Predicting Player Performance-It's not all about the ratings

[Spiccoli]

Or is it?

Ok, I've still very new to this game, still under 1 year. But it's got it's hooks in me... bad.

One of the aspects of OOTP that I love and sometimes baffles me is how players don't always perform per their ratings... Or at least how I expect them too. Therefore, after the seasons starts, I sometimes go with player performance instead of their ratings when setting lineups. It's worked well for hitting so far, though I'm still to trying solve the pitching staff puzzle. Y'all have such varied approaches to setting up bullpens... it's super cool actually.

Anyway... the point of my post. I don't complete understand why Ray Cooper hits so well... and Jayden Harsett doesn't.

Here's their ratings (NOTE - I know that the scouting rating aren't always accurate, but it's what we have)

Ray Cooper - L
(Overall/Vs Left/Vs Right)
Contact 7/6/8
Gap 7/6/7
Power 8/6/8
Eye 6/5/7
K's 7/7/8
Speed - 4

Jayden Harsett - L
(Overall/Vs Left/Vs Right)
Contact 7/6/7
Gap 7/7/8
Power 7/6/7
Eye 9/8/9
K's 5/5/6
Speed - 10

Seriously, their ratings aren't that much different and Hayden has a much better eye. Here's their stats for 2033

Cooper - .309/.394/.559/.953 36 HR's 84 BB's (608 ABs)
Harsett - .214/.324/.381/.711 11 HR's 64 BB's (373 ABs) -> Approx Extrapolate - 20 HR's / 120 BB's

As you can see, there's a significant difference in their performance. It seems to me based on rating, Harsett should be a much better hitter than he is. If you had no past batting history for either player, would Cooper's ratings jump out to you as being that much better than Harsett's? If yes, why? (Note - Harsett has had similar performance the last few seasons, not just last year)

I also see the same thing in pitchers as well. For example, why has Tavio Ciccolella pitched so well over the last two seasons? His stuff is only rated at 6 and movement at 7 (6/7/9)... and he throws 88-90 mph.

I have pitchers on my staff rated better over-all who can't touch him. For example, Jessie Stewart is rated 7/7/8, throws 96MPH and is a career 4.20-4.50 pitcher with a 45-56 career record.

I just thought I'd throw this out for discussion. I'm definitely not complaining about it, in fact, I think it's great that there's this ambiguity between ratings and actual performance. The game would less interesting if the stats directly correlated to ratings.

However, I may also be missing something when I'm evaluating players. In fact, it's highly probable. Maybe it's because of the rounding that a 7 rating is actually everything from a 6.6 to 7.5?

Anyway, I'd love to y'alls opinion on the matter.

Thanks,
Scott

[RonCo]

This is a huge subject. I (and many others) have spent hundreds of thousands of words on it.

All I'll comment on for now is that yes, not all "7"s are equal. Nor are all "6"s, nor all ratings of any kind except things like velocity. In our scale each numerical rating spans 20 or so different performance levels (core performance ratings go 1-200 in general, with a 50 point space above that for superstars). There can be a huge difference between a high 6 and a low 6, for example. And those differences then get further mushed by differences in opponents, or ballparks or defenses or injury impacts or purely the facts of the random nature of mathematical randomness.

[essentially, using your comment, it's really that a 7 is anything between a 7.0 and a 7.95]

This is among the reasons why I say that you really don't need to play with scouts as long as you use 1-8 or 1-10 (or 1-5, whatever) ratings systems. The more coarse the rating system, the more you're really playing a stats only kind of league.

OOTP's results engine is a monstrous thing of beauty, really. And I always find it's ratings/performances are fun to attempt to read and watch.

[Spiccoli]

This is a huge subject. I (and many others) have spent hundreds of thousands of words on it.

All I'll comment on for now is that yes, not all "7"s are equal. Nor are all "6"s, nor all ratings of any kind except things like velocity. In our scale each numerical rating spans 20 or so different performance levels (core performance ratings go 1-200 in general, with a 50 point space above that for superstars). There can be a huge difference between a high 6 and a low 6, for example. And those differences then get further mushed by differences in opponents, or ballparks or defenses or injury impacts or purely the facts of the random nature of mathematical randomness.

[essentially, using your comment, it's really that a 7 is anything between a 7.0 and a 7.95]

This is among the reasons why I say that you really don't need to play with scouts as long as you use 1-8 or 1-10 (or 1-5, whatever) ratings systems. The more coarse the rating system, the more you're really playing a stats only kind of league.

OOTP's results engine is a monstrous thing of beauty, really. And I always find it's ratings/performances are fun to attempt to read and watch.

So, the main theory is that the discrepancy in ratings vs performance is due to the rounding of the actual ratings vs rating system used?

I can understand that, but I'd also like to think (or hope) that there's possibly more to it, maybe hidden (non-hidden) intangibles?

Or not...

[RonCo]

There is much, much more to it. I was only responding to one of your comments.

[bschr682]

Also sometimes players just over/under perform for awhile. Sometimes other team factors can influence stats pretty heavily. A good groundball pitcher on a team with stout infield defense is awesome. Sometimes the park factors play a major role too. Sometimes its just plain random. I can definitively tell you that the only hard truth to it is there is no hard truth. There are a ton of moving parts at play...

[Spiccoli]

Also sometimes players just over/under perform for awhile. Sometimes other team factors can influence stats pretty heavily. A good groundball pitcher on a team with stout infield defense is awesome. Sometimes the park factors play a major role too. Sometimes its just plain random. I can definitively tell you that the only hard truth to it is there is no hard truth. There are a ton of moving parts at play...

Yeah... I mean, look at Ray Cooper’s ratings, 7/7/8/6/7. It’s good, but would you have guessed 3rd place for MVP good? Now, you’re probably right and he should come down off of that this season.

It just baffles me the Harsett is such sub-par hitter compared to Cooper, with his 7/7/7/9/9 ratings. Combined his base running and defense, Harsett should an All Star contender, but the owner is complaining and wants me to replace him.... lol.

Knowing the baseball gods, he’d end up in Yellow Spring or Madison hitting .400 against us and stealing 6 bases per game.

[udlb58]

And then there are some guys that nobody can explain, like Jorge Gonzales. He's been bouncing between 5/6 contact vRHP, with no other rating over 5 either, yet he just keeps hitting .270+ somehow. Or Jose Gamboa, who just keeps hitting .300 with an .800+ OPS every year. Or Jake Blues, who couldn't hit the ground if he fell down despite 7/7/7/5/5 ratings.

[RonCo]

Harsett's AvK rating is lower than you're posting. That's could well be a major source of difference. Harsett strikes out a lot.

[Ted]

The other thing that makes us misinterpret the ratings system is plain old human inability to deal with scaling. While the ratings system goes from 1-13, the actual functional ratings at the big league level are more upper 5 to 13. Furthermore, when you consider the narrow range from upper 6 to mid 8 that most pitchers operate in, you are now looking at a batter's combination's of 6's and 7's in a totally different light. That seven, represents a 20 point spread on a 1-250 scale. Let's look at contact alone. If we say 5 contact is the most marginal thing you can get by with, and 8 is above average, you then have a 60 point window most players are are operating in. OOTP seems to give diminishing returns once you get so much better or so much worse than league average ability. A 13 contact isn't twice as good as a 7 contact. A 2 isn't significantly worse than a 4 at the major league level, because both are too bad to play there. 12 is not as much better than 10 than 8 is compared to 6. It has to work this way because the players ratings are competing with what a probability model says should happen, and any good baseball engine won't let players hit .800, or strike out 22 batters every time out.

So when you look at 6's and 7's that way, then realize the player has five of those ratings, and that pitchers have three (or more if you want to go pitch by pitch) you see that the overwhelming bulk of the difference between player skill at any given level is mathematically derived to matter in a much narrower range than 1-13 (or 1-250).

When you combine this with Brett's addition about plain old randomness, you probably get most of the results. There are internal clutch, and handle adversity/handle success/etc, and other ratings. It is hard to tell how much they matter.

If anyone can do a better job restating that, please do. I'm not very good at explaining sometimes.

[Spiccoli]

Harsett's AvK rating is lower than you're posting. That's could well be a major source of difference. Hasnett strikes out a lot.

Ok. I fixed the OP.

[RonCo]

A few other factors.

- They are LHB, and Cooper is a stronger contact guy against RHP.
- Harsett appears to have bigger splits, making him even more of a platoon candidate than Cooper.
- The skew of pitcher stuff ratings in the league, and the order of the results engine code almost certainly combine to increase the influence of the AvK difference.

[RonCo]

The other thing that makes us misinterpret the ratings system is plain old human inability to deal with scaling. While the ratings system goes from 1-13, the actual functional ratings at the big league level are more upper 5 to 13. Furthermore, when you consider the narrow range from upper 6 to mid 8 that most pitchers operate in, you are now looking at a batter's combination's of 6's and 7's in a totally different light. That seven, represents a 20 point spread on a 1-250 scale. Let's look at contact alone. If we say 5 contact is the most marginal thing you can get by with, and 8 is above average, you then have a 60 point window most players are are operating in. OOTP seems to give diminishing returns once you get so much better or so much worse than league average ability. A 13 contact isn't twice as good as a 7 contact. A 2 isn't significantly worse than a 4 at the major league level, because both are too bad to play there. 12 is not as much better than 10 than 8 is compared to 6. It has to work this way because the players ratings are competing with what a probability model says should happen, and any good baseball engine won't let players hit .800, or strike out 22 batters every time out.

So when you look at 6's and 7's that way, then realize the player has five of those ratings, and that pitchers have three (or more if you want to go pitch by pitch) you see that the overwhelming bulk of the difference between player skill at any given level is mathematically derived to matter in a much narrower range than 1-13 (or 1-250).

When you combine this with Brett's addition about plain old randomness, you probably get most of the results. There are internal clutch, and handle adversity/handle success/etc, and other ratings. It is hard to tell how much they matter.

If anyone can do a better job restating that, please do. I'm not very good at explaining sometimes.

This is where it gets into really deep game engine neep.

One correction: a 5-8 range is an 80 point span, not 60. Note also, that "Contact" is not a meaningful rating to the game engine. It's a calculated construct that represents an expected result if all ratings in a league are with certain boundaries. The actual controller of batting average is a series of calculations that rely on a pitcher's movement and stuff ratings and a hitter's Power, AvK, and hidden BABIP rating (not to mention ballpark factors and defense). Since our ratings are skewed (like the ratings of most leagues are), this means that the "Contact" rating the game calculates and shows us is likely skewed one way or the other, also.

To really get my mind around this, it helped me to pull out the big-ole Bill James Log5 equations and play with order of calculations for a day or so.

---

Regarding hidden internal ratings. I duuno. I'm not convinced "clutch" exists, but I can go with some kind of "big season" slump season thing. Or not. The laws of randomization can still account for a lot of that, though. Which still makes it fun.

---

You can use the game's editor on a test leagues to get an idea of performances around the rating scale of each rating. Takes work, but when you plot every point on a rating, you'll find that the performance scales are not generally linear.

OOTP can be arranged (with league totals and whatnot), to have guys hit .800, or strike out 22 hitters or whatever...once you get a handle on the game engine's structure, entering in extreme control points is a fascinating exercise. I've found it easiest to think of league totals as something you use to define the environment (deadballs/liveballs, raise the mounds, put a league on Mars, all of them can be best dealt with by a twist of league totals). Ratings, then are the universal abilities that people carry with them...potential ratings are OOTP's poorly designed way of playing God.

---

Like I say, there are hundreds of thousands of words that have been written about the game engine and how it creates stats. The first thing to know, though, is that--as the post title says--the ratings can be fickle, so a good GM will be vigilant at learning when and how to read things. It's part art, part science, and part risk assessment.

[RonCo]

Then there's aging, and injury, and talent bump randomness. All of which is uncertain and somewhat unknown.

[RonCo]

On a look at injuries:

Cooper (24) has been injured for only 4 days in 2031, Hasnett (27) has been a walking first aid kit for a couple years. Purely my gut tells me that this is a major source of Harsett's struggles, but I don't really know.

[RonCo]

Here is a 10 year old thread on the OOTP public forums that has several posts from me that get deeper into ratings, Log5s, and league totals.

http://www.ootpdevelopments.com/board/s ... p?t=169727

[RonCo]

Post 30 in this thread is an even deeper dive into the nuances of AvK and pitcher rating skew.

http://www.ootpdevelopments.com/board/s ... p?t=278671

[RonCo]

You don't really _have_ to get into that level of detail, though.

All you really need to do is know that the Contact rating is a calculated rating that is never actually used in the results engine. It's there merely to give you a very general idea of projected batting average, but it can occasionally be considerably wrong (or false) for many reasons.

[RonCo]

Making a long story somewhat longer (and riffing off the OP), these oddities are one reason why I, too, often go essentially with performance over ratings at times.

I'm steeped in details enough that I can often make what I think are reasonable cases about what ratings really exist, but for me the game is surprisingly often played in the mindset of a stats-only league. That may explain why I evaluate certain players differently from how others evaluate them. Dunno.

[Spiccoli]

You don't really _have_ to get into that level of detail, though.

All you really need to do is know that the Contact rating is a calculated rating that is never actually used in the results engine. It's there merely to give you a very general idea of projected batting average, but it can occasionally be considerably wrong (or false) for many reasons.

Wait... what? Really? It's calculated? You mean from Eye & Avoid K's?

I knew Stuff was calculated, but didn't know that about Contact.

I thought Contact was the ability to make a solid hit vs mis-hit... like the difference between a line drive vs a slow ground ball. I mean, that's what making solid contact means to me. Looks like I have some research to do on the those ratings.

[Spiccoli]

Also, thanks Ron for the links. I'll be sure to check them out.

Next Topic... How Coaching really f'ing works in this game....