Home | Webstore
Latest News: - OOTP 15: Update #6 Released! - OOTP 15 Released! - FHM 2014: Version 1.6.19 Available! - iOOTP Baseball 2014 for iOS Available NOW! - Beyond the Sideline Football Announced! - Title Bout Championship Boxing 2.5 released!

OOTP 15 Offseason Special: 50% Off!

  

Go Back   OOTP Developments Forums > Earlier versions of Out of the Park Baseball > Earlier versions of OOTP: General Discussions

Earlier versions of OOTP: General Discussions General chat about the game...

Reply
 
LinkBack Thread Tools
Old 06-30-2011, 10:50 PM   #121 (permalink)
Major Leagues
 
Join Date: Jun 2007
Posts: 397
Thanks: 64
Thanked 29x in 20 posts
Quote:
Originally Posted by SiN8 View Post
Great job on this experiment! Perhaps the cutoff ceiling is 200 and anything above is not accounted for. You might want to retry with players 190 to 200.
That's a good idea. I'll give it a shot at some point this weekend. Gonna play the game for a little while though.
jar2574 is offline   Reply With Quote
Thank you for this post:
Dr. dru (07-01-2011)
Old 07-01-2011, 12:26 AM   #122 (permalink)
All Star Reserve
 
Dr. dru's Avatar
 
Join Date: Apr 2011
Location: Boston, MA
Posts: 647
Thanks: 206
Thanked 64x in 44 posts
Quote:
Originally Posted by jar2574 View Post
I tried to test whether the AI uses scouts in drafts. Unfortunately, the results of my testing were inconclusive.

I took a brand new 10-team league with no history and deleted all the players. Then I created 10 players and turned them into 18-year old catchers with current and potential ratings at 250, 249, 248, etc., respectively, and with experience at 200 for catching and 0 for every other position.

I gave them all max in personality traits and made them all 1's at every injury-rating. They all weigh the same and throw and bat right-handed.

I set player ratings to 100% of evaluation.

I then ran an inaugural draft 6 times, 3 times with scouting and 3 times with scouting off.

The 10 players were taken in the exact same order every time, which would seem to indicate that scouts do not matter. But the order in which they were taken was surprising - 248, 249, 242, 244, 247, 250, 241, 246, 245, and 243. (They were named with the ratings given to them.)

So on the one hand, scouted and scout-less drafts happened in the same order, with random teams in different slots. But on the other hand, I can't figure out why the players went in that order. The only thing I can think of is that (1) I missed a variable in the player editor (not makeup, not handedness, not height or weight, not G/F ratio, not stamina, not velocity, not Hold, and not current and potential ratings, but something else.) Or (2) each player has an internal development engine and will age and progress differently, and that development engine cannot be edited in the player editor but has an effect on AI evaluation. Or (3) something else I'm not smart enough to think of.

Interestingly, the player development report had these guys listed in the same order.

That's enough testing for me for one day. Some mysteries of the game will remain mysteries.

This is some good work. It definitly seems like the AI does not use it's hired scouts. However, it's clear that at times it does use some type of formula other than true ratings (Altough true ratings are definitly a factor and scouts are not).
__________________
If you don't feel well play OOTP all night and call me in the morning!

Dr. Dru is your doctor. Listen to your doctor. He knows what's best for you.
Dr. dru is offline   Reply With Quote
Old 07-01-2011, 06:22 AM   #123 (permalink)
All Star Starter
 
beorn's Avatar
 
Join Date: Dec 2001
Location: near Rochester, NY
Posts: 1,138
Thanks: 31
Thanked 114x in 63 posts
Very interesting results.

A possible explanation for this: The game engine might assign a value to each player, for use by the AI. It may or may not be totally accurate, but it is the same rating used by every AI team, at least during drafts. Possibly this is also what the AI uses for the portion of its evaluation of an active player that is based on ratings?

Wasn't there an option somewhere in setup to assign a draft value to each player? My memory may be playing tricks on me, but I feel I remember that.
beorn is online now   Reply With Quote
Old 07-01-2011, 02:08 PM   #124 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
I think it is in historical drafts. You can assign some sort of draft ratings so the players go to the right teams or at least in the right order. So you can make it where Babe Ruth will be the number 1 pick in the draft.
Biggio509 is offline   Reply With Quote
Old 07-01-2011, 02:44 PM   #125 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Quote:
Originally Posted by jar2574 View Post
I tried to test whether the AI uses scouts in drafts. Unfortunately, the results of my testing were inconclusive.


So on the one hand, scouted and scout-less drafts happened in the same order, with random teams in different slots. But on the other hand, I can't figure out why the players went in that order.
One thing to note here. The AI does not use scouts and does not use scouts in the same way are two different statements. There is not one evaluation AI but several.

AI evaluation
Uses scouted CURRENT ratings as part of the makeup of overall rating or all the overall rating when there are no stats.

Used for
1. Partially in trades. Overall potential ratings based on each potential rating also play a role that I think is affected by favoring prospects, favor vets, or neutral. Also the need for the position the player plays plays a role.
2. Promotions and Demotions (Roster AI)
3. Free agent signings along with potential I presume. Again used with potential and need for position somehow.

Lineup evaluation
Unknown

Used for lineups
AI evaluation has 0 effects on lineups.

Draft AI
unknown

Used for drafts. AI evaluation has 0 effect on drafts and seems to have no effect on initial drafts as shown. Note here only ratings can be used because there are no stats.

BA top prospects Scouting
In earlier versions this was an average scout. In the 11 manual it says this can vary daily depending which OSA staff member does the report. I think the daily scouting aspect is why this is so much better than your scout's every now and then reports. It is still unclear as to how scouting is applied here.

In short some of the AI's "scouting" or better said decision making can be affected by things you choose. Other parts can not.

Scouting is the information received other than stats. AI and humans use some sort of scouting. How that scouting is done depends on AI evaluation for overall but can not be changed for potential.

That information is used somehow. There is no way to make the AI use scouts as a human would. We are not consistent. The AI is. The AI goes by set formulas to use the scouted information. In lineup AI and appears draft AI we can't alter that. One of the things we don't know if is the average to decent scout always under or over rates in the same way by the same amount on the same player, then it may look like they now the real rating when actually they all know the real rating say +10%. If every scout has the exact same error or even close to the same error on every player then the order would not changed. It is unknown if your scout is undervaluing a player if the AI also undervalues by the same or different amount of if some teams might over value them. If all scouts error in the same direction order will not changed unless the players are close and there is big difference in the error. They will still rank the players the same. So lets say there is a 10% difference in player 1 and 2 a five percent error will still get the order right.

Also I think there is some component in the draft for times viewed which improves scouting. It would only make sense that every team in the league would have good info on the 35 y/o superstar and not so good on the 18 y/o hopeful. So what could be happening here at the top of the draft is the times viewed component of scouting accuracy is very high for the top draft picks meaning teams are seeing real or close to real ratings because their scouting is excellent on top players.

I am not sure if the AI is using real ratings or just gets a bonus to scouting the human may not.

Last edited by Biggio509; 07-01-2011 at 02:50 PM.
Biggio509 is offline   Reply With Quote
Old 07-01-2011, 04:57 PM   #126 (permalink)
Major Leagues
 
Join Date: Jun 2007
Posts: 397
Thanks: 64
Thanked 29x in 20 posts
Quote:
Originally Posted by Biggio509 View Post
One of the things we don't know if is the average to decent scout always under or over rates in the same way by the same amount on the same player, then it may look like they now the real rating when actually they all know the real rating say +10%. If every scout has the exact same error or even close to the same error on every player then the order would not changed. It is unknown if your scout is undervaluing a player if the AI also undervalues by the same or different amount of if some teams might over value them. If all scouts error in the same direction order will not changed unless the players are close and there is big difference in the error. They will still rank the players the same. So lets say there is a 10% difference in player 1 and 2 a five percent error will still get the order right.
I do know that if you "act as" different teams and you use the scout for each of those teams, you'll see different ratings. So I don't think scouts err in the same direction - some undervalue while others overvalue.

One of my strategies is hiring a new scout each year during the 5 years that I sim my league before starting an inaugural draft. If you have 5 scouts giving the same ratings about a player, then you can be assured they are correct. But you will often see an outlier where one scout under- or over-values a player. And you generally want to throw out his assessment and use the other 4 guys.
jar2574 is offline   Reply With Quote
Old 07-01-2011, 05:51 PM   #127 (permalink)
Minors (Single A)
 
Join Date: May 2011
Posts: 79
Thanks: 0
Thanked 7x in 4 posts
I wish I could find the thread, but a guy on here was wondering why the AI snapped up all of his apparently rubbish players before they could clear waivers. I surmised at the time that these players may have been better than their ratings and stats would suggest, this thread would explain how the AI 'knows' this...
SteveV is offline   Reply With Quote
Thank you for this post:
Dr. dru (07-01-2011)
Old 07-03-2011, 11:31 AM   #128 (permalink)
Hall Of Famer
 
PSUColonel's Avatar
 
Join Date: Feb 2002
Posts: 5,107
Blog Entries: 37
Thanks: 542
Thanked 865x in 472 posts
Quote:
Originally Posted by SteveV View Post
I wish I could find the thread, but a guy on here was wondering why the AI snapped up all of his apparently rubbish players before they could clear waivers. I surmised at the time that these players may have been better than their ratings and stats would suggest, this thread would explain how the AI 'knows' this...
yup...and still no comment

Last edited by PSUColonel; 07-03-2011 at 11:33 AM.
PSUColonel is offline   Reply With Quote
Old 07-03-2011, 07:13 PM   #129 (permalink)
Hall Of Famer
 
Join Date: Apr 2007
Location: Toronto
Posts: 5,912
Thanks: 554
Thanked 1,665x in 756 posts
Quote:
Originally Posted by Biggio509 View Post
The 11 manual says that AI evaluation uses percieved ratings which I believe are scouted. It may be semantics here but no the AI does not use scouts as we would. The AI does use scouting differently. However, the model for how a scout determines ratings is the same.
I don't know why you're being so insistent about this position when you have no evidence to support it. The facts are these:

* AI teams do not use scouting the way a human team does. That is, you could not ask an AI team for its scouting report on a player and get any kind of answer - they don't keep scouting reports. This is by design; simming would take 10 times longer if the AI used full scouting;

* When scouting is on, the AI's perception of ratings is 'blurred'. That is, the AI sees the true rating plus or minus some 'blur factor', which is presumably determined in part by scouting budgets and scout ratings. There is surely also a random component that differs from player to player. I have no idea how that is implemented, but if you use cloned players, they are all likely to be blurred in the same way if the game is using some random detail from the player record (for example, if the game uses the day of the month on which a player was born to determine how the blur is skewed, cloned players will all be blurred in the same way, which may in part explain Sin8's results). That blurry ratings calculation is done on the fly; it isn't stored anywhere.

So if by 'the AI uses scouting' you mean 'the AI doesn't see the true ratings of each player', that's correct, as far as I know. If you mean that the AI compiles scouting reports, that's just wrong.

I do not hold at all with the position that 'the AI should use scouts just like the user does'. The user is already a billion times smarter than the AI; the last thing I would want is to make the AI any stupider. When the human makes any transactions decision, he incorporates all kinds of information besides the possibly unreliable report from a scout - and the user is very capable of evaluating just how unreliable that scouting report is. If, say, in the amateur draft my scout thinks two players are potential Hall of Fame talents, and one requests a $8m signing bonus and the other is willing to sign for Slot, I'll pick the $8m guy every time and dismiss the other scouting report as hyperbole. I want the AI to be able to do that too, and since the AI is inherently pretty stupid (as is AI in every domestic computer application in the universe), at least letting the AI see a very close approximation of true ratings prevents the game from becoming absurdly easy.

As for in-game reports like the BA prospect reports, those are based on true ratings. The ranking is not, however, based precisely on the same scale as are the POT numbers from 20-80 (or the star ratings, if you use those instead). So if you rank every prospect by their POT rating, that ranking will not match perfectly with the BA prospect ranking. I don't see any issue with that since in real life there are also consensus prospect reports available to all teams; the real life BA reports aren't based on one scout's impression of prospect talent.
injury log is offline   Reply With Quote
3 thanks for this post:
Dr. dru (07-03-2011), Isura (01-03-2012), SiN8 (07-03-2011)
Old 07-03-2011, 07:36 PM   #130 (permalink)
All Star Starter
 
beorn's Avatar
 
Join Date: Dec 2001
Location: near Rochester, NY
Posts: 1,138
Thanks: 31
Thanked 114x in 63 posts
Quote:
Originally Posted by injury log View Post
As for in-game reports like the BA prospect reports, those are based on true ratings. The ranking is not, however, based precisely on the same scale as are the POT numbers from 20-80 (or the star ratings, if you use those instead). So if you rank every prospect by their POT rating, that ranking will not match perfectly with the BA prospect ranking. I don't see any issue with that since in real life there are also consensus prospect reports available to all teams; the real life BA reports aren't based on one scout's impression of prospect talent.
I am not so sure about this part. Run a test league with scouting, and after a few seasons, open up the players at the top of the prospect rankings in editor mode. You are likely to find some pretty questionable players near the top of the list.

I suspect that this ranking is much like AI scouting, based on true ratings, but deliberately blurred.
beorn is online now   Reply With Quote
Old 07-03-2011, 07:44 PM   #131 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Quote:
Originally Posted by injury log View Post
I don't know why you're being so insistent about this position when you have no evidence to support it.
The evidence is in the manual. There is ZERO evidence to support the AI uses real ratings at ANY point. Is there evidence the AI might have scouting cheats or some built in cheats to even the playing field? Yes. However there is ZERO evidence that says the AI must use real ratings.
Biggio509 is offline   Reply With Quote
Old 07-03-2011, 07:48 PM   #132 (permalink)
Minors (Triple A)
 
Join Date: May 2003
Posts: 225
Thanks: 35
Thanked 31x in 22 posts
Quote:
Originally Posted by injury log View Post
I don't know why you're being so insistent about this position when you have no evidence to support it. The facts are these:

As for in-game reports like the BA prospect reports, those are based on true ratings. The ranking is not, however, based precisely on the same scale as are the POT numbers from 20-80 (or the star ratings, if you use those instead). So if you rank every prospect by their POT rating, that ranking will not match perfectly with the BA prospect ranking. I don't see any issue with that since in real life there are also consensus prospect reports available to all teams; the real life BA reports aren't based on one scout's impression of prospect talent.
The 'facts' you state in your post are your conclusions based upon your observations, not facts. For example, at the bottom of the post you state the 'fact' that the BA prospect reports are based upon true ratings. Would you care to share your proof? The prospect ranking in past iterations of OOTP was based upon the ratings of a 'league average scout' according to the description of the report within the program. That explanation was dropped when the affiliation with Baseball America started. That does not necessarily mean the program was rewritten, but perhaps it was. My observations using dynamic scouting convince me that these are not true ratings, but I submit that as an opinion, not a fact. I do find it unlikely that a true rating exploit has been available for years and no one has noticed until now. The only fact I offer to the discussion is this: the only definitive description of the top prospect report was its description as that of a 'league average scout', several years ago. Please share your proof this has changed in OOTP 12.
HH20xx convert is offline   Reply With Quote
Old 07-03-2011, 07:49 PM   #133 (permalink)
Hall Of Famer
 
endgame's Avatar
 
Join Date: Aug 2002
Posts: 12,831
Thanks: 1,695
Thanked 2,016x in 1,055 posts
Quote:
Originally Posted by beorn View Post
I am not so sure about this part. Run a test league with scouting, and after a few seasons, open up the players at the top of the prospect rankings in editor mode. You are likely to find some pretty questionable players near the top of the list.

I suspect that this ranking is much like AI scouting, based on true ratings, but deliberately blurred.
To the best of my recollection- and I'll look into it to support with documented quotes if they're available -the BA rankings and many of the in-game reports are OSA-based. I "believe" they may be, on occasion colored with the blur injury has alluded to in his post above, but for the most part, it's OSA. I can't attest to which report does what, but for the most part, regardless of scouting quality, whose report, the top 10-15 will be the same - with a rare exception - but the order of those rankings may be altered, i.e. #5 on one may be #13 on another. In that respect, it's really not that different than RL sources that disagree on where to place a player in the rankings. But as to true ratings? While they may form the basis of an AI judgment, the only time, to the best of my knowledge, these are utilized consistently is with scouting off.
__________________
"Try again. Fail again. Fail better." -- Samuel Beckett
_____________________________________________
endgame is offline   Reply With Quote
Old 07-03-2011, 07:49 PM   #134 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Quote:
Originally Posted by injury log View Post

* AI teams do not use scouting the way a human team does. That is, you could not ask an AI team for its scouting report on a player and get any kind of answer - they don't keep scouting reports. This is by design; simming would take 10 times longer if the AI used full scouting;
Never denied this. Read again as I have stated the AI can not use scouting as a human does. It has to apply scouting information in a consistent manner. We humans never do that. We use heuristics. AI's use mathematical formulas.

How does one ask an AI team for its scouting? You can't you can only see what your scout is telling you. I was explaining exactly that the AI does not keep scouting reports. There are sort of generated as the AI needs info. That is why when you edit a player and then shop them around the AI knows ratings have changed. This in no way proves the AI knows real ratings. It just shows AI "scouting" picks up on even small changes made in the editor when you click shop player around. All we know is the AI sees ratings have gone up, at least some teams do, not the AI knows the real ratings.

Again what I am simply arguing is that there is no proof the AI knows real ratings and no evidence support that. There is evidence that the AI evaluation system which has some built in obfuscation does pick up on rating changes when you shop a player.

Last edited by Biggio509; 07-03-2011 at 07:56 PM.
Biggio509 is offline   Reply With Quote
Old 07-03-2011, 07:52 PM   #135 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Quote:
Originally Posted by injury log View Post
* When scouting is on, the AI's perception of ratings is 'blurred'.
Again no argument from me. It is Dr. Dru and a couple of others, although I think Dr. Dru and PSU colonel are the same person, that are arguing the AI's perception of ratings are not blurred in trades or any other time. We are in agreement here.
Biggio509 is offline   Reply With Quote
Old 07-03-2011, 07:55 PM   #136 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Quote:
Originally Posted by injury log View Post

So if by 'the AI uses scouting' you mean 'the AI doesn't see the true ratings of each player', that's correct, as far as I know. If you mean that the AI compiles scouting reports, that's just wrong.
The first is what I am saying. I don't think the AI compiles a report on each player it refers back to. Like you said it would take too long. I do think the AI does not even perform an evaluation on a player, in other words look at its blurred ratings and use those in the evaluation process, until it is needed. So the AI might "scout" players when it looks to sign FAs, do a trade, or when you shop a player around. The "scouting" and evaluation is why you see a delay, sometimes a long one when you click shop around.
Biggio509 is offline   Reply With Quote
Old 07-03-2011, 08:00 PM   #137 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Quote:
Originally Posted by endgame View Post
To the best of my recollection- and I'll look into it to support with documented quotes if they're available -the BA rankings and many of the in-game reports are OSA-based.
My theory from looking at the OOTP manual is that it is OSA scouts that do top prospects reports. However, the manual states these can change daily. The manual also states the more times a player is scouted by a scout the more accurate the information is. Therefore if the OSA scout "assigned" to prospects evaluates the same guys daily the information will eventually be very accurate. So accurate that there is no difference than if they saw the true ratings. 100% scouting accuracy means they are seeing the true ratings. That does not mean there is 0 blur in the model just that continual repeated scouting makes that blur close to 0. So after so many reports on the same guys the scouts become something like 99.99999% accurate.
Biggio509 is offline   Reply With Quote
Old 07-03-2011, 08:07 PM   #138 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Quote:
Originally Posted by injury log View Post
If, say, in the amateur draft my scout thinks two players are potential Hall of Fame talents, and one requests a $8m signing bonus and the other is willing to sign for Slot, I'll pick the $8m guy every time and dismiss the other scouting report as hyperbole.
NB: the following statement is pure opinion. I have not gathered the data so this is a hunch. With that out of the way, I believe there seems to be little correlation with salary demands and true ability. I have accidentally signed a few guys in the late rounds that the scout said were scrubs but asking 1.5 to 3 million in bonus. I have never found those guys to be any better than the slot picks in similar rounds. Again I have not meticulously collected data on this and I may have missed some late round surprises who actually asked for a bigger signing bonus.

However, if I am not correct and signing bonus actually reflects the quality of players then the AI in no way is using real ratings in the draft. You would not find guys asking for anything above slot in round 25 or 30. If the bonus were reflective or true ability and the AI knew the real ratings all of these guys would go before guys who only demand slot. I think the bonus is more tied to player greed ratings rather than ability. Again I can't prove that but it is what I believe.
Biggio509 is offline   Reply With Quote
Old 07-03-2011, 08:12 PM   #139 (permalink)
Hall Of Famer
 
endgame's Avatar
 
Join Date: Aug 2002
Posts: 12,831
Thanks: 1,695
Thanked 2,016x in 1,055 posts
Quote:
Originally Posted by Biggio509 View Post
My theory from looking at the OOTP manual is that it is OSA scouts that do top prospects reports. However, the manual states these can change daily. The manual also states the more times a player is scouted by a scout the more accurate the information is. Therefore if the OSA scout "assigned" to prospects evaluates the same guys daily the information will eventually be very accurate. So accurate that there is no difference than if they saw the true ratings. 100% scouting accuracy means they are seeing the true ratings. That does not mean there is 0 blur in the model just that continual repeated scouting makes that blur close to 0. So after so many reports on the same guys the scouts become something like 99.99999% accurate.
Well? In theory, you have a compelling case. Here's the fly in the ointment. OSA doesn't have a scout or scouting director that gains experience like ours have the potential to do. For all we "know", OSA is composed of a variety of news sources and scouting information that in fact does change daily, but not in such a predictable and positive way; sometimes the information on a particular player may get worse or less accurate. Overall? Who knows? Individually? Again, a function of multiple settings and sources from scouting accuracy to which box is checked to which profile page you're viewing to ... ad infinitum. Scouting's a maybe. Always, even everything is set high and Legendary, it isn't improved that much over Decent and Good. With time sure, but those are people, not OSA.
__________________
"Try again. Fail again. Fail better." -- Samuel Beckett
_____________________________________________
endgame is offline   Reply With Quote
Old 07-03-2011, 08:14 PM   #140 (permalink)
All Star Starter
 
Biggio509's Avatar
 
Join Date: Aug 2007
Posts: 1,620
Thanks: 41
Thanked 545x in 302 posts
Very true. Only Markus knows for sure and he might have to look back at the code to remember how it is set up.
Biggio509 is offline   Reply With Quote
Reply

Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT -4. The time now is 07:03 AM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2014, vBulletin Solutions, Inc.
Search Engine Friendly URLs by vBSEO 3.6.1
Copyright © 2013 Out of the Park Developments