For those who play stats only...

[Papa3]

Originally Posted by Questdog

The stats that OOTP generates for the draft eligible players are much more reliable as a guide to quality than stats generated from feeder leagues. Feeder leagues are a cool idea, but hamstrung in OOTP by the fact that these young players' actual ratings have almost 0 correlation to their potential ratings. Thus, drafting based on feeder league stats, especially high schoolers, is about as effective as picking at random.

One advantage I've noticed to having feeder leagues is OOTP's tracking of player development. If I see a guy who's consistently getting rating increases throughout his 4 years of HS or college feeder play, I'm more inclined to pick him over a guy whose ratings are stagnant.

Something about feeders I'm disappointed in, though, is how many players have better seasons as a freshman than as a senior. I don't often see the gradual four-year improvement there should be.

[HerbD]

Originally Posted by Questdog

This is an illusion....in real life the best players in high school and college turn out to be the best players in MLB.....some are busts and some players that weren't rated quite as high turn out to be better, but even they were good players as amateurs......in real life guys that hit .180 in high school don't make it to the pros....but in OOTP, guys that hit .180 could turn out to be superstars.....this is because the ratings that drive the stats in feeder leagues are basically random.....if you do a draft knowing what the potential ratings actually are (according to your scout), you will end up with a more realistic draft....some busts and some low round surprises.

And when you do a feeder league and draft only with stats, only you are looking at the stats...the AI is looking at the ratings......so your draft is basically random (college players are not quite random), while the AI is picking based on the ratings......

This is true and may play out that way in a single player game but when used in an online league everyone is playing with the same "fog of war" and the "illusion" of realism adds a lot to the entire league.

[Questdog]

Originally Posted by HerbD

This is true and may play out that way in a single player game but when used in an online league everyone is playing with the same "fog of war" and the "illusion" of realism adds a lot to the entire league.

It is still an illusion as those stats have absolutely nothing to do with how good or bad the player will end up being.....In an online league drafting with stats only (no peeking at reports?!), you might as well just pass the players out at random.....

[The Wolf]

"It's your game, play it your way."

Meanwhile, I'll play it my way.

Enjoy!

[HerbD]

Originally Posted by Questdog

It is still an illusion as those stats have absolutely nothing to do with how good or bad the player will end up being.....In an online league drafting with stats only (no peeking at reports?!), you might as well just pass the players out at random.....

There is no way that I know of to get rid of the OSA reports so they are fair game.

[Questdog]

Originally Posted by HerbD

There is no way that I know of to get rid of the OSA reports so they are fair game.

Then they are your best friend at draft time.....

[Fyrestorm3]

Originally Posted by SirMichaelJordan

For that very reason I will never turn off individual ratings (I may only use potential sometimes) my version of "stats only" is to hide OVR and set my AI Evaluation to favor stats. Besides that real life scouts use a number rating system. I always have scouts set to very low accuracy.

Interesting approach; I might try that sometime.

I usually leave OVR/POT on, as stars, and with low scouting accuracy. That gives me a quick glimpse of how the AI sees a player and allows for me to scan over my minor leagues and pick out my top prospects easily. Other than that, I ignore them entirely.

Your idea is a pretty good compromise. You can see how good a player COULD be, while still having to evaluate based on his performance. And there's no worries about having to turn on ratings at draft time. I like it.

[Honorable_Pawn]

Originally Posted by Questdog

.....this is because the ratings that drive the stats in feeder leagues are basically random.....

I think the problem with feeder league stats is a function of degrees of separation from the highest rated player to the lowest rated player. In the ML level (on a scale from 1-00...I know the engine uses 256 but it is easier for me to talk 1-100) the range runs from the top guys in the 90's all the way to the 40's and 30's. There is a large degree of separation from the best rated player to the lowest rated player. This allows the league totals modifiers to be distributed so that the difference over a few hundred AB's can be noticed.

The problem with feeders is that you are seeing ranges from single digits to teens. The degree of separation is too small so that the stats come out looking mostly random. There isn't enough separation for the league totals to be distributed so that it doesn't look random.

I have an idea for a potential solution.

If Marcus used decimal points for any ratings below a certain threshold that would allow him an extra 100 points of separation. This would allow the league totals modifiers to be distributed so that guys that really are better will perform better.

$0.02

[Questdog]

Originally Posted by Honorable_Pawn

I think the problem with feeder league stats is a function of degrees of separation from the highest rated player to the lowest rated player. In the ML level (on a scale from 1-00...I know the engine uses 256 but it is easier for me to talk 1-100) the range runs from the top guys in the 90's all the way to the 40's and 30's. There is a large degree of separation from the best rated player to the lowest rated player. This allows the league totals modifiers to be distributed so that the difference over a few hundred AB's can be noticed.

The problem with feeders is that you are seeing ranges from single digits to teens. The degree of separation is too small so that the stats come out looking mostly random. There isn't enough separation for the league totals to be distributed so that it doesn't look random.

I have an idea for a potential solution.

If Marcus used decimal points for any ratings below a certain threshold that would allow him an extra 100 points of separation. This would allow the league totals modifiers to be distributed so that guys that really are better will perform better.

$0.02

That is possible, but the real problem is that these single digit ratings have no correlation to the potential ratings.....

A player with an 18 contact rating may have a potential contact rating of 40 and a player with an 8 contact rating may have a potential rating of 90.....now I don't believe every player with a higher actual rating should have a correspondingly higher potential rating, but it should be the norm rather than the exception.....

[Honorable_Pawn]

Originally Posted by Questdog

That is possible, but the real problem is that these single digit ratings have no correlation to the potential ratings.....

A player with an 18 contact rating may have a potential contact rating of 40 and a player with an 8 contact rating may have a potential rating of 90.....now I don't believe every player with a higher actual rating should have a correspondingly higher potential rating, but it should be the norm rather than the exception.....

Oh definitely this is a problem. But if we check favor talent for the feeder leagues in v. 15 that "8-guy" should get more playing time. So the total number of AB's should be a good indicator.

Maybe the rating distribution in feeders should be weighted. The acutal number of "hits" that a plyer gets will not be based upon actual only like it is now. What if the number of "hits" was 75% potential and 25% actual?

[Syd Thrift]

Originally Posted by Honorable_Pawn

I think the problem with feeder league stats is a function of degrees of separation from the highest rated player to the lowest rated player. In the ML level (on a scale from 1-00...I know the engine uses 256 but it is easier for me to talk 1-100) the range runs from the top guys in the 90's all the way to the 40's and 30's. There is a large degree of separation from the best rated player to the lowest rated player. This allows the league totals modifiers to be distributed so that the difference over a few hundred AB's can be noticed.

The problem with feeders is that you are seeing ranges from single digits to teens. The degree of separation is too small so that the stats come out looking mostly random. There isn't enough separation for the league totals to be distributed so that it doesn't look random.

I have an idea for a potential solution.

If Marcus used decimal points for any ratings below a certain threshold that would allow him an extra 100 points of separation. This would allow the league totals modifiers to be distributed so that guys that really are better will perform better.

$0.02

Yeah, that is a really, really awful idea. Looking at this from a programmer's perspective... if Markus is using 8-bit integers to store this data (which I am pretty positive he is because we've run into issues in the past with values over 255 "clipping" back down to zero and vice versa), even using the double type in C# would increase the storage space required for each rating 8-fold, as the double type is 64 bits in length. I'm not saying it would increase the memory for *everything* by 8 - I'm equally sure that many of the stats are using larger ints and floating-point numbers of their own - but it'd significantly increase the size.

On top of that, Markus would probably have to rewrite large chunks of his code to interact with floating point numbers, and that in turn would probably expand the amount of RAM used by the application as well (because every time you stick someone's rating in the stack, you have to allocate 8 times as much as before, as above).

[Honorable_Pawn]

Welp...that leaves weighting potential higher than actual in feeder leagues in order to distribute the league total modifiers.

[Fyrestorm3]

Originally Posted by Honorable_Pawn

Welp...that leaves weighting potential higher than actual in feeder leagues in order to distribute the league total modifiers.

Which really isn't a bad idea.

[sprague]

Originally Posted by Honorable_Pawn

Welp...that leaves weighting potential higher than actual in feeder leagues in order to distribute the league total modifiers.

Isn't that how it is?

It always seemed to me that the stats for players were being generated only by potential, that actual in feeder leagues had no impact at all

[sprague]

And a question for SO guys...as I am doing a 3 year test run now to catch my errors before going to a full league

in rating scales: actual potential other

Should those be on? or should they be with none displayed?
Could someone post a screen shot of the settings for ratings in SO

I'm not sure...

Thanks

EDIT re above
Read through again "how I learned to love" thread
what I get from that is consensus is
a- turn off actual ratings
b- some suggest keeping potential ratings on, but at very low scale 1-5, or 2-8
c- most seem to suggest keeping other ratings (field run) on as those are abilites one can see w naked eye, and that seems logical

I may try w the low potential ratings for the first league, to see how that goes, see if it still gives me the feel of too much of an advantage in knowing players

[Papa3]

Originally Posted by sprague

And a question for SO guys...as I am doing a 3 year test run now to catch my errors before going to a full league



in rating scales: actual potential other



Should those be on? or should they be with none displayed?

Could someone post a screen shot of the settings for ratings in SO



I'm not sure...



Thanks

I think the answer to your first question is checking a box in Global Setup that says something like Use AI Evaluation For Overall Rating. I use that setting but I'm also advocating the use of ratings. I use the 2-8 scale but there's also a 1-5 scale if you prefer it to be a bit more vague.

[sprague]

Originally Posted by Papa3

I think the answer to your first question is checking a box in Global Setup that says something like Use AI Evaluation For Overall Rating. I use that setting but I'm also advocating the use of ratings. I use the 2-8 scale but there's also a 1-5 scale if you prefer it to be a bit more vague.

Thanks. I edited my q above after reading through more threads

Going to be an interesting test...

[Fyrestorm3]

Originally Posted by sprague

And a question for SO guys...as I am doing a 3 year test run now to catch my errors before going to a full league

in rating scales: actual potential other

Should those be on? or should they be with none displayed?
Could someone post a screen shot of the settings for ratings in SO

I'm not sure...

Thanks

EDIT re above
Read through again "how I learned to love" thread
what I get from that is consensus is
a- turn off actual ratings
b- some suggest keeping potential ratings on, but at very low scale 1-5, or 2-8
c- most seem to suggest keeping other ratings (field run) on as those are abilites one can see w naked eye, and that seems logical

I may try w the low potential ratings for the first league, to see how that goes, see if it still gives me the feel of too much of an advantage in knowing players

Everyone does it differently. Some say all ratings should be off, others keep just the speed and fielding ratings on (which admittedly makes the most intuitive sense), and others like myself use some form of OVR/POT as well. Some people play with Normal scouting accuracy, others say it has to be set to Very Low. Totally up to the player. The essence of stats-only is getting rid of the ability ratings and making your evaluations based on stats alone.