Mean and Variance for Player Creation and Engine

LordTC · 05-19-2021, 03:12 PM

Would it be possible to offer control over mean and variance for things like player creation modifiers or engine behaviors? (Instead of just mean)

I think it would be cool to have tunable knobs for how big the difference between top and bottom players are. Both for fictional player creation and draft classes.

It might also be useful to have tunable knobs for how wide the spread of talent in draft class is. Some leagues might want to set a higher mean and lower variance so they get a lot of regulars but very few top tier players.

To give an example for the engine. If the average was set to 1.000 and the variance was set to 2.000 for pitcher errors then pitchers with good error ratings would make fewer errors and pitchers with bad error ratings would make more errors but the total errors for pitchers would be the same. For instance if the normal distribution was 7 errors a year with a standard deviation of 2, then a 1SD below average pitcher would average 9 errors in classic OOTP, but 11 errors with this new variance mod. A 1 SD above average pitcher would average 5 errors normally but 3 with this mod.

It's fine for me if the feature doesn't work quite as cleanly as this because of how it's coded under the hood. But having a knob that lets you adjust the performance difference between best and worst players would be extremely helpful for some of the leagues I want to run.

Cobby · 05-19-2021, 04:45 PM

Yes, I second this. I have found that when I do historical-fictional, that in the early years (pre-1945 or so) I almost never get any outstanding players. I was curious about this once so I took historical data and just looked at batting or pitching qualifiers and did a standard deviation for things like batting average and home run rate etc. Sure enough the standard deviation for almost everything was considerably higher pre-1945 and mostly stabilized post 1945. (I can't find my actual data at the moment, but it would be easy enough for someone to replicate) My hypothesis is that the player creation algorithm is tuned more for post-1945.

It would be cool to have the option of player creation variance dynamically changing through history.

native son · 05-19-2021, 06:00 PM

I third this. It's exactly the reason why I jump through so many hoops to tinker with player creation modifiers by position. I can live with the flattening of the rest, but batter home runs and pitcher strikes outs are maddeningly flat for fictional players from the deadball era until just about expansion era in nearly every version of OOTP for at least a decade.

KBLover · 05-20-2021, 01:12 PM

Quote:

Originally Posted by LordTC

Would it be possible to offer control over mean and variance for things like player creation modifiers or engine behaviors? (Instead of just mean)

Yes pls!

Makes it easier to fine tune results if I find that the lower end players are doing TOO poorly with the rest in line.

I would adore something like this.

I remember Baseball Pro '98 - a lot of the game had this kind of mean and variance engine tuning to it.

Would love for OOTP have something like this as well!

05-19-2021, 03:12 PM	#1
LordTC Major Leagues Join Date: Aug 2009 Posts: 479	Mean and Variance for Player Creation and Engine Would it be possible to offer control over mean and variance for things like player creation modifiers or engine behaviors? (Instead of just mean) I think it would be cool to have tunable knobs for how big the difference between top and bottom players are. Both for fictional player creation and draft classes. It might also be useful to have tunable knobs for how wide the spread of talent in draft class is. Some leagues might want to set a higher mean and lower variance so they get a lot of regulars but very few top tier players. To give an example for the engine. If the average was set to 1.000 and the variance was set to 2.000 for pitcher errors then pitchers with good error ratings would make fewer errors and pitchers with bad error ratings would make more errors but the total errors for pitchers would be the same. For instance if the normal distribution was 7 errors a year with a standard deviation of 2, then a 1SD below average pitcher would average 9 errors in classic OOTP, but 11 errors with this new variance mod. A 1 SD above average pitcher would average 5 errors normally but 3 with this mod. It's fine for me if the feature doesn't work quite as cleanly as this because of how it's coded under the hood. But having a knob that lets you adjust the performance difference between best and worst players would be extremely helpful for some of the leagues I want to run.

05-19-2021, 06:00 PM	#3
native son Minors (Double A) Join Date: May 2002 Location: New York, NY Posts: 100	I third this. It's exactly the reason why I jump through so many hoops to tinker with player creation modifiers by position. I can live with the flattening of the rest, but batter home runs and pitcher strikes outs are maddeningly flat for fictional players from the deadball era until just about expansion era in nearly every version of OOTP for at least a decade. __________________ life in lowercase

05-19-2021, 04:45 PM	#2
Cobby Minors (Triple A) Join Date: May 2016 Posts: 252	Yes, I second this. I have found that when I do historical-fictional, that in the early years (pre-1945 or so) I almost never get any outstanding players. I was curious about this once so I took historical data and just looked at batting or pitching qualifiers and did a standard deviation for things like batting average and home run rate etc. Sure enough the standard deviation for almost everything was considerably higher pre-1945 and mostly stabilized post 1945. (I can't find my actual data at the moment, but it would be easy enough for someone to replicate) My hypothesis is that the player creation algorithm is tuned more for post-1945. It would be cool to have the option of player creation variance dynamically changing through history.