THE BOOK--Playing The Percentages In Baseball

Oh, and there’s no way that the uncertainty level of your team wins estimate is 10 wins for the season.  The maximum it could be is 6.36.  My guess is that the numbers John was quoting was 2 SD.

Tango

You are right—not quite what went on with the SD numbers, I’ll have to recheck.

Definitely an interesting area. There is definitely scope to apply some sort of real options to baseball, especially for trades.

Yes, I see the mistake. The standard deviations were both sides of the mean, so 2SD, as you say! Doh! I have amended the original article.

What about some people’s belief that WPA is not ‘real’, that it is only some kind of statistical illusion. Would any of these people actually bet money to prove that WPA is an illusion?

There’s a difference between win expectancy (WE) and WPA.

Everyone agrees that WE is perfectly suited for betting.

As for betting on WPA or LWTS, I would bet on LWTS for sure.  Take the top 25 guys in WPA and in LWTS, scratch off the guys who are in both lists, and I’ll take the 10 best remaining in LWTS and you take the 10 best remaining in WPA, and I’ll win in WPA *and* in LWTS.

I was referring to WPA as it was used in the article (where it showed the win probability throughout the game).

I don’t think everyone agrees, for example, that a home run in the top of the 1st of a tied game is worth less than a home run in the top of the 9th of a tied game.

Well, the article was wrong.  WE is WE and WPA is WPA.  The focus of this thread should be WE, not WPA.  Look for the Jeter/clutch thread if you want to talk about WPA.

Insofar as WE is concerned, everyone agrees that the HR in the the latter situation is worth far more, in real-time.  Worth, as in dollars, pulse, and blood pressure.

"WPA” would be better used as a term that focuses on accountability to players, as opposed to simply being the change in WE.

After the games are over and both teams end up winning does that change the ‘value’ of the home runs from what they were when they were hit?

Chris,

Your comments are better suited to the Jeter thread.  We’ve talked about it at length already.  Please click on my name, read that thread, and post any comments you have on “value” there. 

Thanks, Tom

If you have the numbers for the top 25 WPA and LWTS from last year, how do they compare this year?

Chris,

Where in the article do I say that a HR in the to of the first is worth more than one in the 9th? Nowhere.

I think in the article there is a bit of ambiguity as to how I have used the WPA acronym. When I say WPA I refer to the theoretical win probability change.

All I show is the market matches the model pretty well. No more no less. I disagree that anything was inaccurate or wrong.

The whole concept of markets in sports is in fact fascinating (and this thread could be quite lively I think).

One of the interesting questions is whether the “collective wisdom” of the participants in these markets is better than that of, say, a sabermetric expert (a sabermetrician who is an expert in computing WE before and during a game, and other things that these markets address, like assessing team w/l records and playoff chances).  I don’t think it is a given that it is.

Collective wisdom tends to be optimal when you have a variety of experts (not necessarily experts as we traditionally define “expert") using different methodologies (and sometimes the “methodologies” are no more than experience, instinct, and the like) working on the same “project.”

For example, the w/l over/under betting “markets” before the baseball season begins reflects very poor wisdom.  The betting markets during the season reflect pretty good, but not great wisdom (which is why there are many successful sports bettors).

Granted these “tradesports” markets tend to be played by a bunch of very smart people, as opposed to the average bettor that essentially sets the “line” for pre-season w/l over/unders and daily prices on games, but there is still no guarantee that collectively they are better than any one or more experts.  Of course, the more experts there are that infiltrate the market, the fairer and the more optimal the market will be.  It just depends on the ratio of experts’ to non-experts’ influence on the market.

Another example of collective wisdom that is good but not great is in horse racing.  Some people are amazed at how good the “public” (all the bettors) is at handicapping horses when conventional wisdom has it that so many race bettors (presumably) simply bet on the horse whose name they like.  The fact is that there are many experts who can do a lot better than the entire pool of bettors on horse racing.  Of course, the “juice” in horse racing is so large (15-21% or so) that even an expert handicapper will have trouble beating the game.

As far as teams using these markets to assess the value of anything, that is fine and dandy, however…

You can’t even get more than a handful of “teams” to employ and/or listen to a competent sabermetrician (who can save them millions of dollars and/or add several wins to their team success with their eyes closed, and essentially do the same thing as the futures market), and most that do, usually do it in a half-ass fashion, how would you expect those same teams to start “listening to” a sports futures market?

MGL—you say at over/under betting at the start of the season reflects poor wisdom. I am not sure this is necessarily true, I think it reflects high uncertainty. For instance, projection systems for W-L don’t outperform betting systems, and neither should they given that the “information” is available in the market.

It could be that there are a bunch of proprietary systems that are better than the market but I would doubt that.

Markets like tradesports tend to be pretty good because you often have market makers. Also because bettors may have open contracts any change in win expectancy is quickly arbitraged away.

As to your point about teams incorporating futures markets in their decision making process as I with you that the conservative nature of most front offices make it unlikely in the near future ... but you never know with some of the more progressive organisations out there!
_______________

Just to clarify my previous post on WPA—I should add that I am only looking at theorertical win probability change from a team perspective and not individual players.

John, right you are talking about WE and change in WE.  “WPA” is typically referred to player totals and accountability.  We should focus on WE here.

***

I agree with MGL that it is fascinating.  Anything involving “Wisdom of the crowds” is fascinating to me.  I am continually amazed at how well a semi-knowledgeable, yet independent, crowd can do.  Here’s another example, involving chess.

Point taken on WPA—the focus should be on WE.

For those interested there is a play-money futures market at: http://sports.us.newsfutures.com/

Not the most liquid of markets but it tends to work reasonably well during ball games.

Great thread.

The final paragraph of MGL’s post raises another fascinating (if somewhat unrelated) question, IMO.

That is, what is the value of a competent sabermetrician? Put another way..."What is MGL Really Worth?” [Insert winking smiley-face emoticon here].

Against what measures should a competent (and paid) sabermetrician’s advice be measured? How MUCH better are the proprietary defensive systems than others that are freely available? How MUCH better are teams’ internal scouting departments than the reports anyone can find for 50 bucks in BA or PGCrossChecker? How MUCH better is Bill James at projecting Red Sox players’ future performance than the posters on SoSH? How MUCH better are proprietary methods for determining a player’s appropritate salary than what one can find on THT or BP?

Prediction markets are a fascinating avenue for future sabermetric research because, (in addition to the low-cost alternatives mentioned above), they could provide a type of S&P 500 Index Fund check on what teams are getting for their money.

Sorry about the hi-jacking.

Thanks.

- Kent

There are many people, whose names shall not be diclosed, who slaughter the pre-season over/unders (team w/l market).  There is NOT a whole lot of wisdom in those “lines.” The simple reason is that the structure is such that the liquidity ("action") is dominated by the ignorant general public (the squares).  It is a little different from a pure futures market, like Tradesports, with lots of liquidity, and which, as I said, is dominated by very smart people.

For example, the Yankees might be initially set at 98 wins.  The “smart” bettors realize that this is too high, and bet it down to 95.  The public then bets it back up to 98.  A few more smart bettors bet it down to 95.  The public bets it back up to 98.  There are no more smart people left (even though the original smart bettors can bet it again of course) and it stays at 98 (or 97 or whatever).

That is very simplistic, but that is the way it works, more or less.

One can estimate (WAG) what a sabermetrician is worth.  Of course, many people will argue that a sabermetrician is worth nothing or negative.

Player contracts, trades/acquisitions, and roster construction: 3-5 mil per year (more for teams with larger payrolls).

In-game strategies, lineup construction, and bullpen managment: 3-4 wins or 7-10 mil a year.

Miscellaneous: 1-3 mil per year.

Draft, player development, etc.:  Who knows?

Hmm—interesting. You are correct that there isn’t a whole lot of liquidity in the over/under lines and I have no doubt that some people make an absolute killing on these but I wouldn’t have thought that the movement is as dramatic as you say (is it really say 3 games, or more like 1 game). For a start bettors don’t alternate betting in groups of smarts and stupids—but I agree the effect will be present.

Over/under isn’t a true market beacuse it is made an backed by bookies so there will be an overround effect. Also bookies tend to take positions in sports betting as they tend to be more savy than the stupids. I do think that future markets (be it fixed future contracts or call / put options) will be more accurate as they are not made by bookies and have higher liquidity.

I am not sure if it is something that you discuss in the projection roundtable at THT but std dev of over/under and, say, PECOTA are similar—not sure what that says about either though!

_________
MGL—on your esimates of sabermetrician worth I sincerely hope that you are fairly paid!