Irisheyes\' MLB Overnight Robot
 

YTD 0-0 +0u

Well, it's been quiet here on the MLB front for a little while. And as I'm forcibly weaning myself from Crockpot's picks *sniff* and with great thanks and respect to the posters here I've designed a model that I think is different than TomG (and also unfortunately different than Crockpot). In no way do I claim it is better, and as this is a new system--fade/follow at your own risk. Some of you may recall I started this season blindly fading the Red Sox. Caveat bettor.

What is different about it? Weighted lineups, attention to DIPS/sabermetric stats, and blending projections (PECOTA, etc.) and season to date stats.

I've been winning with it but given the ridiculously small sample size I could just as easily prove the world is flat.

Clearly, I will get crushed.

6/28
PIT@FLA
FLA -156
Risk 3.12 units to win 2

CIN@PHI
CIN +142
Risk 1 unit to win 1.42

STL@NYM
STL +167
Risk 1 unit to win 1.67

COL@HOU
COL +158
Risk 1 unit to win 1.58

CWS@TB
TB -129
Risk 1.29 units to win 1

Good luck.

Re: Irisheyes\' MLB Overnight Robot
 

Good luck. I'm actually heading down a similar path myself in regards to blending in more BaseballProspectus stats and creating my own model. When I have things updated I will bump my thread with picks. Long live Robot betting! [img]/images/graemlins/smile.gif[/img]

Re: Irisheyes\' MLB Overnight Robot
 

Some brief reasoning behind the picks.

PIT @FLA: The fish have the edge in starter, bullpen, offense, and defense. And they're at home.

CIN @PHI: In my opinion, the most marginal of my plays today. Belisle trumps Eaton, but the Reds lag the Phils in run production. Coin toss on a neutral field, and if so the -152 is unjustified.

STL @NYM: Value in Wainwright. He's coming back to form.

COL @HOU: Close value play on COL. Hopefully the National League's worst bullpen gets a chance to give up a few late runs once Oswalt is done.

CWS @TB: Battle of the league-worst bullpens. Vazquez is overperforming and Kazmir is underperforming. I believe the line reflects that, and I think the line is wrong. Same basic logic as the FLA pick above.

Re: Irisheyes\' MLB Overnight Robot
 

What's your winrate been like so far? I ask because I have basically an identical system set up though it's not for gambling purposes. I'm thinking it might be time to repurpose it.

I've also got a couple of little twists in it that I'm not sure anyone else is using.

Re: Irisheyes\' MLB Overnight Robot
 

Tom, I'm a big fan of your thread and I owe you a debt of gratitude. Looking forward to your continued posting. It will be quite interesting to see where and if we differ.

Re: Irisheyes\' MLB Overnight Robot
 

Justin,

you've basically hit on the key to why I started this thread. I have not been religious in keeping my own system bets separate from picks from TomG, Crock, and other sources.

YTD, I'm 169-153 betting 62% underdogs/38% favorites. 85%+ of these picks were sourced elsewhere. Clearly it's been a good year, but I'll never know if it's truly sustainable unless this model proves out.

Re: Irisheyes\' MLB Overnight Robot
 

Can you explain why you risk x units to win 1 when betting on a favorite and not x units to win 1 when betting on an underdog?

I think this has always been the case with Crock (or has it) and I've never done it, which makes me feel like a square.

Re: Irisheyes\' MLB Overnight Robot
 

Crock does the same thing as he just explained on the blog. My reasoning is a bit different, and might be flawed.

I do it so that the average win/loss is 1u. I should win more often with favorites, but that is canceled out the costlier losses--so it should average out to 1 unit per win. Similarly with underdogs I lose more frequently, but when I win it hopefully makes up for the additional losing events -- again averaging at or above 1 unit per win.

Re: Irisheyes\' MLB Overnight Robot
 

good luck with it

Re: Irisheyes\' MLB Overnight Robot
 

6/28 Results 1-3

YTD 1-3 (-1.29u)

Tough losses. Two walk-offs. Them's the brakes.

Re: Irisheyes\' MLB Overnight Robot
 

6/29 -- Two plays today, both in NL

MIL@CHC
MIL +114
Risk 1 unit to win 1.14

SD@LAD
LAD +108
Risk 1 unit to win 1.08

Good luck.

Re: Irisheyes\' MLB Overnight Robot
 

I also like both LAD and Mil... but I was reluctant to bet Mil b/c Yovani Gallardo is still an unknown.

Re: Irisheyes\' MLB Overnight Robot
 

I'm curious what in your algorithm favors LA. Offense? There's not much data on Kuo; Young's stats have improved every year and he has a nearly .666 career winning %. Do you count pitcher win % at all or is that just considered a loosely correlated byproduct of a pitcher's ability.

Re: Irisheyes\' MLB Overnight Robot
 

A pitcher's W/L record is a meaningless stat. The sportsbooks have the wrong team favored IMO. Kuo hasn't played as much as Young but his peripherial stats thus far indicate he is as good if not a better pitcher than Young (Kuo's 105 strike-outs in 92 innings pitched is very impressive). But I'm willing to call pitching a push and let LAD be favored based on only SD's piss poor hitting and home-field advantage.

Re: Irisheyes\' MLB Overnight Robot
 

I just went through all of my numbers and I'm in agreement with you for the Dodgers. My system has it as a really bigplay. Apparently, it likes what it sees in Kuo. I'll be interested to see how it holds up tonight.

Re: Irisheyes\' MLB Overnight Robot
 

Frau, with the original formula I learned first in TomG's thread from the Michael Murray book (Betting Baseball 2007) the Dodgers have nearly a 5% edge in batting. The basic calc is OBP x SLG.

Re: Irisheyes\' MLB Overnight Robot
 

Tom,

even though Gallardo is clearly unproven, I gave him the benefit of the doubt by setting him equal to Hill and the Brewers still had an edge. Even so, I think this guy is the real deal.

I didn't see the end of the game, and find it hilarious that three of my first five public bets with this system have been decided by walk-offs in the 9th inning.

What's that smell? Mmmm.... variance.

Re: Irisheyes\' MLB Overnight Robot
 

Josh,

agree with you. Kuo doesn't appear to suck as bad as his ERA would have us believe. Sample size too small to really say, but I felt it was worth a small gambool [img]/images/graemlins/smile.gif[/img]

Re: Irisheyes\' MLB Overnight Robot
 

It seems I have a taste for betting games with multi-run 9th inning rallies (the wrong kind).

6/29 Results 0-2

YTD 1-5 (-3.29u)

Re: Irisheyes\' MLB Overnight Robot
 

Barring some significant line moves, there doesn't look to be anything interesting on tomorrow's NL card. My system is still far too manual for my taste, and I'll probably look at the AL matchups in the morning.

Re: Irisheyes\' MLB Overnight Robot
 

[ QUOTE ]
Josh,

agree with you. Kuo doesn't appear to suck as bad as his ERA would have us believe. Sample size too small to really say, but I felt it was worth a small gambool [img]/images/graemlins/smile.gif[/img]

[/ QUOTE ]

I think if their is an object lesson here it is that extrapolating based on small sample sizes is risky. My own analysis was that Young's stats have the appearance of "accelerating"(consistently improving over his 3 seasons) thus making overweighting his recent performance appropriate and making him a decent bet over an unknown quantity. Unfortunately I allowed myself to be swayed by you guys and decided to lay off the game instead of betting Young. But hey, it's just one game and maybe you just got unlucky for the 5th time in 6 games.

Re: Irisheyes\' MLB Overnight Robot
 

IMO, CY is "accelerating" because he moved from pitching in Rangers Ballpark to Petco. I could pitch in Petco and have a respectable ERA. FWIW, the Dodgers did have 14 hits to San Diego's 9, but Chris had a very nice outing and Kuo obviously did not.

It's a long season, and while 1 for 6 sucks, it's happened before and will happen again. If you believe there's value in Young next time he pitches, I suggest you go for it. I would wager there will be plenty of people here who will line up to bet the other side.

Re: Irisheyes\' MLB Overnight Robot
 

YTD 1-5 (-3.29u)

Only going to have time to do the NL this weekend. One play tomorrow:

MIL@CHC
MIL +102
Risk 1.5 units to win 1.53

Good luck.

Re: Irisheyes\' MLB Overnight Robot
 

Well, I'm sure everyone has been hitting 'refresh' every few minutes to make sure they haven't missed my latest pick. Great fade material here, it would seem.

7/1 Results 0-1

YTD 1-6 (-4.79u)

Re: Irisheyes\' MLB Overnight Robot
 

Rough start but don't give up too soon. Give yourself a chance to let the other side of variance hit you.

Re: Irisheyes\' MLB Overnight Robot
 

1-6 is nothing! Be happy you're not 1-15 or some nonsense like that (yes, it will happen).

Re: Irisheyes\' MLB Overnight Robot
 

Thanks Tom & rjp,

7/2 YTD 1-6 (-4.79u)

PHI@HOU
PHI -113
Risk 1.13u to win 1.00

Prepare to ship it. Good luck.

Re: Irisheyes\' MLB Overnight Robot
 

Wow, you guys are quick. I barely get this posted and already it's been bet down to -115. That will teach me to post before I bet!

Amended pick:

PHI@HOU
PHI -115
Risk 1.73 to win 1.50

With so many followers, how can I miss [img]/images/graemlins/wink.gif[/img] ?

Re: Irisheyes\' MLB Overnight Robot
 

On it - thanks Irish [img]/images/graemlins/smile.gif[/img]

Re: Irisheyes\' MLB Overnight Robot
 

[ QUOTE ]
Wow, you guys are quick. I barely get this posted and already it's been bet down to -115. That will teach me to post before I bet!

Amended pick:

PHI@HOU
PHI -115
Risk 1.73 to win 1.50

With so many followers, how can I miss [img]/images/graemlins/wink.gif[/img] ?

[/ QUOTE ]

6am EST: line is down to -110 at WSX; (-112 at pinnacle). Agree that this a strong pick.

Re: Irisheyes\' MLB Overnight Robot
 

7/2 Results 0-1

YTD 1-7 (-6.52u)

Re: Irisheyes\' MLB Overnight Robot
 

In case anyone thought otherwise, I was kidding of course about the line move and having followers. I wouldn't want anyone to follow where I've been for the past week.

I need some good picks to pull me out of this funk, and I just don't see much of value on tomorrow's card. Is Zambrano worth -190 (Pinny) on the road? Gut says NFW. But that line isn't up for me yet because of the pitching change, so we'll see tomorrow.

I also lean towards fading Randy Johnson, but don't like the Redbirds enough at +124. I need a break, and I have a feeling the lines will be much more friendly on the Fourth of July.

Good luck, and for those of us stateside--enjoy the holiday.

Re: Irisheyes\' MLB Overnight Robot
 

irisheyes, what's interesting about these various methods of handicapping is how people can end up on different sides of games based on what information the choose to use. I've done pretty well in baseball using some of the methods described in murray's book but I was often times on different sides of games than TOMG. For the batters I've been using this season, last season, and the season before that. However if a player is getting worse due to age i make some minor adjustments.

For instance tonight my line on philly was -108. I would've made a play on houston at +116 or so but I wasn't at my computer before game time. I've spent a little time now focusing on totals as well as team totals and have done okay. Don't worry about this rough posting patch as you said that you've been on a roll so you know that eventually you regress back to where you should be.

good luck man. you irish i guess? I am too.

Re: Irisheyes\' MLB Overnight Robot
 

newb,

thanks for the encouragement. I feel like my weighting of YTD versus regressed/projected stats is off--I believe I am placing too much weight on YTD. An interesting thing to note is that(in my limited experience) YTD appears to be a better predictor of the book's line than projections or previous year's stats (at least at this point in the season).

You may want to check out the various projection models (Pecota/ZiPS/etc.), since they take regression and player age modeling to a level I would never be willing to do with my current mashup of Excel and Java.

If I didn't misunderstand you, if you calc'd the true line at -108 for Philly, you'd bet Houston at +116? If so, that's only a 1.8% edge. I don't usually push any edges less than 5%. If you meant true -108 for HOU getting +116, that's a good 5.6% edge.

As far as totals, I'm pretty much terrified to trust my Murray-based model for that. My initial attempts have not been stellar, and there are a lot of manual factors (weather, umps, etc.) that come into play.

What I really need the time off for is to automate this. I'm moving from Excel to Java/JSP because I'm spending upwards of 3-4 hours/day handicapping.

Anyone know of any good sources of 2007 stats in database/machine readable format? I already have stats through 2006. I really don't want to rely on screen-scraping (web-scraping actually) to pull current stats.

Good luck to you as well. If I post some lines that look errant to you, by all means jump in. It will help me look for flaws or *gasp* improvements in the model.

And yes, I am a mick, but lately me luck's run a little dry [img]/images/graemlins/wink.gif[/img]

Re: Irisheyes\' MLB Overnight Robot
 

[ QUOTE ]


An interesting thing to note is that(in my limited experience) YTD appears to be a better predictor of the book's line than projections or previous year's stats (at least at this point in the season).

[/ QUOTE ]

I think this is exactly what makes betting on baseball possible. Books know that a large % of the public bases it's bets on YTD performances and creates lines accordingly. We know that a better predictive tool uses some sort of rolling average of stats for a period of years (with, of course exceptions made for injuries), so we can discount current hotness or coldness and find errors in the lines. Of course, some players do dramatically improve or decline, but I don't know of any way to know for sure who is a statistical anomaly and who has had a real change in ability. There might be some value in charting trend lines.

I think betting gets harder as the season progresses because YTD stats begin to regress to the mean as time progresses. It seems that early in a season, you get a lot of easy pickings where the betting public will get behind a pitcher who has had a couple of quality starts in a row and a YTD low ERA, despite a much higher historical ERA. Milwaukee's Chris Capuano is/was a good example of this phenomenon this year.

Re: Irisheyes\' MLB Overnight Robot
 

irish, the reason I chose to go back three season was mainly because last season was somewhat of an anamoly with the high scoring which most likely had something to do with pitchers not taking steroids anymore. But I'm sure using something like Pecota would be far superior as it probably has more predictive validity than what i do.

the philly example was philly -108 for me, I've noticed that the small edge bets have done quite well which obviously means that my line isn't entirely accurate but I think in baseball with all the various factors it's really hard to produce an exact line, which is why i just flat bet these games. When I bet hockey I have more faith in my line and use kelly criterion.

that would be a good idea to automate the process through something like Java.. I've never programmed before so i'd have no idea what to do. For the 2007 stats I just use a web query in excel but the tedious thing for me is getting the pitchers era and then the bullpens era as the players available switch from day to day. I'm not especially skilled in excel either, a lot of what i do is just formulas and the occasional macro with the recorder.

For the pitching I've been using this season as well as the previous two seasons. I think tomg said he used that much data and I found it better than using just the current and previous season.

Re: Irisheyes\' MLB Overnight Robot
 

mess with VLOOKUP in excel. it will automate a bunch of it.