Projections

mathfuture  6/22/2015 11:25 AM

Vegas has CHC projected to score 4.13 runs vs. LAD tonight, this site projects Kershaw to give up 1.5 ER in 7.1 IP so you're telling me that the dodgers bullpen is supposed to give up 2.63 runs in 1.2 IP?? I understand slight differences but stuff like this really makes me scratch my head

ken_admin  6/22/2015 11:56 AM

We likely do project Kershaw higher than most sites. His ERA to-date is 3.29 which admittedly is more in line with today's odds (after unearned runs are removed).  I see there are heavy winds blowing out in Chicago today, and I'll do some checking to see if that's affecting the odds.  I can also check the factors since he's getting a boost today likely due to Chicago's higher strikeout rate.

descender  6/22/2015 11:59 AM

Where are you seeing that the Cubs are projected to score 4.13 runs?  The sites I frequent only show to total O/U line for the matchup (currently 9).

ehrenberggreg  6/22/2015 1:18 PM

I'm seeing the Cubs over under at 3.5 what websites has their line over four I would love to bet the under on that 

Cameron1Turner  6/22/2015 1:50 PM

Yeah no way Kershaw should be projected to score 8 more DK points than anyone else. Winds blowing out and Cubs are projected to score 3.9 runs and the Dodgers 5.1 for a total of 9

mathfuture  6/22/2015 2:08 PM

https://rotogrinders.com/schedules/mlb?

descender  6/22/2015 4:01 PM

Those are actually rotogrinders projected runs, not Vegas. Not sure that changes the crux of the issue but you'd basically be trusting that one projected source over another. I wouldn't put much faith in the cubs scoring 4 runs tonight personally. More likely the game will be 6-2 Dodgers than 4-4. The extra runs should be coming from the Chicago bullpen with Wada only projected for 5 innings.

mathfuture  6/22/2015 4:09 PM

It's not rotogrinders projecting runs, they're posting the pythagorean expectation using the vegas over under and run line.

At LAD -184, CHC +169

LAD win rate = 184/284 = 64.8%

CHC win rate = 100/269 = 37.2%

64.8 + 47.2 = 102%

LAD no vig win rate = 64.8%/1.02 = 63.5%

CHC no vig win rate = 37.2%/1.02 = 36.5%

Pythagorean win loss says win% = (runs scored)^2/(runs scored ^2+ runs allowed^2)

For CHC runs scored we have .365 = (x)^2/(x^2 + (9-x^2)

which simplifies to .265x^2 +6.57x -29.565 = 0

apply quadratic formula and we have x = 3.9

rotogrinders is higher likely because they use a lower exponent

descender  6/22/2015 5:38 PM

I am genuinely curious here, so don't take any of this the wrong way. :)

It is certainly an interesting way to reverse engineer the Vegas line, though I have my doubts about it's accuracy and usefulness.  I wasn't aware that people would (could?) actually apply the pythag win expectancy formula in that manner.  Have they produced any data that show how accurate this application of the formula is?  While Pythag win expectency generally has a 90% correlation or better with actual wins, some data I mined up from their forums suggested only a 60.4% accuracy on projected runs scored in 2014.  That's neither good nor useful.

To wheel back on topic for a second, Kershaw is projected to give up 2.5 runs by BBM (meaning 2 OR 3, obviously it can't be a half) and Vegas expects 1.5 more runs (meaning 1 OR 2) than that to be scored.  That seems to be right in line with what your Vegas line is telling you.   3 ER between Kershaw and the bullpen, with a possible 4th run from "somewhere". Happy hour + math = wrong.

Improper application of formulas to make predictions should very obviously leaving you scratching your head :p  That's kind of what I see going on here.

mathfuture  6/22/2015 7:20 PM

You are basically saying x correlates to y but y does not correlate to x which is impossible. I can say without a shred of doubt this is a correct application of the expectation formula. You are literally just plugging in the variables and doing solving for an unknown. Furthermore what does 60.4% accuracy mean? I haven't seen any correlation analysis with results in terms of "accuracy".

And regarding the Kershaw projection I don't know what you're looking at but I see BBM projects Kershaw to give up 1.6 ER (it was 1.5 earlier in the day).

ehrenberggreg  6/22/2015 8:10 PM

Although with this said vegas had the Cubs run line set at 3.5 with juice on the under 

descender  6/23/2015 8:26 AM

I would love to see any research done on this subject (application of ptyhag win expectency to vegas odds lines), I can't find any.

It is, as you say... mind-numbingly obvious application of the formula... I have a hard time believing something so obviously useful would only be utilized by one by 1 major fantasy site.

mathfuture  6/23/2015 10:03 AM

Dude you are just plugging in variables into a proven formula, why do you need validation from fantasy sites for you to believe it works? - it is, as you say, common sense... furthermore, i even did the math in a previous post, if you can find a logical inconsistency please let me know. I think the one caveat to this methodology is that vegas' goal isn't to exactly project the outcome it is simply to balance the action on their books, but that is still an issue with the source data not the methodology.

Cameron1Turner  6/23/2015 11:53 AM

Jonathan Bales and CSURAM's site projected 3.9 runs for the Cubs and they scored 4. I don't know what they use, but it's been pretty accurate. Plus it gets updated way more often than Rotogrinders

descender  6/23/2015 12:22 PM

There is really nothing wrong with the methodology here, the source data is exactly what I am taking issue with (sorry if that hasn't been clear).  They really need to show some work that proves the way Vegas sets their line weights actually correlates to the win% the way you are using it in the formula.  Remember their goal is actually to get you to place a bet, not to give you an accurate prediction.

Something tells me that the underlying correlation between Vegas run line and expected wins isn't very strong.  This would easily explain why so few sites use this method to projected daily runs scored.  I'd love to be wrong about that but without someone doing (and showing) the dirty work... it's tough to come to any other conclusion.  If there was a strong correlation I don't think they would be shy about it either, and there is almost no information about the method on the Rotogrinders website.

mathfuture  6/23/2015 12:55 PM

Vegas run lines and money lines are set (initially by oddsmakers but I mean the final lines) according to what the public perception of the runs scored and win chance of each team are. We are just adding one more step to that process and extrapolating the run expectancy based on those two factors. To reiterate, we are not inferring any correlation here, we are simply taken the data we are given and manipulating it to produce another data point. The only potential source of error here is that although public expectancy is strongly correlated with outcomes it is obviously not perfect.