the "my computer thinks it's smarter than me" system.

[suicidekings]

So recently I've been working on a computer model for baseball to do most of the number crunching for me. The past seasons game data came from http://www.retrosheet.org/ and other specific player stats came from covers.com and yahoo sports.

The model has been doing well so far, so I thought I'd post the picks it makes for fun. I'm definitely not placing any money on these picks. I'll probably run this until the end of April, until the system craps out badly, or until I get bored with it. Day 1 & 2 were previously posted in the daily CK thread and I've moved the records here.

Day 1: (11-3-1)
MIL, DET, PHI, FLO, BOS, STL, CLE, BAL, TB, TOR, HOU, NYM, COL, LAA, LAD

Day 2: (6-4)
ATL, PIT, CLE, PHI, SD, LAD, TOR, LAA, STL, TB

YTD: 17-7-1 (68.1%)

[shoebox]

good start suicide!!

[suicidekings]

Thanks Shoe.

Friday, April 17:
STL, FLO, ATL, SD, MIL, HOU, COL, SF, NYY, TOR, BAL, TB, KC, MIN, SEA

[yort05q]

good stufff, keep posting daily.. always nice to have a model to lean on at the least.

[suicidekings]

Friday, April 17:
STL, FLO, ATL, SD, MIL, HOU, COL, SF, NYY, TOR, BAL, TB, KC, MIN, SEA

The computer pulled off an 8-7 record on Friday, which I don't feel too bad about, seeing as Friday was a rough day (ATL, HOU...)

YTD: 25-14-1 (63%)


Saturday, April 18:
STL, WAS, ATL, SDN, NYM, HOU, LAD, SF, CLE, TOR, BAL, TB, KC, MIN, DET

[Justin7]

Suicide,

Are you using moneylines on these? It's hard to gauge picks unless you know the moneylines that go with them.

[suicidekings]

Justin:

Yes. The final output comes from comparing my system results with opening moneylines for the games. I haven't been posting them just to keep it simple. I'm sure there's better ways to go about this than I've been using, but I started from scratch, and thought I'd see how it went.

Basically, I went about the problem trying to develop 2 lines for each game that I deem fair. One is based on expected team offense (from retrosheet data, broken down inning by inning, and sorted by home/away games). The other line focuses on pitching, using a weighted average between the the starting pitchers and bullpens for each team for comparison (by ERA). Both of these methods then use Pythagorean expectation to develop the lines, and those lines are compared to each other and to the opening lines for each game to see where they sit. The picks are then taken as the team that has the more favourable position in comparison to the opening lines, with the line from pitching carrying more weight (still fine tuning exactly how much weight each should have).

For example, today's SD/PHI game. I had the pitchers (Hill/Myers + bullpens) lined up as SD/PHI = +108/-108, the team offence as SD/PHI = +185/-185 (based primarily on 2008 offence), and the opening line was +150/-170. The Phillies came out as the team that was more overvalued in the opening lines as a (-170) fave, so the pick for the game was San Diego.

The plan is to eventually superimpose these two into one system equation, however I have other factors I want to incorporate as well (game location, team defence, and more elaborate pitching criteria) before I go any further. I'm not really sure which direction I want to go with it now, and I just got super busy at work, so the time I have to play with it right now is limited... So the for the moment it's just a fun pastime while I think about it. If you have any suggestions for factors I haven't considered, I'd love to hear them.

Edit: The two projected lines are combined prior to comparison to the opening lines for games, but the method of combining them is not fixed as yet (relative weight). Numbers posted below on April 19 onwards will be using a 65-35 weighting, favouring the pitcher criteria. That ratio is just a working estimate based on my opinion. Any second opinions on the weighting from anyone would be welcome.

[suicidekings]

Rolling along...

Saturday, April 18: (9-6)
STL, WAS, ATL, SD, NYM, HOU, LAD, SF, CLE, TOR, BAL, TB, KC, MIN, DET

YTD: 34-20-1 (63%)

[suicidekings]

Sunday, April 19

STL, FLO, PIT, SD, MIL, CIN, COL, ARI, CLE, TOR, BAL, TB, KC, LAA, DET

[j0hnnyv]

please start providing the ML to go along with these selections. thank you.

[suicidekings]

SUNDAY'S BREAKDOWN:

APRIL 19: (5-9-1) (33%)
YTD: (39-29-2) (55.7%)

When I input the opening lines late last night for today's games, I just flat out made mistakes in entering the data for the ATL/PIT, KC/TEX, and COL/LA games. All three of the posted teams lost, and the record will take the three losses due to my errors. These three are shown at the bottom with the corrected lines.

MATCHUP..... OPENING LINE..... MY PROJECTED LINE......PICK

STL/CHN......(+115/-135)..........(-103/+108).......STL
FLO/WAS......(-140/+120)..........(-234/+240).......FLO
SD/PHI......(not available )..........(+105/+123).......SD (based on projection only)
MIL/NYM......(+135/-155)..........(+129/-129).......MIL
CIN/HOU......(-130/+110)..........(-145/+217).......CIN
ARI/SF......(+105/-125)..........(-128/+129).......ARI
CLE/NYY......(+210/-250)..........(+147/-147).......CLE
OAK/TOR......(+105/-125)..........(+233/-232).......TOR
BAL/BOS......(+190/-230)..........(+121/-119).......BAL
CWS/TB......(+145/-165)..........(+192/-186).......TB
LAA/MIN......(+135/-155)..........(+104/-100).......LAA
DET/SEA......(+100/-120)..........(-134/+135).......DET

Corrected lines:
ATL/PIT......(-115/-105)..........(-108/+109).......ATL
KC/TEX......(-110/-110)..........(+112/+104).......TEX
COL/LA......(-110/-110)..........(+119/-112).......LA

So, for example, if you were looking at the ARI/SF game, ARI's projected line is (-128), while the actual line opened at (+105), and SF's projected line was (+129), but their actual line opened at (-125). Choosing ARI gives me more value for money spent, getting the team I feel should be favoured at +money.

I'll post the selections in this format in future days, time permitting.

[suicidekings]

Monday, April 20:

MATCHUP..... OPENING LINE..... MY PROJECTED LINE......PICK

ATL/WAS......(-150/+130)..........(-195/+195).......ATL
SD/PHI......(+135/-155)..........(+210/-208).......PHI
FLO/PIT......(-105/-115)..........(-164/+173).......FLO
CIN/HOU......(+125/-145)..........(+107/-104).......CIN
COL/ARI......(+100/-120)..........(-131/+139).......COL
BAL/BOS......(+170/-200)..........(+123/-121).......BAL
OAK/NYY......(+170/-200)..........(+131/-121).......OAK

[suicidekings]

Ouch. 1-4-2 on Monday. YTD: 40-33-4.

No time for details today...

ATL, SD, FLO, LAD, COL, MIN, OAK, CHC, STL, TOR, CLE, DET, TBA, BAL, PHI

[suicidekings]

APR 21: 6-8-1
YTD: 46-41-5

Not surprisingly, the system's win % is evening out. Time to start applying filters to the picks.

[suicidekings]

Not much time today. here's a couple the system and I agree on:

MATCHUP..... OPENING LINE..... MY PROJECTED LINE......PICK

KC/CLE......(+170/-200)..........(+259/-247).......CLE
SD/SF......(-110/-110)..........(-150/+153).......SD
CWS/BAL......(+105/-125)..........(+103/+101).......CWS