Win Predictor Fun

[Guest Edgy DC]

You all saw the home opener and the see-saw battle that ultimately sawed decisevely in the Mets favor. What was the big blow --- the offensive play that advanced the Mets cause most effectively?

For definition's sake, I used the win expectance finder to calculate the value of the various Mets' offensive plays.

[Centerfield]

I don't know about "big blow", but the one play that made me shift from pessimistic to optimistic was the walk to Old Man Franco.

[Guest Edgy DC]

Good choice. That was the third-most-damaging offensive play by a Met, advancing their win probablility (based on 2001-2006 game data) by 19.2 percentage points, from 38.2% to 57.4%.

One reason why pinch-hitters matter. They may only get 20 hits in a season, but they're in high-leverage spots. Go old guy.

[TransMonk]

Yup, I agree with CF...I was expecting an inning ending DP from Julio.

[Guest Edgy DC]

Ah, monk, you're in new territory. I'm measuring the difference between where a team starts and where a team ends. You're measuring the difference between where a team ends up and where they could have ended up given the worst possible outcome.

[metirish]

That's really cool,and it's based on real data,right?.....Joe Morgan would love it.

[Benjamin Grimm]

How about Wright's double?

[Guest Johnny Dickshot]

Was it Rollins' error?

[Guest holychicken]

Edgy DC wrote: You all saw the home opener and the see-saw battle that ultimately sawed decisevely in the Mets favor. What was the big blow --- the offensive play that advanced the Mets cause most effectively? For definition's sake, I used the win expectance finder to calculate the value of the various Mets' offensive plays.

Do you also use this win expectance finder to determine your PoTG points?

[Guest Edgy DC]

Yancy Street Gang wrote: How about Wright's double?

Wright's double was vicious, but the Mets were already up by a few and it was late. It advanced the likelihood of a Mets win 3.3 points, from 94.3 %to 97.6%. The game wasn't in bed, but the lights were out and the covers were pulled.

="Johnny Dickshot"]Was it Rollins' error?

Very very close. Reyes' conspiring hit to Mr. Team-to-Beat pushed our scheme ahead by 22.8 points, from 57.4% to 80.2%.

Rollins might have realized that, though his team led in score, they trailed in probablility. Had Rollins relaxed and gotten the runner at second, the Mets would certainly have advanced less, from 57.4% to 72.5%. The double play would have cut the Mets chances all the way down to 14.6%. So was it worth the gamble to rush the play going for two? I dunno, depends on what you thought of Reyes' running that one out. Frayed Knot can help us here.

But there was one other play that pushed the Flushing Nine cause ahead by 22.9 points. What was it?

="holychicken"]Do you also use this win expectance finder to determine your PoTG points?

No, but I'd consider it.

[Guest Edgy DC]

By the way, I'm taking 2001-2006 data.

[metirish]

The wild pitch by Geary?

[Guest Edgy DC]

That brought in the go-ahead run, but the likeliness that it was going to be brought in one way or another at that point apparently supressed the overall statistical impact. It advanced the marker 12.1 points, from 80.2% to 92.3%.

I thought Barajas should have partial blame on that one. It wasn't that unplayable.

[metirish]

This is fun,it must be the sac-fly by Beltran.

[Guest Edgy DC]

Two guesses is your limit. You're driving.

[Guest Johnny Dickshot]

Whatever LoDuca did after Reyes. Walk, I believe

[Guest Edgy DC]

Most of what Lo Duca's walk would have accomplished was already accomplished in the wild pitch. The walk only advanced the Mets 1.8 points, from 92.1% to 93.9%.

[Guest Johnny Dickshot]

Yeah, that's why it wasn't my first guess.

[Guest Edgy DC]

Two things that the predictor has taught me have grossly over-rated value, that should be evident but aren't always.

1. Sacrifice bunts
2. Walks behind runners on second and third.

[Benjamin Grimm]

See the next page...

[Benjamin Grimm]

Here's the inning:

- M. Alou singled to deep right center
- S. Green singled to left, M. Alou to second
- E. Chavez ran for S. Green
- J. Valentin grounded into fielder's choice, M. Alou out at third, E. Chavez to second
- J. Franco hit for P. Feliciano
- J. Franco walked, E. Chavez to third, J. Valentin to second

Edgy DC wrote: That was the third-most-damaging offensive play by a Met, advancing their win probablility (based on 2001-2006 game data) by 19.2 percentage points, from 38.2% to 57.4%.

- J. Reyes reached on fielder's choice, E. Chavez scored, J. Valentin to third, J. Franco to second on J. Rollins' fielding error

Edgy DC wrote: Reyes' conspiring hit to Mr. Team-to-Beat pushed our scheme ahead by 22.8 points, from 57.4% to 80.2%.

- J. Valentin scored, J. Franco to third, J. Reyes to second on wild pitch

Edgy DC wrote: That brought in the go-ahead run, but the likeliness that it was going to be brought in one way or another at that point apparently supressed the overall statistical impact. It advanced the marker 12.1 points, from 80.2% to 92.3%.

- P. Lo Duca walked

Edgy DC wrote: Most of what Lo Duca's walk would have accomplished was already accomplished in the wild pitch. The walk only advanced the Mets 1.8 points, from 92.1% to 93.9%.

- J. Lieber relieved J. Lieber
- C. Beltran hit sacrifice fly to center, J. Franco scored
- C. Delgado walked, J. Reyes to third, P. Lo Duca to second
- D. Wright doubled to deep left, P. Lo Duca and J. Reyes scored, C. Delgado to third

Edgy DC wrote: Wright's double was vicious, but the Mets were already up by a few and it was late. It advanced the likelihood of a Mets win 3.3 points, from 94.3 %to 97.6%. The game wasn't in bed, but the lights were out and the covers were pulled.

- M. Alou singled to left, C. Delgado and D. Wright scored
- E. Chavez grounded out to shortstop


So the big leap had to have come early in the inning.

Interesting.

My second (and final) guess: Green's single?

[metirish]

I think we found a new game to entertain us.

[Guest Edgy DC]

No to Green's single and no to Beltran's sac fly.

It's 4:42, so I'll lay the numbers out now. Two notes before I do:

1. We shouldn't be arguing about leaving Burgos' in to pitch. We maybe should be arguing about bunting Valentin. Getting the runner at third there was huge for the Phillies (not huge enough, in the end, from 78.0% to 38.2%.
   
   It's the sort of play that makes one (again) appreciate Keith Hernandez.
   
2. Howard's homer nonetheless cut the Mets win expectancy from 65.2% to 28.7%. Ouch.

Here we go.

Delgado's second-inning single: .584 to .623 = .039

Maine's third-inning single: .450 to .478 = .028

Reyes's third-inning single: .478 to .498 = .020

Beltran's fourth-inning single: .468 to .523 = .055

Delgado's fourth-inning single: .523 to .562 = .039

Green's fourth-inning walk: .439 to .429 = -.020

Valentin's fourth-inning two-RBI two-out single, swinging a 1-0 deficit into a 2-1 lead with two out: .439 to .668 = .229

Reyes's fifth-inning reach on a 2-base error: .598 to .672 = .074

Lo Duca's fifth-inning "productive out," moving Reyes to third: .672 to .637 = -.035

Beltran's fifth-inning walk: .637 to .626 = -.011

Delgado's fifth-inning sacrifice fly: .626 to .720 = .094

Beltran's fifth-inning caught stealing: .720 to .688 = -.032

Alou's sixth-inning walk: .248 to .286 = .038

Lo Duca's seventh-inning single: .162 to .178 = .016

Beltran's seventh-inning walk: .178 to .235 = .057

Delgado's seventh-inning RBI single: .235 to .301 = .066

Alou's eighth-inning leadoff single: .295 to .419 = .124

Green's eighth-inning single: .419 to .580 = .161

Franco's eighth-inning four-pitch walk: .382 to .574 = .192

Reyes's eighth-inning tie-run-scoring grounder to Jimmy Rollins who misplays it: .574 to .802 = .228

Lo Duca's standing there as Geary wild-pitches in the go-ahead run: .802 to .923 = .121

Lo Duca's eighth-inning walk: .921 to .939 = .018

Beltran's eighth-inning RBI sacrifice fly: .939 to .979 = .040

Delgado's eighth-inning walk: .979 to .943 = -.036

Wright's eighth-inning two-run double: .943 to .976 = .033

Alou's eighth-inning two-run single: .976 to 1.000 = .034

[A Boy Named Seo]

From FanGraphs.com:



The play log has win expectancy for ever batter of the game, plus some other neat stuff.

Edit: Got rid of the graph. I don't want to fuck up the game. You can still get it here.

[nymr83]

cool toy.
Rollins' leadoff homer dropped the Mets chances from.540 to .440, going with only 2000-2006 it went from .539 to .431
I'm suprised. that means that in the past few seasons a lead-off homer is slightly MORE likely to hold up than it was in the past, i'd think the greater frequency of runs scored would make the opposit true. i guess the sample size is a bit small, i'll have to play with it later

[nymr83]

edit- wrong thread

[Guest holychicken]

A Boy Named Seo wrote: From FanGraphs.com: The play log has win expectancy for ever batter of the game, plus some other neat stuff. Edit: Got rid of the graph. I don't want to fuck up the game. You can still get it here.

Anyone a programmer (besides me) who is motivated (probably not me)?

I think it would be interesting to use this tool to grab the data from every game and calculate which players contribute the most to increasing (and decreasing) our probability of winning over the season.

As I mentioned to edgy in another thread, this might be a interesting way to generate another "PoTG" type of ranking.

Any takers? Or am I the only programmer here?

[Guest Edgy DC]

I think, and have always thought, it's an excellent way to compare players and generate MVP candidates and such.

The thing is that a tool like this can both (1) iron out clutch performances over the long haul, and (2) reward you with credit (and no more or less than you actually deserve) when you actually are clutch.

It's also a good way to see if late inning relievers and pinch-hitters actually belong where we put them, value-wise. They give you fewer reps but in high-leverage situations, and this system measures the leverage.

By this system, in a typical year, a team with the bases loaded, down by three, with two outs in the bottom of the ninth, has a 10% chance of winning. A player who hits a grand slam in such a situation would get credit for .90 win shares. If he makes the final out, he gets -.10 win shares.

Of course, in order to get a keep a bunch of people from hovering around zero, you'd still want a computation to distinguish his performance from that of a replacement player.

[Benjamin Grimm]

holychicken wrote: ="A Boy Named Seo"]From FanGraphs.com: The play log has win expectancy for ever batter of the game, plus some other neat stuff. Edit: Got rid of the graph. I don't want to fuck up the game. You can still get it here. Anyone a programmer (besides me) who is motivated (probably not me)? I think it would be interesting to use this tool to grab the data from every game and calculate which players contribute the most to increasing (and decreasing) our probability of winning over the season. As I mentioned to edgy in another thread, this might be a interesting way to generate another "PoTG" type of ranking. Any takers? Or am I the only programmer here?

I was thinking the same thing.

I might be willing to program a spreadsheet, but I won't take custody of tracking the results.

[Benjamin Grimm]

Okay, I went ahead and did it. I slapped together a little spreadsheet that counts up the "points" for each player.

Paste the table from this page (this is from April 9) into the spreadsheet, run a macro, and you get the following results.

J Rollins	-0.087
J Maine	-0.094
S Victorino	0.151
C Utley	0.017
J Reyes	0.269
C Hamels	-0.115
P Lo Duca	-0.06
C Beltran	0.089
R Howard	0.319
P Burrell	0.005
C Ruiz	-0.069
A Rowand	0.099
A Nunez	-0.083
C Delgado	0.323
D Wright	-0.238
M Alou	0.05
S Green	0.095
J Valentin	0.164
A Burgos	-0.297
J Smith	0.037
D Easley	-0.04
M Smith	-0.128
G Geary	-0.531
P Feliciano	0.055
J Franco	0.147
J Lieber	-0.081
E Chavez	0
B Wagner	0.003

What's cool about this is that players can get negative points as well as positive points.

If anyone wants to take ownership of this just let me know and I'll send you the spreadsheet. You'd have to come up with your own way of tracking season totals. And you'd also have to filter out the numbers for the opposing players.