THE BOOK--Playing The Percentages In Baseball

What do you mean “match the experts as a group?”

These numbers mean nothing without knowing the Vegas pointspreads.  If you are just picking teams to win, not against the spread, I assume that you will do the best in the long run by picking all of the favorites.  I am not sure off the top of my head how often the faves collectively win in NFL, but I can look it up easily enough.

If you are the greatest handicapper in the world and you are picking games NOT against the spread, you are still going to pick almost all of the favorites.  Occasionally you will pick some dogs when the spread is small.

This Accuscore/StatShark company makes some bold claims about their projections.  All handicapping services make similar claims and they are all B.S.

So ESPN and other mainstream media companies are now in the business of using and touting “tout services.” That’s interesting.

Nowhere in their website, which is extensive, do they (StatShark) give even a hint of their methodology.

As is almost always the case, anyone who can actually sustain a large advantage in sports betting and can do so while betting on a substantial number of games, is almost always a zillion times better off betting than running a tout service. That is one reason why 99.9% (maybe 100%) of all tout services do NOT have a long run advantage in sports betting.

There are 3 kinds of sports handicappers in the world.  One are those that think they can “beat” a sport or sports but actually can’t (maybe they can neutralize the juice or even have a slight edge, but that is about it).  Many of them are “touts” (run for pay handicapping services).  There are zillions of these people.

Two are those who know that they can’t “beat” sports but run tout services nonetheless.  They are pure crooks.  There are not a zillion of these people but there are many of them.  And many of the tout services fall into this category.

Three, there are really “smart” people who can actually get a significant edge in (betting) one or more sports.  Depending on the cutoff point of the long run edge they have (and no one really knows for sure what that edge is), these people are very, very rare.  Of these exceedingly small amount of persons, vrey few if any sell their knowledge because, as I said, it is generally more profitable to bet rather than tout.  Of course there are some legitimate reasons why a person may decide to sell their handicapping expertise rather than bet.  Some do both of course.

The above is why 99.9% of all people/companies that sell handicapping information and sports picks do NOT have a significant edge over and above the juice.  Most of them (95+%, maybe more) will not do better in the long run than throwing darts.

That being the case, even though StatShark and Accuscore sound wise and have a really nice website, the default opinon on them regardless of the great results they claim (virtually ALL handicapping services boast great “documented” results) must be that they are a fraud like all other tout sites.

I am actually shocked that mainstream media would actually hook up with a clearly handicapping, touting service.

I was wondering during the past few weeks how I’d be able to do in baseball betting.  As a guy who’s spent a lot of free time reading about and working with baseball stats, part of me thinks I’ll have some kind of advantage… primarily, I’d guess that there’s some opportunity around games in which one or both starters have a large disconnect between their DIPS (or FIP) ERA and their actual ERA, for the current season to that point.  Are there any sabermetrics guys out there who have tried gambling and written about their failures or successes?

I rarely follow the NFL, but one year I was in a pool where you
a - select the outright winner
b - slot them from 14 down to 1, where you get the number of points in the slot if you win

Like I said, I rarely follow the NFL, so I decided to just go with the spread.  The higher the spread, the higher the slot.  Only on pick-em’s did I put in my opinion.  I ended up winning.  I would never win any single week, but after 16 weeks, people started dropping off, leaving me (or rather, the system) at the top.

Mick, if you can tell me what the winning record is for favorites (outright win) in 2006 (through week 12), that’d be good.  I would guess it would be somewhere in the 106-109 wins.

I couldn’t agree more with MGL in #1.

I’ve done “confidence” pools like tango described as well, and his strategy works well as long as there aren’t too many players in the pool.  Another strategy would be to try to win each week by flipping the high (14) pick (bet the largest underdog), but picking the rest according to form.  Most of the time you’ll lose, but eventually that big upset will happen and you’ll almost certainly win that week.  Unless someone else had the same idea.

In the NFL, favorites won about 66% of games from 1998-2003, but it obviously varies by point spread (78% on 7+ point lines, 55% on 3 or less).

This year must have alot of upsets, since Accuscore is at .617 (and highest among all the “experts"). The overall 98-03 average is .660 according to post #4, and that’s barely less than the “locks” picks of .684.

You said 55% on the 3 points or less (which is similar to the nonunanimous picks of .566). What’s the winning record on:
a. 4 points or less
b. more than 4 points
?

In my dataset, <=4 points is 59% (n=707), >4 points is 73% (n=773).

I don’t have 2006 stats compiled yet, but if you’re interested I have the raw data as an unparsed text file.

That’d be nice!  Send it to:
tangotiger~@~yahoo~.~com
(remove all ~)

As suspected (sample size caveat of course), .617 is a joke.  If you just picked the faves, you would be at .642 in 190 games (all games that were not “pick") so far in 2006.

Of course that is not fair as I assume that all of the experts had to pick “pick’em” games as well.  If you include pick’em games, and assume a 50% win rate on those, then a person who picks all faves and flips a coin on pick’em games would have picked .541 (there were only 2 pick’em games so far this year, and indeed that is rare for some reason).

So the .617 is indeed a joke.

As a prevous poster said, if you are competing against other players whom you have to beat, and ties are not good, then it may behoove you to NOT pick all of the faves (maybe, as someone said, picking one big dog is the best strategy), especially on a week by week basis (when lots of people are picking all the faves and the chance of a multi-way tie is great).  If it is just highest percentage wins a certain amount of prize money even if there are multiple ties, or there are enough games, maybe even for one whole season, then I suppose picking all the faves is the optimal strategy.

As far as “locks” go, of course in any game where the line is like -3 or higher, all of the so-called experts should be on the same side.  If anyone is picking a dog when the line is -3 or higher, they are either a genius or an idiot, overwhelmingly likely the latter of course.

If the “locks” (every expert is on the same side) are games in which the fave is greater than -3 or something like that (locks SHOULD be games with large point spreads, right?), then .684 is also a joke, right? 

In games lined more than a 3 point spread, the fave won 69.8% of the time in 06.  Greater than 4 points, the fave won 70.3%.  4 or less won 56.4%, again in 2006.

MGL, do you mean through 2005 or 2006?  Through week 12, ESPN is showing 176 games played, not 190.  I’m guessing you are reporting 2005, through week 13?

(And, obviously, your typo of .541 was means to be .641.)

I am definitely reporting 2006 through the last week played.  I’ll have to check again the number of games played.  The other numbers (the percentages) seem about right.

Yes, that .541 should be .641.

No need to check total number of games played.

It is now week 13.  So 12 weeks have been played, that’s 11 games and 1 bye week for each team (as all teams have had their bye weeks).

11 games for each pair of teams (16) is 176 total games played so far in 2006.

I’m sure your point stands, MGL, (as although an excellent one, it should be obvious to everyone already) but it would be nice to see the actual numbers.

I included week 13 games by accident.  Here are the correct numbers:

In 176 games:

All faves (no pick’em) won 64.4%
Faves including half of all (2) pick’em games won 64.2%
Faves of more than 4 points won 72.0%
Faves of more than 3 points won 71.6%
Faves of 4 points or less won 54.7%

It is rather unreal to think that Accuscore would win 108 of 176, while “the market” would win 113 of 176.

And Accuscore has supposed some great simulator engine that works on PBP. 

I mean… really?  Really?  Chalk up yet another in the long list of Wisdom of the Crowd kicking a$$ against smarty-pants-with-high-tech-gizmos.

Maybe I am confused, but how does a great PBP engine help you pick the results of a game before it starts?

OK, maybe in baseball it can.  But in football?  I think it would be currently impossible, based on the variables involved in football, to have the input for such a simulator anywhere near precise enough to do such a thing.  In fact, if I were to build a PBP simulator, the first version would start by using the vegas line as one of the input parameters, and work from there.

Yes, baseball is really the only major sport than can remotely be analyzed via a PBP sim.  I have no idea what they would be doing in the NFL and NBA.  I would also suspect that this company might have no idea either.

Again, I am shocked (not from a morals perspective) that a company like ESPN would be involved with a company that clearly is promoting themselves as an aid gamblers, and one that is likely a scam as well, despite the “nice looking, high-tech” website.

Tango, the differnce between 113 and 108 out of 176 is not even close to being statistically significant (it is like .5 SD), so I don’t think we can say with any certainty at all who is likely to beat whom in the long run.

The clue in the web site that this is a scam organization is their “data claims.” If the historical numbers they claim were true, they and their followers would now or very soon bankrupt every sports book in the world.

Hate to bust your bubble MGL, but StatShark is for real.  You sound disgruntled and are slamming a service which your pretty much know nothing about.  The community is filled with VERY intellgent sports investors that are making a lot of money utilizing the simulations and the StatShark methodology.

I’m a paying member at StatShark and I hate to see you slam a good service when you know very little about it.  The animations you see on ESPN pretty much scratch the surface of what the simulation engine can do.

1) The simulations are not picks, they are tools...just like historical data.  Most users on statshark have their own systems and use the forecasts to narrow down or strengthen certain plays.

2) The numbers are true, members have access to the Daily Line Report which clearly documents all of there plays.

3) Statshark is NOT a tout service.  You never see them posting “stone cold locks” or any other BS.  They don’t even promote sportsbooks on their site.  The numbers pretty much speak for themselves.

4) Paying members get tools that let you get predictions analyze ANY vegas line.  I put in the line from my sportsbook and I get the % chance of covering the spread or the over/under.

5) Just because you can’t build a play by play simulator doesn’t mean it can’t be done.  Details about the simulator are here:  http://statshark.com/About_Us/

Statshark just posted there 2006 NCAAF results.  No scam here…
http://statshark.com/NCAAF/NEWS/NCAAF_2006_Season_Recap/

Supposedly last years MLB was supposedly hot but I wasnt a member at the time.  Check out this article.
http://statshark.com/MLB/NEWS/MLB_2006_Season_Recap/

If you want to join a real site and learn how to handicap games yourself while using statshark forecasts as a tool, come to statshark.

Otherwise listen to the haters and wander message boards aimlessly looking for the next good thing.  These guys are for real, I’d jump onboard before they’re gone.

wildman, I am not against promotion on this site, as long as its verifiable and not some “wild-man” claims.  If you are not a shill, please be so kind as to email me at:
tangotiger
at
yahoo
dot
com

***

Here is a link which people with more knowledge of gambling that I have can look into:
http://statshark.com/images/dlr/dlr2006.xls

The value in D9 is what seems to interest me the most.  If I’m reading it right, the 56.5% represents the % of times that the simulator called for a win on a team actually won. 

That number seems pretty unimpressive, since just betting on the home team would traditionally (I don’t know about 2006) give you a 53-54% chance of winning.  This includes betting against a powerhouse on the road against an also-ran.

no shill:

wildman
75
96
at yahoo.com

are you referring to this past NCAAF season?

According to the article at:

http://statshark.com/NCAAF/NEWS/NCAAF_2006_Season_Recap/

Game winners were 75.8% and ATS was 59%.

Not necessarily here to promote, but to defend the system that I feel was unfairly bashed.

Of course no simulator is going to be right 100% of the time and the system wasn’t designed to give you picks. 

Out of the simulations, trends emerge.  For example the simulation engine may have a great record for Kansas State point spread forecasts.  I know little about Kansas State football but I’ll definitely look into it if StatShark has a high accuracy rate.

Conversely, statshark may have a horrible record for accuracy on game winners for Florida State.  In that case it might be a good to go against the statshark prediction.

Its a valuable tool that needs to be tried to be understood.

That is insane.  I don’t have a dog in this fight, not being a gambler, but you will never have the sample size to say something like this system is good for Kansas State but bad for Florida.  All it means is they got lucky on one school more than the other.

Even if there was something to it, long before you have enough of a sample to know which school they can pick correctly, the team’s roster will turn over 100%.

Thanks for validating yourself.

If you click on the link I provided from stats shark, it shows that it’s from MLB.

No simulator in any domain of anything you can think of will ever give you 100% forecast.  It should go without saying.

The question is: if you bet the favorite in MLB, what percentage of the time would the favorite win?  As a guess, it seems to me that 56.5% would be about right.  It should be easy enough to figure out.

According the data in the link, the simulator predicted the correct winner 56.5% of the time.

So, I’m not excited over here.

I don’t understand anything else in that spreadsheet, so if you can walk through an example, using the link I provided, feel free.

It is possible that if you “roll your money over”, that you will end up ahead, if you started off particularly lucky.  That is, rather than betting an equal amount each time, you bet an equal amount, plus a share of any profits you have.  If you got lucky to be ahead early, then that’s extra money that you get to parlay.

So, to be fair, I’d first say: assume the betting was an equal amount for each game.

I wasn’t in baseball for last season so I haven’t looked in the daily line report in full detail, but I’ll have a go at it.

In the spreadsheet it says the total overall accuracy is 56.5%, but that is over all of the games. 

Each game is given a percent chance of winning.  According to the spreadshett, when the forecast had the home team winning over 70%, the simulation was 66.2% correct. 

When the simulation had the home team winning 60-69% of the time the simulator was 63% correct.  Naturally as the percentage of simulations won by a team goes down, the accuracy will go down also.

The basic premise is, go with the plays in which the simulator provides a confident prediction(maybe 55% would be a confident prediction) and stay away from the ones that are around 50% (basicall considered a toss up).

Once again, baseball isn’t reall my thing, but I’ll learn more about the strategy closer to the start of the season.  I’ll share my findings if your interested.

Those predictions are without any odds or line.  So, if I always bet on say the Yanks, I’d be right 60% of the time.  That doesn’t mean I should bet on them, since the odds might have been too much to overcome.

My basic point is that the favorites probably win 56.5% of the time in baseball.  (I don’t know what the number is.) If that’s the case, then their sim doesn’t help me, since they forecasted the winner 56.5% of the time.