Handymans Guide #3: Stats from Ratings

[biggerme]

It's been a while that I've done some analysis on OOTP and I wasn't really sure where to post this, but as results are derived from statistics gained from Tournaments, here it goes.

It's common knowledge that a rating of a player results with a certain event with a certain probability, e.g. Contact determines how often a batter will succeed in getting a base-hit. What I have done is use a little tool called Knime and feed statistics from Tournaments into it, you can also use Excel for similar stuff. Difference is in quality of understanding results.
Introduction of Tournaments created a unique opportunity to get relatively comparable sandbox, especially Bronze and Silver ones, which really helped remove these pesky extremities. And an added bonus, these things just run somewhere and all I had to do was export statistics from completed tourneys, instead of running anything on my own setup.

Results where interesting and predictable, "Adjusted R-Squared" being consistently >95%, either for Linear or Polynomial Regression models. And that means there is a very good correlation. Presented statistics are for at least 2000 AB's by rating, this eliminates anomalies of not having enough data.

Little more info, ratings are on scale of 100Pts, resulting numbers are not guaranteed and should be used only to compare player A to B. There will be a major difference for Batting Average when opposing Pitchers Stuff is 60 or 100 or etc. It is a probability against same opposing Pitcher in that case.
One interesting finding, pitchers Stuff rating also is linked directly to how many Extra Base Hits will be allowed, I was not expecting that one at all.


And voila, Batter vs Pitcher chart is born.


Table with some numbers, no point of going below 40 as I suspect it's becoming rather obvious.


And a strange conclusion about what's the magic #, it looks like CON-POW-EYE 70-55-53 STU-MOV-CON

Enjoy …

[macacks]

Originally Posted by biggerme

It's been a while that I've done some analysis on OOTP and I wasn't really sure where to post this, but as results are derived from statistics gained from Tournaments, here it goes.

It's common knowledge that a rating of a player results with a certain event with a certain probability, e.g. Contact determines how often a batter will succeed in getting a base-hit. What I have done is use a little tool called Knime and feed statistics from Tournaments into it, you can also use Excel for similar stuff. Difference is in quality of understanding results.
Introduction of Tournaments created a unique opportunity to get relatively comparable sandbox, especially Bronze and Silver ones, which really helped remove these pesky extremities. And an added bonus, these things just run somewhere and all I had to do was export statistics from completed tourneys, instead of running anything on my own setup.

Results where interesting and predictable, "Adjusted R-Squared" being consistently >95%, either for Linear or Polynomial Regression models. And that means there is a very good correlation. Presented statistics are for at least 2000 AB's by rating, this eliminates anomalies of not having enough data.

Little more info, ratings are on scale of 100Pts, resulting numbers are not guaranteed and should be used only to compare player A to B. There will be a major difference for Batting Average when opposing Pitchers Stuff is 60 or 100 or etc. It is a probability against same opposing Pitcher in that case.
One interesting finding, pitchers Stuff rating also is linked directly to how many Extra Base Hits will be allowed, I was not expecting that one at all.


And voila, Batter vs Pitcher chart is born.
Attachment 663669

Table with some numbers, no point of going below 40 as I suspect it's becoming rather obvious.
Attachment 663671

And a strange conclusion about what's the magic #, it looks like CON-POW-EYE 70-55-53 STU-MOV-CON

Enjoy …

This looks interesting...math isn't my strongest attribute...so what the heck are you saying.

[QuantaCondor]

From which population was this data sampled? Entirely in bronze and silvers? An even mix? There is a major, major difference between the important batter and pitcher stats in, say, a silver or bronze tourney vs. open/cap/Perfect leagues/etc., as you mentioned. And if you collected data over a bunch of different types of tournaments, it might just be better to present those fits separately rather than together, otherwise it's hard to really apply any of this to a region of the game. I also am unsure if 2000 ABs is necessarily the dividing line to use. I mean, for 100 CON for example, there aren't that many players out there being used in tourneys that fit that classification, and probably most of your ABs in the limits are coming from one specific guy's team (or, at minimum, one or two players used commonly on a few different teams). Especially true for high MOV and even high CON stats as well. Or is this just raw player overall you're using on the left side there?

For hits allowed, it's especially not obvious what's the underlying statistic being compared against. Stuff? Player overall? And if it's Stuff, you might expect that a pitcher with high stuff might just end up having high other stats as well. I'd guess you would probably have to remove that correlation to get a true statement of the effect of stuff on hits allowed. There's a lot of potentially interesting data in there, but all the wires are tangled together right now and the threads need to be separated to give them all their proper context.

[biggerme]

Sure, it's not an exact science just an interesting finding. There is anyway L/R split for each player which can have substantial impact and only having more data samples improves accuracy.
For clarification, measures are grouped by rating, e.g. group by Contact and get a batting average for that. Pitchers are then against the same pool of batters from same tournaments which makes it comparable.

There is a major, major difference between the important batter and pitcher stats in, say, a silver or bronze tourney vs. open/cap/Perfect leagues/etc., as you mentioned. And if you collected data over a bunch of different types of tournaments, it might just be better to present those fits separately rather than together, otherwise it's hard to really apply any of this to a region of the game.

And 2 separate pools; Bronze and Silver for statistics, both of them ending up with practically the same regression so final chart is using only Silver as having more data samples. And yes, mixing Open/Cap/... would be a bad idea, it would work but would need about 50X more data samples to get anywhere reasonable.

Resulting batting average example is not something that is a result, changing World to year 2000 League Totals will affect that one, example model AVG = 0.002 X Contact + 0.14, so anyone can have a batting average of at least .140

On chart findings, Contact to Batting Average, Stuff to Hits Allowed is linear, CON50 to CON60 increases probability of getting a hit by 20%. But Eye to Walks Drawn and Control to Walks Allowed is Progressive, as a result increase from rating 50 to 60 is not just 20%, it's instead close to 30%. And same goes for batting Power to HR's and pitching Movement to HR's Allowed.

Batting Gap rating is an oddity and totally not expected, assumption was that all extra base hits are already included in Hits and Gap is only responsible for telling how many of those Singles should be instead Dbls or Trpls.

[One Great Matrix]

Thanks for the pretty graph. Those strange magic #'s? At leasts? 70+ contact = approved?, etc.?

[biggerme]

Originally Posted by One Great Matrix

Thanks for the pretty graph. Those strange magic #'s? At leasts? 70+ contact = approved?, etc.?

Not sure, might be just a coincidence, currently these are breaking points.
Batter with 55 Power will have same amount of HR's as pitcher with 55 Movement will allow HR's.

[Morgans Magic]

So if I read this correctly extra base hits is an independent measure based on stuff vs. gap power?

[biggerme]

Originally Posted by Morgans Magic

So if I read this correctly extra base hits is an independent measure based on stuff vs. gap power?

It does look like that, batters Gap rating is just more relevant in achieving that.

[OMGPuppies]

I like the OP loved the character from In Living Color

[Tony Pellegrino]

Thank you for providing this data. I've been playing around with it today and found something interesting that you may have missed:

HR Allowed for Pitchers is actually linear with Movement

It only appears to curve because it bottoms out at 0.0% before Movement = 100. Looking at the values 40-95, it's a straight line at -0.06% per Movement point.

Thanks again!

[One Great Matrix]

The angle on the gray lines is less than on the green lines, could be that the increased extra-base hits are just a product of increased hits overall with respect to pitcher stuff / hitter contact... if I'm reading the graph half-way right and your comments...

I'm not sure there's really a DIRECT correlation to extra-base hits allowed & stuff except that they are allowing more hits overall, maybe.

[Tony Pellegrino]

Could I bug you to calculate the BA, OBP and SLG for your sample?

Setting all ratings to 60 gives an OPS of 840 for both Hitters and Pitchers, which I found interesting.

[Tony Pellegrino]

Originally Posted by One Great Matrix

The angle on the gray lines is less than on the green lines, could be that the increased extra-base hits are just a product of increased hits overall with respect to pitcher stuff / hitter contact... if I'm reading the graph half-way right and your comments...

I'm not sure there's really a DIRECT correlation to extra-base hits allowed & stuff except that they are allowing more hits overall, maybe.

From looking at the data, there is small influence on XBH Allowed from STUFF but it is there. The data below is Stuff vs XBH Allowed as a percentage of Hits Allowed.

STU XBHA/HA
40 22.40%
45 22.36%
50 22.08%
55 22.04%
60 21.74%
65 21.77%
70 21.45%
75 21.48%
80 21.15%
85 21.17%
90 20.82%
95 20.83%
100 20.54%

[biggerme]

Originally Posted by Tony Pellegrino

Could I bug you to calculate the BA, OBP and SLG for your sample?

Setting all ratings to 60 gives an OPS of 840 for both Hitters and Pitchers, which I found interesting.

One of my favorites in Silver leagues with resulting stats:

[One Great Matrix]

Originally Posted by Tony Pellegrino

From looking at the data, there is small influence on XBH Allowed from STUFF but it is there. The data below is Stuff vs XBH Allowed as a percentage of Hits Allowed.

STU XBHA/HA
40 22.40%
45 22.36%
50 22.08%
55 22.04%
60 21.74%
65 21.77%
70 21.45%
75 21.48%
80 21.15%
85 21.17%
90 20.82%
95 20.83%
100 20.54%

Yeah, going through it I figured that would make sense... I guess movement is the main influence there?...but it would make sense that a pitcher that gave up far less hits would on average give up less extra base-hits (per total hits) as well...stuff being nastier, higher % infield hits, bloopers per total hits...perhaps their movement is just on average higher as well or perhaps the stuff alone being higher keeps the extra-base hits slightly lower. You see it is not absolutely direct there with 90 being slightly lower than 95 and a few other 5 increments demonstrating the same not-exactly-direct pattern.