Winshare Formula in Cato

[Comedian2004]

Does anyone know if Cato uses the Short Formula for figuring out Winshares or the Long Formula?

Also, See if I understand this right.

End of the season a guy was figured to have 53 Winshares. (Using the short formula). His team won 79 games.

So, I say 79 * 3 = 237 (This is how many Winshares this team has)

I then have to add up all the winshares that I figured out and see what they total. Let's say they totaled 300.

Now, I have to adjust the '53' to the team total, so that would be:

237 divided by 300 multipled by 53 = 42 (rounding to nearest integer)

So, this guy actually had 42 Winpoints. Does this make sense (To anyone that knows the Winpoint formula)?

[Jestor]

Bumping this for Comedian.

[arod23]

I am pretty sure it uses the long formula, or at least as much of it as possible, given that all of the inputs may not be available. At the individual player level, you can see in the data file the it calculates multiple Runs Created formulas which appear to then be adjusted based on team Runs Created. From memory, that is part of the long formula process.

Unfortunately, the Win Shares calculations are part of cato's code and not done in Access like many of the other calculations. So, I can't see exactly what it is doing. Maybe one of the programmers out there can peek at the code.

[Comedian2004]

I don't think it can use the long formula and be accurate without every single available data that is require. The ball park adjustment factor would be just about impossible to figure out, unless he just used the numbers supplied by James, i.e. Arizona's is 1.014 for Park Adjustment and 1.020 of Park Factor. Just about every entry is modified by this. Also, there is not enough fielding data to properly figure that, and just ignoring the fielding or the batting or anything else that is in the long formula would make the formula worthless.

Why would catobase figure out RC, since it is already a stat displayed by OOTP?

Originally Posted by arod23

I am pretty sure it uses the long formula, or at least as much of it as possible, given that all of the inputs may not be available. At the individual player level, you can see in the data file the it calculates multiple Runs Created formulas which appear to then be adjusted based on team Runs Created. From memory, that is part of the long formula process.

Unfortunately, the Win Shares calculations are part of cato's code and not done in Access like many of the other calculations. So, I can't see exactly what it is doing. Maybe one of the programmers out there can peek at the code.

[gmo]

As best I can tell CatoBase goes about as complicated/precise as it can. I believe some things are guestimated, but I do not think there is much simplification of the full calculations.

http://www.ootpdevelopments.com/boar...ad.php?t=66594

[Le Grande Orange]

I've been wondering about adjusting a player's stats based on his team's park factor. Such corrections are applied to a player's stats for the entire season.

But there's a problem with this: his team does not play its entire season in its home park. It only plays half its season there.

By extension, this means that individual players will also only play around half their games in their team's home park. So why then is the park factor being applied to a player's stats for the entire season? Isn't this an inaccurate correction? Shouldn't the home park factor correction only apply to the stats that the player actually put up in his home park? Or have I missed something in the way park factors are determined?

[arod23]

Originally Posted by Comedian2004

I don't think it can use the long formula and be accurate without every single available data that is require. The ball park adjustment factor would be just about impossible to figure out, unless he just used the numbers supplied by James, i.e. Arizona's is 1.014 for Park Adjustment and 1.020 of Park Factor. Just about every entry is modified by this. Also, there is not enough fielding data to properly figure that, and just ignoring the fielding or the batting or anything else that is in the long formula would make the formula worthless.

Why would catobase figure out RC, since it is already a stat displayed by OOTP?

From what we can see outside of the actual code, it is more likely that Catobase calculates Win Shares from the long formula. It does calculate ballpark factors and he doesn't ignore pitching or fielding. If you look at just the Access files, you will see all of the detail. I don't have my Win Shares book handy but the data files contain all of the data needed at the team level.

There are about 1000 different RC formulas out there, who knows which one OOTP uses. I am sure Catobase recalculates RC so that it matches the forumla used in Win Shares, which includes marginal runs created et al.

Outside of WS, Catobase also includes Furtado's Extrapolated Runs (XR) for individual batting rather than RC since RC is not meant for individual performance, it was designed for team performance. XR was specifically designed to evaluate batting at the individual level. If you have ever compared RC in OOTP to XR in Cato you will sometimes see quite a difference in the leaders.

[Comedian2004]

That is an interesting point. It also relates to something I bring up. Some ratings are compared to the 'league average'. The league average in the AL contains the pitcher's batting. So, the NL averages are going to be lower than the AL averages. It also includes guys that are brought in for defense replacements.

So, a guy in the NL will get higher winshares because of the lower total numbers.

Throw in other things like.. a team winning 12-3, the team then brings in the backup players to give the starters a rest...

Even though baseball if a numbers game and based so much on stats, I do not think there will ever be a formula that is 'perfect'. There always has to be that human factor thrown in. For example, a catchers arm or an outfielders arm cannot be measure on stats only. (He threw out 75 percent). Players playing the game know the arms of outfielders and catchers. They do not run on Pudge as much as they do on other catchers. We are also aware of how much the pitcher has to do with the success of basestealers, not to mention the situation. (Down 6-1, 2 outs, they are not going to care if you steal 2nd base).

(I do not see runners taking 2nd base in OOTP and no throw being made and no SB given either)

Even stats like the RBI. I would 'guess' if you went through the RBI leaders for each team over the past 100 years, that a larger percent of them appeared from the #4 position. Same thing with runs scored. Bat the guy in the leadoff spot and he will more and likely score the most runs.

How many times have you seen the #6, 7 or 8 spot leading the team in RBIS?

ERA is the one that is really silly. How about this one...

Smith gets the 1st two batters out, then walks a hitter. They take Smith out and bring in Jones. He allows a HR. Smith is charged with 1 run, Jones is also charged with 1 run. I think that Smith should be charged with 1/2 run and Jones should be charged with 1.5 runs. That would reflect the situation much better and make the ERA more reliable.

Here is my favorite. 2 outs, no one on. A ground ball to the pitcher, he muffs the throw to 1st and the hitter goes to 2nd. An error is charged on the pitcher. The next batter singles and a run scores. An unearned run is charged to the pitcher.

Originally Posted by Le Grande Orange

I've been wondering about adjusting a player's stats based on his team's park factor. Such corrections are applied to a player's stats for the entire season.

But there's a problem with this: his team does not play its entire season in its home park. It only plays half its season there.

By extension, this means that individual players will also only play around half their games in their team's home park. So why then is the park factor being applied to a player's stats for the entire season? Isn't this an inaccurate correction? Shouldn't the home park factor correction only apply to the stats that the player actually put up in his home park? Or have I missed something in the way park factors are determined?

[Le Grande Orange]

Originally Posted by arod23

There are about 1000 different RC formulas out there, who knows which one OOTP uses.

Well, maybe not a thousand, but there are definitely several.

Originally Posted by arod23

Outside of WS, Catobase also includes Furtado's Extrapolated Runs (XR) for individual batting rather than RC since RC is not meant for individual performance, it was designed for team performance.

XR also has the benefit of being a much less complex formula than the "thorough" version of RC.

One could also consider ERP (Estimated Runs Produced) or its enhancement NERP (New Estimated Runs Produced), both of which are similar to XR in formula construction as well as producing reasonably accurate results.

[nestajones]

Originally Posted by Le Grande Orange

I've been wondering about adjusting a player's stats based on his team's park factor. Such corrections are applied to a player's stats for the entire season.

But there's a problem with this: his team does not play its entire season in its home park. It only plays half its season there.

By extension, this means that individual players will also only play around half their games in their team's home park. So why then is the park factor being applied to a player's stats for the entire season? Isn't this an inaccurate correction? Shouldn't the home park factor correction only apply to the stats that the player actually put up in his home park? Or have I missed something in the way park factors are determined?

park factors are one of the most complex things in baseball statistics, and seem to be one of the most misunderstood. there are a lot of different ways you can measure them depending on your goal. with regard to Win Shares, i think what is done is comparing all runs scored in a team's home games to all runs scored in a team's road games to calculate the park factor. then, yes, if you determine his park increases scoring by 10%, you should decrease the value of the player's total runs by 5% (assuming half the games played at home). i can't remember exactly how Bill James handles it in Win Shares, but i assume he does it correctly and by extension so does catobase.