Player Evaluation: Fielding Broken?

redmarkYankees · 06-18-2006, 10:37 AM

Is this a (partial) cause of dodgy trades and roster decisions?

In the SQL dump is an interesting table called Player Value. Playing around with this data, something becomes obvious very quickly: fielding evaluation is broken - all players have a value of zero - the question is, is this in-game, or just in the SQL dump?

If this is indeed broken in-game, this suggests a number of knock-on effects:

- Travis Hafner and Derrek Lee are more valuable than Alex Rodriguez and Vladimir Guerrero. Basically, first baseman are overrated.
- Defensive players are totally undervalued. Explains odd gold glove awards?
- All pitchers are overrated vs all hitters, as hitters defensive value is not taken into account.
- Trading AI is dealing with inaccurate evaluations, then further affected by GM tendencies and scouting skills... no wonder some trades look strange.

A follow up question... I'm assuming that in a trade evaluation, the AI looks at Overall Value, Talent Value and Career Value - then adjusts the relevant weight of these factors based on GM tendencies and game settings (favour veterans, favour prospects, etc). Is this broadly correct, without giving away any trade secrets?

redmarkYankees · 06-18-2006, 03:43 PM

Just bumping this as it might be a potential biggie (and big win to fix, if it effects everything else like trading). Certainly bigger than teams conceding 6 HRs in a game...

redmarkYankees · 06-18-2006, 04:24 PM

Found another issue with the Player Value SQL dump, that certainly seems to be a problem if the SQL is accurately portraying what's in the game (in this case, possibly why MRs seem to be overvalued in trades):

Amongst other ratings, there are Pitching Value, Pitching Value VSR and Pitching Value VSL. Then, Overall Value - and Overall Value VSL and Overall Value VSR. For MRs, Overall Value is generally significantly less than Pitching Value, while for SPs it isn't (i.e. the game is reasonably reducing overall value for a reliever, as their contribution is limited in scope).

However, Overall Value VSL and Overall Value VSR are NOT reduced in line with 'overall' Overall Value; they are the same as Pitching VSL/R. IF the trading AI considers bullpen needs specifically VSL/VSR, this could significantly be overestimating a players Overall Value, according to the Overall Value itself.

Raidergoo · 06-18-2006, 04:52 PM

Neat.

I am going to make a TT. It is TT#1889.

Please create a second, simple fictional league, sim for a year or two, run a data dump, and confirm this error.

This may prove to be an interesting bit of data.

redmarkYankees · 06-18-2006, 05:21 PM

Done; created a default fictional league, created dumps prior to seasons 1, 2 and 3. Also created csv to check against SQL (no difference).

Confirm:

1. Fielding_Value0 to Fielding_Value9 (all positions, presumably) are zero for all players in all dumps.

2. For MRs/CLs, the overall_value, talent_value and career_value are significantly reduced from the pitching_value. However, overall_value_vsl and overall_value_vsr are exactly the same as pitching_value_vsl and pitching_value_vsr.

Raidergoo · 06-18-2006, 10:55 PM

Thanks for the extra effort on this item, redmarkYankees.

I don't know what the net result of this is. I don't know if this is due to a bug or a design choice.

I do know this much, it is flagged in the Test Track and it seems to me to be an interesting item. Therefore, I feel that I have done my best on this issue.

I may be wrong, as I often am.

redmarkYankees · 06-19-2006, 03:13 AM

Thanks for logging it, Raidergoo, I don't know either exactly what effect this has or if the bug (if there is one) is ingame or in the SQL dump. It looked worth asking about and that it might be a contributory factor in some odd roster/trade decisions being discussed, which I thought appeared might be due to a common denominator in a player evaluation issue rather than in the trade AI or roster movement AI, for example.

dougaiton · 06-19-2006, 06:15 AM

Great work, RMY!

Hammer755 · 06-19-2006, 09:32 AM

Quote:

Originally Posted by dougaiton

Great work, RMY!

"" !

redmarkYankees · 06-21-2006, 07:08 AM

A gentle bump to see if anyone is looking at this.

It occurred to me that if there is a minor issue with young players squeezing out average older players too early, this might also be a contributory factor in that. If the career_value is excluding fielding_value, then the career_value is lower than it should be; for all players, but with a particular relative impact on the older players who aren't offensive superstars - hence HOF type players may play until they are 40+, but journeymen hitters/fielders may be being particularly impacted by this issue.

Raidergoo · 06-21-2006, 07:27 AM

Perhaps I should try to explain my MO in a friendly fashion.

Once it is in TT and has a #, bumping does little good without new data.

The bug is in my name, so when I log in at the Test Track, I see it and others staring back at me. The first line of each bug in this list (for most of my entries, at least) have the URL, the initial post, the TT name ( usually the thread name ) and any additional commentary. When people have valuable additions to make, I copy/paste into the existing ticket.

This way, the red flag gets raised for the developers, who don't have all the time in the world.

This is why questions "are you patched?" and "can you get a screenshot" are repeatedly asked. If a bug can be quickly and clearly identified in a thread, it makes it that much easier to eliminate.

redmarkYankees · 06-21-2006, 07:34 AM

Fair enough, though the gentle bump was more an opportunity to post thoughts on a possible link with possible issues regarding possible uneven age distribution posted elsewhere.

I don't entirely get the relevance of the rest of your post, as I haven't been asked for screenshots, or whether I've patched (I have), or any more data. Which is a shame in a way, as I'm quite interested in the outcome so if some busy tester or developer wanted me to run some further tests or something, I could. (Like, I could run an initial draft, do a SQL dump, and check the draft order against the player evaluations or something).

Regardless, I won't bump again. Unless something else possibly relevant occurs to me.

dougaiton · 06-21-2006, 07:54 AM

Indeed. RMY was to indicating the new dimensions to the problem. Had someone read the thread previously, they might not have made the logical jump.

Joshv02 · 06-21-2006, 09:00 AM

This is a relatively big deal, it appears.

Quote:

Originally Posted by Raidergoo

TT and has a #,

Would you mind giving us the TT# so that when the next patch comes out (presumably with a list of TT#s fixed) we can see if this is one of them?

tia.

Raidergoo · 06-21-2006, 09:14 AM

Quote:

Originally Posted by redmarkYankees

Fair enough, though the gentle bump was more an opportunity to post thoughts on a possible link with possible issues regarding possible uneven age distribution posted elsewhere.

I don't entirely get the relevance of the rest of your post, as I haven't been asked for screenshots, or whether I've patched (I have), or any more data. Which is a shame in a way, as I'm quite interested in the outcome so if some busy tester or developer wanted me to run some further tests or something, I could. (Like, I could run an initial draft, do a SQL dump, and check the draft order against the player evaluations or something).

Regardless, I won't bump again. Unless something else possibly relevant occurs to me.

Discussion is encouraged. I don't know the relevance of this issue; I think it is important enough to flag at the highest level possible, partly because I don't understand the ramifications. It leaves me scratching my head and a slightly sick feeling. If discussion is what you want, you'd better bump early and often. If it is the attention of the developers you want, well, you have it with the TT.

The rest of my post was informational in nature, trying to explain what I do. It is background material, it can be processed and filed away. I figure the more people know, the better.

Raidergoo · 06-21-2006, 09:15 AM

Quote:

Originally Posted by Joshv02

This is a relatively big deal, it appears.
Would you mind giving us the TT# so that when the next patch comes out (presumably with a list of TT#s fixed) we can see if this is one of them?

tia.