A Look into Draft Sizes

[Kett]

I have spent the last few days look at the number of draft rounds and how they affect the distribution in talent in leagues. I remember someone doing this kind of analysis for ver 4 but I figured, a completly new version must bread completely different results. And as it turns outs, it is different, as the previous one recomended 3 rounds.

I did this analysis, using the default league settings and the default league file. As a measure of ability I used OPS as my stat of choice. Why? Because its easier to calculate than runs created, and besides runs created its the best stat to rate an individuals performance.

I simmed 25 seasons for various rounds of drafting. To analyze, I took the players with the top 200 plate appearance totals for the last season simmed, breaking down OPS in .05 intervals to split up all the players. I then compared that season with the breakdown for the actual 2003 MLB Season.

My Results: I found that the best number of rounds is 5. For round amounts less than that, there is not enough talent to spread out over the entire league, so the poor starters are poorer than those of the Major Leagues. Also, because the talent level is lower overall, the really good players look even better, so the league is top heavy if you will. As for round amounts greater than 5, the talent starts to become too good for league (this is all in comparison to the 2003 season, so the results may desired for a fictional league). What ends up happening, is the really good players become just good, with the top players numbers decreasing relative to the league. And overall there are just more good players hitting at or near the average OPS.

Actually, I just looked at the distribution for 1965, and it is different from the current year. In 1965, from looking at the numbers, the league did not have as much talent, so the proper number of rounds is more like 3 rounds. And for 1930 it looks like it should be about 2 rounds.

Well thats about it, just though I'd share that with the OOTP world, contribute my little bit, to such a great game.

[Dan Theman]

I've been running a 25-player draft in my multi-player league (28 teams, two leagues actually) and after 108 games, the current major league leader in hitting is batting .381 with a major league average of .270, while the major league leader in ERA is at 1.65 when the major league average is 4.81 (both leagues use the DH, by the way).

I think that's a pretty big difference. By the way, getting the league totals (and creation modifiers after the conversion to v.6) to get these kinds of results was NOT easy, but I think it's worth it and most importantly, so do my owners. They love the "realism" of a long draft. Well, some do - some could really do without, but they're okay either way.

I hope this was of some help.

- Dan

[Matter2003]

Dan, what modifiers are you using to get the proper mix? I would love to be able to have a longer draft....

Matter2003

[Dan Theman]

Unfortunately, the creation modifiers won't be of any use because your league-wide ratings heavily depend upon what vesion you started your league. I'd recommend you first decide how many rounds you want, get averages and distribution threads for your leagues current ratings, then run a test league through at least 17 seasons (the length of time to completely "even out" after the initial creation players are gone) with the desired number of rounds and check a) actual league totals and b) distribution and averages of ratings. The easiest way to do this is by exporting the players to a CSV file and using Excel. Then make creation adjustments to a new test league, do it again, and see what you get. It's all trial and error, unfortunately. Once you get the right blend, you'll be able to seamlessly adapt your old league to the longer draft.

My league totals - which are also relative to the ratings, but perhaps a bit more useful - are as follows:

AB:166234
H:45000
2B:9750
3B:928
HR:5458
BB:15806
HBP:1890
SO:29750
ABIP: .302

- Dan

[Henry]

Dan,

Great study I'm wondering if the recommended number of rounds would thus be based on some equation that has number of teams and Era as the primary variables. As you said, maybe the Deadball era with 16 teams would only have 2 rounds while a 1965 league with 20 teams would have 5. If the basic theory is that more teams and later history requires more rounds, then some equation should be possible.

[reslfc]

Interesting. I wonder if it has anything to do with the number of rookies coming into the majors each year. There were many more players who made their debuts in the last few seasons than in 1965 or 1930. Obviously the increased number of teams is one reason, but the debut numbers jumped up before the '98 expansion, around '95 or '96. I looked at this a while ago because I was going to import historical players as free agents every year into a league of mine instead of holding the amateur draft with fictional players.
http://slfc.netfirms.com/YearlyDebutNumbers.htm
and a chart of that data is at http://slfc.netfirms.com/DebutYearsChart.png

[Kett]

Originally Posted by Henry

Dan,

Great study I'm wondering if the recommended number of rounds would thus be based on some equation that has number of teams and Era as the primary variables. As you said, maybe the Deadball era with 16 teams would only have 2 rounds while a 1965 league with 20 teams would have 5. If the basic theory is that more teams and later history requires more rounds, then some equation should be possible.

I'll look more into it later, as after simming over 200 years, I'm kinda tired of it, but I'm not sure whether it is a function of league size, or just a function of changing times, and more people in the world and what not. Come to think of it, I did all the tests on a full current structure league, so I would assume the differences would be more a function of era and less of league structure, but I'll look into that. That will be a good thing to test next.

[Henry]

Kett,

DOH... sorry for the "Dan" response. I sometimes run through these things too fast

Yes, please let us know if a formula can be created. It would be a nice addition to the OOTP "library" !!

[Dan Theman]

LOL - hey, it's always nice to get credit for work I haven't done

Good luck, Kett - if you need any help, let us know.

- Dan

[Matter2003]

Dan, this is my first year at OOTP, so I started my league with version 6.03...

I have a great engine setup for Halofan's roster set that I am using, but I was distressed because they were markedly different from when I tested with a default league.

Basically I tuned the engine to give almost perfect stats comparble to the 2003 MLB season with a default league, but when I switched to the Halofan rosters and used these settings, the stats were way off...the batting average dropped from .264 to about .248 and all offensive numbers were down bigtime, creating a pitcher dominated game....

I then had to "re-test" with his rosters and get correct settings, but am now worried that when new computer generated players are created, the league will become pitching dominated again after 4 to 5 years...

Any suggestions Dan? Will recalculating based on last years stats work to alleviate this problem or do I need to actually do more work than this....

This is why I suggested being able to, in addition to recalculating based on era, being able to recalculate based on a year....all the information is already contained in the .dat file, and it would able to recalculate settings based on whatever year's stats you wanted to use...Howver this suggestion was met pretty harshly by many others...

Any suggestions would be helpful...my settings for Halofan's rosters that give very good results are:

AB 166566
H 44090
2B 9175
3B 1255
HR 5135
BB 16150
HBP 1672
K 36050
ABIP .297

I also have been thinking that instead of adjusting the settings, the game should adjust the RATINGS based on the settings you put in to maintain stability...

Matter2003

[Dan Theman]

Matter2003 -

I use fictional rosters, so my experience with Halofan's rosters is a bit limited. Most modern roster sets are balanced to a 10-round draft because it's the default, but that doesn't mean they're balanced perfectly. I'd still recommend the trial-and-error approach if you're aiming for consistent careers and not just consistent league totals. It takes awhile, but I think it's worth it in the long run. Because you have a league that's already been tested for consistency, you should be able to get it done relatively quickly - set up the trial league with your exact settings (the easiest way is to copy your league folder into an empty folder, rename it, and then copy it back), let it run overnight, and then take a look at what you've got. Hopefully all will be well

As far as the idea to adjust ratings, while appealing on one level, it's a bit of a Pandora's Box. First you get into the idea of "maximum" and "minimum" performance, going against the philosophical grain of most baseball fans, and then you get into the even heavier issue of player development, etc.. Still, the benefits of that kind of a system would certainly make for an easier time in many ways. Someday I hope we can marry the two.

- Dan

[The Interloper]

Somebody posted something in Talk Sports not long ago that showed all kinds of technical stuff about real baseball history. Maybe you could compare that to your leagues and see what's most accurate. It had standard deviations and all that.

[ctorg]

Originally Posted by The Interloper

Somebody posted something in Talk Sports not long ago that showed all kinds of technical stuff about real baseball history. Maybe you could compare that to your leagues and see what's most accurate. It had standard deviations and all that.

That was me, if it's what I think you're talking about, and it was only for batting averages.