Simple Suggest: Add an Analyst

[BaseballATeam]

Now that advanced stats have landed, add a toggle-able "analyst" similar to the "incorporate stats in scouting reports" toggle. The analyst adds a current and (when relevant) future projection at the end of a scouting report. This should serve as a less-accurate version of the expected stat lines produced in the player editor.

For all of OOTP history, I've been frustrated by the inability to ask my scout "how do you think he'll perform?" They just give you one of ~five relevant canned responses (and I'm not saying you need more ways of reporting that a player is middling) and the number grades. I'm left to figure out if they'll hit .220/.310/.450 or .290/.320/.390--which is fine (I guess) but usually isn't worth the effort.

[ColumbusJets]

Advanced stats do not give any insight into future performance.

[BaseballATeam]

Are you saying OOTP's implementation of advanced stats doesn't correlate with future output? I wouldn't be surprised if their first release isn't a total success.

If you're instead asserting that advanced stats aren't useful for assessing irl performance, that's just incorrect.

More importantly, my comment isn't really about the analyst "using" the advanced stats to set a future expectation. Without using commish mode and going into the player editor pages, the game is missing a built-in tool for assessing a player's expected/future contribution. We get all this information, tool grades, stats, write-ups, and leave out the one detail real front offices want to know--what do we think this guy can do?

And I suggest making that a toggle so OOTP purists can continue to do that assessment themselves.

[locuspc]

There's a contingent of people who gripe about the implementation of advanced stats because they want OOTP to be a physics simulation instead. Just ignore the trolls. It's easy to verify that advanced stats do in fact correlate with future performance in the game engine.

[ColumbusJets]

Originally Posted by locuspc

There's a contingent of people who gripe about the implementation of advanced stats because they want OOTP to be a physics simulation instead. Just ignore the trolls. It's easy to verify that advanced stats do in fact correlate with future performance in the game engine.

Advanced stats show what happened. Same with regular stats. They don't reveal what will happen in the future. If a player over performs or under performs ratings then that's randomness. and there's a lot of randomness in the game.

Statcast simulations are based on ratings. They are not based on play outcomes. So what they show is not related to the performance seen in advanced stats. People are looking at these two things as related when they are not.

With a sufficient sample size Statcast simulations could be a clue that scouting is inaccurate. A sufficient sample size is unlikely.

If, as you say, "it is easy to verify" that advanced stats predict future performance then do it. Run some tests with sufficient sample sizes and results that can be duplicated by others. And if performance gets better, explain how better performance happened during the prediction period without ratings to drive it and how better performance during the prediction period resulting in better ratings to drive future performance.

[ColumbusJets]

Originally Posted by BaseballATeam

Are you saying OOTP's implementation of advanced stats doesn't correlate with future output? I wouldn't be surprised if their first release isn't a total success.

The only thing the game is giving you that is a clue about future performance is the scout's potential rating.

Actually advanced stats is nearly a total success. Most people have already forgotten the interview where Matt waffled about the effect and people are believing the stats do what they want (hope?) they do.

[jcard]

Originally Posted by BaseballATeam

Now that advanced stats have landed, add a toggle-able "analyst" similar to the "incorporate stats in scouting reports" toggle. The analyst adds a current and (when relevant) future projection at the end of a scouting report. This should serve as a less-accurate version of the expected stat lines produced in the player editor.

For all of OOTP history, I've been frustrated by the inability to ask my scout "how do you think he'll perform?" They just give you one of ~five relevant canned responses (and I'm not saying you need more ways of reporting that a player is middling) and the number grades. I'm left to figure out if they'll hit .220/.310/.450 or .290/.320/.390--which is fine (I guess) but usually isn't worth the effort.

Glad to see someone else pick up this mantle. I have on multiple occasions raised the same issue, in virtually identical terms (right down to the editor parallel). I certainly hope that you prove more effective than I in generating community interest and support and compelling developer attention.

[PSUColonel]

Originally Posted by jcard

Glad to see someone else pick up this mantle. I have on multiple occasions raised the same issue, in virtually identical terms (right down to the editor parallel). I certainly hope that you prove more effective than I in generating community interest and support and compelling developer attention.

I have always been for more detail no matter what the aspect of OOTP. The problem is there’s a large contingency of players (and also developers) who don’t want to turn OOTP into FM.

[locuspc]

Originally Posted by ColumbusJets

Advanced stats show what happened. Same with regular stats. They don't reveal what will happen in the future. If a player over performs or under performs ratings then that's randomness. and there's a lot of randomness in the game.

Statcast simulations are based on ratings. They are not based on play outcomes. So what they show is not related to the performance seen in advanced stats. People are looking at these two things as related when they are not.

With a sufficient sample size Statcast simulations could be a clue that scouting is inaccurate. A sufficient sample size is unlikely.

If, as you say, "it is easy to verify" that advanced stats predict future performance then do it. Run some tests with sufficient sample sizes and results that can be duplicated by others. And if performance gets better, explain how better performance happened during the prediction period without ratings to drive it and how better performance during the prediction period resulting in better ratings to drive future performance.

Advanced stats come from ratings. Future performance comes from ratings. You understand both these things. How is it not immediately obvious to you that this means future performance is correlated with advanced stats?

Yes, if you're playing in commissioner mode and looking directly at the underlying ratings in the editor, the advanced stats don't tell you anything you couldn't figure out on your own. Most people don't play that way. You shouldn't go around saying things "advanced stats do not give any insight into future performance" without caveating it with "once you already know a player's ratings with perfect accuracy".

[ColumbusJets]

Originally Posted by locuspc

Advanced stats come from ratings. Future performance comes from ratings. You understand both these things. How is it not immediately obvious to you that this means future performance is correlated with advanced stats?

Yes, if you're playing in commissioner mode and looking directly at the underlying ratings in the editor, the advanced stats don't tell you anything you couldn't figure out on your own. Most people don't play that way. You shouldn't go around saying things "advanced stats do not give any insight into future performance" without caveating it with "once you already know a player's ratings with perfect accuracy".

It is incorrect to say that all advanced stats come from ratings. I have been careful to state Statcast simulations as a subset of advanced stats. That is because they come from ratings. Others are based on performance in the game. Some of those are BB% for batters and HR/FB for pitchers. These tell only what happened not what will happen.

The simulation of Statcast is different. As you say, it is based on ratings. That means given sufficient sample size they could be used to evaluate the accuracy of the scouting rating. The sample size is insufficient.

I've tried to identify advanced stats (based on performance) and Statcast simulations as separate things. I'm sorry I haven't been adequately clear on the difference.

You've stated its easy to verify that advanced stats can be used to predict future performance. I contend that 1) those advanced stats based on performance cannot be used to predict future performance, and 2) the sample size of Statcast simulations too small to be useful. So if you have some tests that show otherwise for either 1) or 2) that are conclusive and can be duplicated by others please post them.

[mcercone]

I ran through a season and noticed how Judge had the highest EV, and Kwan was one of the leaders in contact %. Therefore, I am seeing a correlation to where the players rank in real life with some of these statistics. Are they based upon the ratings? Probably. to a certain degree because that is how the "engine" determines these Advance statistics. I know they are not arbitrary. So can they predict future results? Like any statistic, I believe they can to a certain degree. But there is such things as sample sizes, averages. Therefore you have to take everything into account when analyzing a players performance.

[mcercone]

I ran through a season and noticed how Judge had the highest EV, and Kwan was one of the leaders in contact %. Therefore, I am seeing a correlation to where the players rank in real life with some of these statistics. Are they based upon the ratings? Probably. to a certain degree because that is how the "engine" determines these Advance statistics. I know they are not arbitrary. So can they predict future results? Like any statistic, I believe they can to a certain degree. But there is such things as sample sizes, averages. Therefore you have to take everything into account when analyzing a players performance.

[BaseballATeam]

You've stated its easy to verify that advanced stats can be used to predict future performance. I contend that 1) those advanced stats based on performance cannot be used to predict future performance, and 2) the sample size of Statcast simulations too small to be useful. So if you have some tests that show otherwise for either 1) or 2) that are conclusive and can be duplicated by others please post them.

I don't know where you're getting your assertion for #2. One of the reasons analysts like Statcast metrics is the relative speed with which they become statistically relevant. A max EV, for instance, requires a sample of 1. Avg EV and EV90 take 50-60 BIP. Launch angle takes a while to settle down, but you can usually discern a lot from 30 BIP. Running speed is instantly predictive. Bat speed requires 3-10 swings.

These are facts you can google yourself if you want to question them. As to how they function in-game? I don't know! But, that's not relevant to the topic.

All I'm saying is since the game has accepted that advanced stats are important, why not incorporate "front office staff" that perform the role of analyst? Whether or not they actually "use" the advanced stats, it would be flavorful to have a staffer tell me "this is a classic second hitter with an OBP-over-power profile."

(PS, I'm available to write the lines and help develop the concept. I'm a former FanGraphs, NBC, MLBTR staffer. Lauerman knows how to get in touch with me.)

[ColumbusJets]

The performance stats are not predictive of future performance.

Some of the performance stats are simulated. Made up. For batters, chase rate, swinging at strikes, swing percentage are totally made up. The game calculates outcomes per PA not per pitch.

The Statcast "data" is simulated. Made up.

A batter is created by entering PA, AB, H, 2B, 3B, HR, BB, HP, and SO. These are all derived from per PA data. Nothing about his swing and miss rate or his EV.

Then the game arbitrarily decides if he's a pull hitter or a spray hitter and his ground ball/fly ball ratio. And that doesn't affect his output at all. Nothing but flavor.

Performance stats, whether actual or concocted, whether regular or advanced, show how a player performed but do not show why. This is contrary to the statement on the Buy Now page "This is a breakthrough for modern fans, who now have all new tools to understand how their teams are performing – and why!"

The simulated Statcast "data" with sufficient sample size might give an indication the scout has misevaluated a player. It does not show how a player performed, again contrary to the statement on the Buy Now page.

Sufficient sample sizes for some real life Statcast data is viewed as easily obtainable. Fine, but does that apply to simulated Statcast "data" derived from a source other than the outcome of the play?

How is this stuff useful? None of the performance data is. The Statcast simulation is marginally useful in that in some cases it could give a clue that perhaps the scout's rating might possibly be a bit off when scouting is set to less than 100%. People should go back and watch Matt waffle and refuse to commit in the road to release interview.

But then, there is nothing easier to accept as true than something a person wants to be true.

[BaseballATeam]

You're missing the point. I'm concerned about flavor and verisimilitude. It ultimately doesn't matter if the advanced stats are or aren't useful/predictive/whatever (which, btw, check out the threads on using AI to evaluate the advanced stats).

The game makers have correctly identified that a baseball simulator without Statcast in some form is not viable. If the game makers accept this, then they should take the next logical step and insert an analyst function into the game. For verisimilitude.

We all understand the baseball simulator is simulating.

[Sweed]

Originally Posted by BaseballATeam

You're missing the point. I'm concerned about flavor and verisimilitude. It ultimately doesn't matter if the advanced stats are or aren't useful/predictive/whatever (which, btw, check out the threads on using AI to evaluate the advanced stats).

The game makers have correctly identified that a baseball simulator without Statcast in some form is not viable. If the game makers accept this, then they should take the next logical step and insert an analyst function into the game. For verisimilitude.

We all understand the baseball simulator is simulating.

Spot on.

Though I'm not as confident as you with the last sentence.

[BaseballATeam]

Originally Posted by ColumbusJets

BaseballATeam refers to Statcast earlier in his post. Since that was the established context I took it that his comment "we all understand" referred to his comments that Statcast simulation was flavor. Most people do not understand that and considering his prior posts in this thread that is a recently gained understanding on his part.

Nope.

[BaseballATeam]

Originally Posted by ColumbusJets

No body would have to get personal if just one somebody would post proof advanced performance stats can be used to predict future performance and Statcast "data" has anything to do with the associated play.

You can make that thread if you'd like. The topic here is incorporating an analyst role into the game. Because it's realistic. That's all.

I apologize for confusing matters by linking the implementation of the advanced stats with my idea for an analyst. To me, 'if advanced stats, then in-game analyst' is faultless logic. You seem to have read "if advanced stats" and melted.