See copyright notice at the bottom of this page.
List of All Posters
SABR 301- Win Probability Added (June 26, 2003)
Discussion ThreadPosted 8:37 a.m.,
July 1, 2003
(#2) -
Sky
Tango, in the Win Shares pdf you and Rob Wood put together, the main conclusion was that Win Shares "fails" at a basic theoretical level in its attempt to assign absolute win shares to specific players since it can result in players getting credit for more win shares than game shares. Is this a problem inherent to James' Win Shares method specifically, or any attempt to assign absolute wins? It seems like WPA would run into the same problem (not that you're trying to give credit for absolute wins, just marginal wins above average).
Wouldn't it make more sense, when trying to assign absolute wins, to compare every player to the worst possible players (100% SO for hitters). Only a team of perfectly bad hitters could manage to lose every single game. Replacement players win their fair share and should still get credit for some absolute wins, if not many. And on the flip side, the only team wouldn't lose any game would be a team of perfectly good hitters (1.000 OBP) since even a team scoring 20 runs per game would still lose occassionally.
What if, to measure Bonds' absolute wins, one computed the expected wins with him in the lineup (actual, pythagorean, whatever), and then subsituting in a .000 OPS guy in his spot. That would give his absolute wins, no? The other end's a little harder, since as your near infinite run scoring, all positive hitting outcomes converge towards equal value, but in the MLB environment, they definitely aren't. Would you say a perfect hitter gets all walk? All doubles? All homeruns? There's a big difference.
Chance of Winning a Baseball Game (October 20, 2003)
Posted 8:08 p.m.,
October 20, 2003
(#1) -
Sky
That's pretty interesting, thanks. The 4.3 RPG is key, but doesn't it also matter the distribution of those runs (which is probably dependant on the component events that result in the runs)?
An extreme example - a league that only hits singles, but hits .300 might average the same number of runs as a league that hits .100, but only hits HRs. I would think the run distrubutions per half inning would be different in those cases...
What's a Ball Player Worth? (November 6, 2003)
Posted 7:02 p.m.,
November 11, 2003
(#26) -
Sky
If you were to find what the average value of a homerun is in regards to Win Probability Added (ignoring score, base state, and number of outs), would it hold the same value as it does in linear weights? I guess that's really two questions in one. Should it, by definition have the same value. And if not, does it end up being close, anyways? Are there specific stats that have comparable value in WPA and linear weights, and stats that don't? This thought was triggered by the fact that an IBB has some value in linear weights, but basically zero value in WPA.
Win Shares, Loss Shares, and Game Shares (November 15, 2003)
Posted 2:06 p.m.,
November 17, 2003
(#15) -
Sky
Win Advancements seem pretty interesting, Tango. Please keep the info coming. Of course, if I have to wait for the book, I will.
My thought is that the concept of advancements could be adapted to the entire season. "Playoff Advancements" or something like that. The goal of the season, after all, isn't just to win games, but to win enough games to make the playoffs. Now, most teams are on pretty equal footing early in the season, giving each win pretty similar value. But when you get to September, it's true that the value of ARod helping the Rangers towards the playoffs is extremely low. And wins for the Cubs/Astros during the 2003 stretch were EXTREMELY valuable.
A method like this might work: For each game, compute the probability change for a team making the playoffs. Use these probabilities as weights
I don't necessarily think this would be a great measure of player value or ability (definitely not ability), but it would definitely answer the version of the MVP question typically asked by Joe Morgan/Jayson Stark type writers.
Win Shares, Loss Shares, and Game Shares (November 15, 2003)
Posted 4:38 p.m.,
November 17, 2003
(#18) -
Sky
Tango, I also believe that a player's goal is to help his team try to win each individual game, but since the MVP is decided upon based on regular season performance and many many sportswriters believe making the playoffs is basically a pre-req for winning the MVP, a Playoff Advancement metric would be the technical answer to their question. It would pretty much be a worthless stat, in my opinion, but I just thought I'd point it out.
Again, thanks for all the work on Win Advancements.
HOOPSWORLD.com Review: Pro Basketball Prospectus 2003-04 Edition (November 18, 2003)
Posted 7:15 a.m.,
November 20, 2003
(#4) -
Sky
For solid football analysis, check out FootballOutsiders.com
FANTASY CENTRAL (February 21, 2004)
Posted 12:45 p.m.,
March 7, 2004
(#112) -
Sky
I'm a little confused about this whole standard deviation thing. Why use standard deviations instead of just raw stats above replacement position?
Sure, it's true that the standard deviation for stolen bases is similar to that for HRs, just with a much lower mean. But even accounting for replacement level, you need way more HRs just to get into the range of earning points, where the standard deviation becomes important.
Thus, even beyond replacement level, there are a certain number of HRs per player that aren't used to move past other teams, but are used just to to earn the possibility of moving past teams if you get even more HRs.
I believe the standard deviation option is basically a standings gain points model, which I didn't think was as accurate as a pure value over replacement model.
Thanks in advance for clarifying.
FANTASY CENTRAL (February 21, 2004)
Posted 10:31 a.m.,
March 8, 2004
(#116) -
Sky
Thanks, Tango. I guess I missed the part where you subtracted out league-average stats from each player's category totals before dividing by the standard deviation. (Which makes perfect sense when normalizing a value, obviously.)
I've read a lot of criticism of the Standings Gain Points method, which is similar to what this SD method is. The main fault is that the SGP method doesn't account for a barrier threshold that's needed to start accumulating points. This SD method DOES deal with that, so I'll have to compare it to what I currently use. Anything inherently wrong with...
Subract out replacment levels stats from each player. Add up the "useful" stats of all players and figure out $$/stat ($/HR, $/SB, etc) for each category. Multiply useful stats by $$/stat to get value.
Also, how does this model deal with the counting stats, such as batting average?
FANTASY CENTRAL (February 21, 2004)
Posted 7:49 p.m.,
March 11, 2004
(#120) -
Sky
Still not convinced, although I'd like to be, one way or the other.
Assuming you want to use replacement players, not replacement categories, couldn't you just approximate a replacement level player by averaging the statistical categories of the worst+5 through worst-5 (a range of players around the worst) players at each position, assuming the groups are large enough? Then you would compute stats above this theoretical replacment level in each category?
FANTASY CENTRAL (February 21, 2004)
Posted 10:25 p.m.,
March 11, 2004
(#122) -
Sky
Nod - up to now, I've added up all the Rs, HRs, RBIs, SBs, and xH above replacement level and divided by the total number of that stat above replacement level. So if everyone in the draftable player pool had a total of 1000 HRs above replacement level, Player A would have .5% of the pool. Since for auctions you usually deal with dollars, you convert the .5% to .5% of the $$ allocated to HRs. If you assume 12 teams, $260 and a 2/3 hit/pitch split, Player A would earn about $2 for his HRs. Repeat for each category.
FANTASY CENTRAL (February 21, 2004)
Posted 7:10 p.m.,
March 12, 2004
(#125) -
Sky
(homepage)
I make separate lists of catchers, middle infielders, and 3B/1B/OF/DHers. Now, 3B are almost as bad as the the middle infielders these days, but at least you have the 1B to fill up the CI positions. With these three different lists, you can create a different replacement level for each group. So to determine useful SBs for middle infielders, subtract out the middle infielder SB replacement level, which is different from the catchers SB replacement level, etc. You could theoretically separate out ALL positions, but that gets to be a pain in the ass.
A slightly simpler method is to use the same replacement level for everyone and add on a certain amou (equal to $1 minus the worst catcher vale ) to every catcher's value. For a 12 team league-specific league, it's about $3 to $4. Adding this much doesn't really affect the total money in the pool, so you don't need to adjust the rest of the league.
FANTASY CENTRAL (February 21, 2004)
Posted 10:24 p.m.,
March 15, 2004
(#132) -
Sky
(homepage)
Michael, that is an extreme, although representative example. Todd Zola over at Mastersball has pretty much shown that for typical player populations, taking the highest valued player (above replacement) every round is the best way to go.
If, in your example, we assume the entire league is made up of 10 teams each picking one 3B and one 3B and we have the first pick, it doesn't matter who we pick because we'll get a 60 and a 10. The second pick will get a 60 and a 10.
If, more realistically, there are a lot of players at other positions left and some players are already off the board and it's our turn at some random point in the draft, it still really doesn't matter which $60 player we pick. Assuming the league is somewhat rational and has left many players at other positions in the 30 to 60 range, if we take the 60 1B, we can fill up on the other positions before taking a 3B at the lower end.
There are very few decisions at a draft where thinking about the game theory side of things will prove useful.
FANTASY CENTRAL (February 21, 2004)
Posted 3:08 p.m.,
March 19, 2004
(#143) -
Sky
Here's a question i'd like to propose to folks concerning points leagues...
In roto leagues, an uneven split between hitters and pitchers is common place, mostly because more pitcher value is available from the free agent pool throughout the season than hitter value because pitcher performance has more variability given the traditional roto categories.
In points leagues, pitchers are probably still more unpredictable, but how does that manifest itself in some sort of weighting? If you figure out how many points the pitchers and hitters are projected to earn, you should probably discount pitcher points somewhat due to their unpredictability. But how? In roto, half the categories are pitching and half are hitting, by definition. But with points leagues, one group could be given more importance because of the point values.
If youre draftable player pool for hitters contains as many points as the draftable player pool for pitchers, then you might want to apply the same roto split (something like 1:2). But for anything else, how do you know?
MGL - Questec and the Strike Zone (March 20, 2004)
Posted 4:40 p.m.,
March 20, 2004
(#2) -
Sky
There should be a file called questec.pdf