.70 = 0.14
I'd guess that a 70% baste stealer would go up to around 80-85% vs a 20% catcher.

I tried to put these data to AI to get it to find me a formula but it gave me all sorts of formulas that didn't make sense, including the one you used.

The mystery remains.

In short, they use an era based weight first to adjust the SBA metric to determine whether a SB is even attempted, then they use a similar era based weight to adjust the CS% (think of it like normalization, but instead of to the catcher and basestealers years they are just normalizing the catchers CS% based on an era (probably something like 1885-1919, 1920-1960, 1961-2024, or something vaguely close to that)). So the raw CS% we see isn't what actually gets used in the formula, but whatever comes out of that weighted adjustment/normalization. Once they get that value, they just multiply it by the batters SB% rate and roll the dice to determine SB or CS.

In my example above I inverted it, keep it as I have it and just apply it to the other side (CS instead of SB, so .55% of SB in my example, and 86% in yours), or it could be 1-(.5*.9), or in your example, 1-(.2*.7), for the success rate to show the opposite of what I put.

My educated guess would that it uses a "Log5" (https://en.wikipedia.org/wiki/Log5) formula with some baseline CS% that is hidden from us. Whenever comparing the success of "A" vs the failure of "B", a league average is needed. Batting average is a great example. How would a .406 Ted Williams fair versus a pitcher whose average against is .350? Without a league baseline, you would might say Williams would bat between .350-.406. But we know .406 is outstanding and BAA of .350 is terrible. We could define our variables as A = .406, B = .350 and L = .260. Using the "log5" formula of [A*B*(1-L)]/[(A*B)-(A*L)-(B*L)+L] would result in Williams batting .511.

If we assume the hidden baseline CS% is around 33%, then a 90% baserunner vs 50% catcher would be 81.5%. A 70% baserunner versus a 20% catcher would be 82%.

A link to graph/calculator if you wanted to play with some numbers: https://www.desmos.com/calculator/qqw0jdkpqj

Just4me has an excellent point where the baseline CS% could be dynamic to different eras.

Where does the speed of the player come into it? A guy with 90 speed and 80% RL success is always more successful in the sim than a guy with 70 speed and the same RL success rate, no?

speed doesnt factor into the stolen base success rate. its used for taking extra bases when already on base and beating out double plays.

Slightly off topic, but has anyone noticed weirdness with 2024 players' SB rates? I'm in a 2024 prog that just finished and there were a ton of players who doubled their real life numbers, and with much better success rates. Part of it may be that they ran only on 2024 catchers, and for some reason there are only two with A+ arms, neither of them full timers.

Here are ten players from the league, with sim stats (SB/CS) first and real life stats second:

Elly De La Cruz 103-7 (67-16)
Andres Gimenez 88-3 (30-5)
Jarren Duran 80-3 (34-7)
Jacob Young 71-7 (33-10)
Brenton Doyle 67-4 (30-5)
Jackson Chourio 62-12 (22-7)
Maikel Garcia 56-0 (37-2)
Blake Perkins 52-2 (23-5)
Zack Gelof 50-0 (25-3)
Mookie Betts 37-2 (16-2)

Posted by crazystengel on 5/20/2025 1:49:00 PM (view original):
Slightly off topic, but has anyone noticed weirdness with 2024 players' SB rates? I'm in a 2024 prog that just finished and there were a ton of players who doubled their real life numbers, and with much better success rates. Part of it may be that they ran only on 2024 catchers, and for some reason there are only two with A+ arms, neither of them full timers.

Here are ten players from the league, with sim stats (SB/CS) first and real life stats second:

Elly De La Cruz 103-7 (67-16)
Andres Gimenez 88-3 (30-5)
Jarren Duran 80-3 (34-7)
Jacob Young 71-7 (33-10)
Brenton Doyle 67-4 (30-5)
Jackson Chourio 62-12 (22-7)
Maikel Garcia 56-0 (37-2)
Blake Perkins 52-2 (23-5)
Zack Gelof 50-0 (25-3)
Mookie Betts 37-2 (16-2)

The effect of the larger bases, new pick off/ throw to first rules and teams and players having now had time to adapt to them maybe?

no one ever stress-tested an era where success rates are so high since there never was one and it broke the sim

I used 2024 Duran in an open league recently, playing in Kaufman. His hit 26 triples (led the league with 14 in real life) which I wasn’t too surprised at, but his steals were way up also. 59 steals and only caught 10 times.

Posted by crazystengel on 5/20/2025 1:49:00 PM (view original):
Slightly off topic, but has anyone noticed weirdness with 2024 players' SB rates? I'm in a 2024 prog that just finished and there were a ton of players who doubled their real life numbers, and with much better success rates. Part of it may be that they ran only on 2024 catchers, and for some reason there are only two with A+ arms, neither of them full timers.

Here are ten players from the league, with sim stats (SB/CS) first and real life stats second:

Elly De La Cruz 103-7 (67-16)
Andres Gimenez 88-3 (30-5)
Jarren Duran 80-3 (34-7)
Jacob Young 71-7 (33-10)
Brenton Doyle 67-4 (30-5)
Jackson Chourio 62-12 (22-7)
Maikel Garcia 56-0 (37-2)
Blake Perkins 52-2 (23-5)
Zack Gelof 50-0 (25-3)
Mookie Betts 37-2 (16-2)

Wow, those are pretty crazy numbers...

My first thought was that, as italyprof posted, with the new rules and larger bases it would definitely make it easier for base runners to steal bases, thus negatively impacting catcher's CS %, but the numbers you posted are totally off the charts.

Would be interesting to know the various catcher's arms in that league.
Too bad there was no Ohtani in that league, he could have gone like 90-0 or something.

I've had a look at the 2024 catchers.
Of those with more than 400 PA, the best arm is B+, then there is one C and everybody else is C- to D-
Probably the weakest year ever for catcher's arms which totally explains those SB numbers in that 2024 prog league.

The MLB SB success rate was almost 80% in 2024; there are only 6 players rated at catcher who have an arm rating above C+