Trading AI Tweak Ideas

TwinMonster · 03-07-2003, 01:09 PM

Trading AI Tweak Ideas
First, it is my hope that we can stick to ideas ONLY in this thread. Since we have another tread where we've debated the issue (actually, a couple I think) perhaps it would be good to have one where we post ideas for tweaking the trading AI. My present ideas probably would be rather time consuming to code. Perhaps they won't happen for v.5 and that's cool.

I was contemplating a previous post mentioning teams rebuilding, preferring veterans, etc when trading. I took this to the next level so to speak. Why not have team building ratings for GMs. There are talent evaluation ratings which is great. But why not add preference for veterans, young players, etc. Or perhaps a front office rating. Teams go through a cycle of rebuilding, veterans again, rebuilding, etc. Cleveland has hit that point. The Yankees never seem to. But than they are free spenders. So, then there would need to be a rating for teams that just keep adding old farts like Mondesi (and we'll leave that one right there.) Perhaps it would be better to have owner ratings than. In truth, this could affect more than just trading. It could get into what prospects get to play and when. Maybe some of this is already in the game. The point is, it would be great to have some idea of what a team is looking for in a trade. Do they want veterans, rookies, prospects? Believe me, I wouldn't want to have to code it, but if it could be done it would give the game even more depth.

Anyone else with ideas? Mine are still raw and I'll keep thinking.

Joshv02 · 03-07-2003, 03:17 PM

Wonderful thread to start!

It may be important to build in different ‘personalities’ for team building. This would have to be nuanced, as it’s not clear what should be valued most in building a team. At the least, there is disagreement. Even among sabermatricians, there is significant disagreement. (E.g., look at differing recommendations for team building in Colorado; or differing emphases on pitcher value in avoiding hits.)

A ‘perfect’ AI system with this in mind would also have various manners of evaluating trades. This would not merely be smart-to-stupid continuum. There would have to be different value systems in place for different teams/GMs. You may want a smart-to-stupid continuum, too.

Different personalities should be different from merely being different places on BasballProspectus’s idea of the cycle. Some GMs LOVE vets. This may not be stupid from a purely baseball point of view (from an economic point of view, it’s a little less smart – but there can be market corrections for that). Vets provide more information and therefore may provide -- on the whole -- less overall risk. Or, at least, the risk can be better quantified. This may be a reason why GMs are willing to ‘overpay’ for someone else’s reclamation project (e.g., Matt Stairs or Mike Fetters). Some teams may be more risk adverse for very decent reasons. Other teams may be more interested in continuity. Others in youth for youth’s sake. Still others build a team for profitability’s sake. (How to attain profitability –again- may differ from city to city. Compare profitability in Seattle vs. Minnesota.)

Then, there is the question of what talent to go after. Drafting young, high ceiling talented pitchers has worked for Atlanta for a decade (although, this is debateable). Drafting college tested, older pitchers with high SO rates has worked for Oakland for the past half decade. Drafting toolsy hitters hasn’t worked for Pittsburgh, but it may for New York. I love walks and power with sound defense. Others may love speed and average. Others may love power and walks, but care little for defense. I usually look for high SO rates in pitchers. Others for low HR rates. Others for low hit rates. What a GM is looking for in a player should matter (although, probably not be determinative).

Lastly, I’d think it would matter what position a team is in. Some teams should be fleeced in desperate situations (e.g., July 29th the # clear 1 SP goes down on the Yankees). Some teams should be so risk averse to never want to trade unless they clearly “win.” Others may compare Win Shares from past years to “pre”-determine if a trade is valuable. Others may really care about today (e.g., Larry Anderson for Jeff Bagwell). Others may have to pair salary. Others may have to get people into the stands (e.g., Juan Gonzolez in Det), regardless of the improvement.

A truly “logical” trade AI should have significant variation in what the baseline is for ‘logical.’ I don’t think there should be a clear right answer.

HA—how to code that? Yeah right… it’s hard enough to write it in English.

Stephen19 · 03-07-2003, 05:01 PM

I think the previous two posts made alot of very valid and intelligent suggestions for how to tweak the trading AI.
That said I believe it's something that will never be done to the satisfaction of everyone who plays the game.
It's not possible. Too many variables. Too many different ideas of what is good and what sucks.
But, I believe there is room for everyone to make suggestions on what they'd like to see changed for the next patch or the next game itself.
That way whatever seems most popular as a better chance of being added.
Since the patch came out I've found the trading to be vastly improved.
Moreso from the offers I've received than when I propose a trade that is.
Before I never got offered a trade that was reasonable. Now I get a couple that I think are fair and really that's all I would want.
When it comes proposing a trade I find that a little harder to complete. Maybe it's just me but when I propose something it's almost always rejected and when they make a counter-offer they tend to ask for my best hitter or best pitcher 90% of the time. Regardless of who I'm asking for.
I haven't done any studies on this and this just my first impression, but I'd like more variety when proposing a trade.
I'd also like the ability to counter-offer trade proposal's the computer gives me.
All in all I can live with things the way they are, but I think any improvement is welcome.

03-07-2003, 01:09 PM	#1
TwinMonster Major Leagues Join Date: Jan 2003 Location: Minnepolis Posts: 325	Trading AI Tweak Ideas Trading AI Tweak Ideas First, it is my hope that we can stick to ideas ONLY in this thread. Since we have another tread where we've debated the issue (actually, a couple I think) perhaps it would be good to have one where we post ideas for tweaking the trading AI. My present ideas probably would be rather time consuming to code. Perhaps they won't happen for v.5 and that's cool. I was contemplating a previous post mentioning teams rebuilding, preferring veterans, etc when trading. I took this to the next level so to speak. Why not have team building ratings for GMs. There are talent evaluation ratings which is great. But why not add preference for veterans, young players, etc. Or perhaps a front office rating. Teams go through a cycle of rebuilding, veterans again, rebuilding, etc. Cleveland has hit that point. The Yankees never seem to. But than they are free spenders. So, then there would need to be a rating for teams that just keep adding old farts like Mondesi (and we'll leave that one right there.) Perhaps it would be better to have owner ratings than. In truth, this could affect more than just trading. It could get into what prospects get to play and when. Maybe some of this is already in the game. The point is, it would be great to have some idea of what a team is looking for in a trade. Do they want veterans, rookies, prospects? Believe me, I wouldn't want to have to code it, but if it could be done it would give the game even more depth. Anyone else with ideas? Mine are still raw and I'll keep thinking.

03-07-2003, 03:17 PM	#2
Joshv02 Hall Of Famer Join Date: Feb 2003 Location: , " Posts: 3,082	Variability Wonderful thread to start! It may be important to build in different ‘personalities’ for team building. This would have to be nuanced, as it’s not clear what should be valued most in building a team. At the least, there is disagreement. Even among sabermatricians, there is significant disagreement. (E.g., look at differing recommendations for team building in Colorado; or differing emphases on pitcher value in avoiding hits.) A ‘perfect’ AI system with this in mind would also have various manners of evaluating trades. This would not merely be smart-to-stupid continuum. There would have to be different value systems in place for different teams/GMs. You may want a smart-to-stupid continuum, too. Different personalities should be different from merely being different places on BasballProspectus’s idea of the cycle. Some GMs LOVE vets. This may not be stupid from a purely baseball point of view (from an economic point of view, it’s a little less smart – but there can be market corrections for that). Vets provide more information and therefore may provide -- on the whole -- less overall risk. Or, at least, the risk can be better quantified. This may be a reason why GMs are willing to ‘overpay’ for someone else’s reclamation project (e.g., Matt Stairs or Mike Fetters). Some teams may be more risk adverse for very decent reasons. Other teams may be more interested in continuity. Others in youth for youth’s sake. Still others build a team for profitability’s sake. (How to attain profitability –again- may differ from city to city. Compare profitability in Seattle vs. Minnesota.) Then, there is the question of what talent to go after. Drafting young, high ceiling talented pitchers has worked for Atlanta for a decade (although, this is debateable). Drafting college tested, older pitchers with high SO rates has worked for Oakland for the past half decade. Drafting toolsy hitters hasn’t worked for Pittsburgh, but it may for New York. I love walks and power with sound defense. Others may love speed and average. Others may love power and walks, but care little for defense. I usually look for high SO rates in pitchers. Others for low HR rates. Others for low hit rates. What a GM is looking for in a player should matter (although, probably not be determinative). Lastly, I’d think it would matter what position a team is in. Some teams should be fleeced in desperate situations (e.g., July 29th the # clear 1 SP goes down on the Yankees). Some teams should be so risk averse to never want to trade unless they clearly “win.” Others may compare Win Shares from past years to “pre”-determine if a trade is valuable. Others may really care about today (e.g., Larry Anderson for Jeff Bagwell). Others may have to pair salary. Others may have to get people into the stands (e.g., Juan Gonzolez in Det), regardless of the improvement. A truly “logical” trade AI should have significant variation in what the baseline is for ‘logical.’ I don’t think there should be a clear right answer. HA—how to code that? Yeah right… it’s hard enough to write it in English.

03-07-2003, 05:01 PM	#3
Stephen19 Minors (Single A) Join Date: Jul 2002 Posts: 50	I think the previous two posts made alot of very valid and intelligent suggestions for how to tweak the trading AI. That said I believe it's something that will never be done to the satisfaction of everyone who plays the game. It's not possible. Too many variables. Too many different ideas of what is good and what sucks. But, I believe there is room for everyone to make suggestions on what they'd like to see changed for the next patch or the next game itself. That way whatever seems most popular as a better chance of being added. Since the patch came out I've found the trading to be vastly improved. Moreso from the offers I've received than when I propose a trade that is. Before I never got offered a trade that was reasonable. Now I get a couple that I think are fair and really that's all I would want. When it comes proposing a trade I find that a little harder to complete. Maybe it's just me but when I propose something it's almost always rejected and when they make a counter-offer they tend to ask for my best hitter or best pitcher 90% of the time. Regardless of who I'm asking for. I haven't done any studies on this and this just my first impression, but I'd like more variety when proposing a trade. I'd also like the ability to counter-offer trade proposal's the computer gives me. All in all I can live with things the way they are, but I think any improvement is welcome.