BRF GJP Trial Poll-1: How many seats will NDA win?

Kanson · Post by **Kanson** » 08 Apr 2014 08:39

300+ for BJP n NDA

devesh · Post by **devesh** » 08 Apr 2014 08:43

200-220

didn't realize it was NDA, and not BJP. for NDA, 240-260

archan · Post by **archan** » 08 Apr 2014 09:03

niran wrote:>300, yeall know me, me is cowbelti know nothing probability wability
i go by house flies wishpers

aap ke muh me ghee aur shakkar, woh bhi shuddh. But we are a bit wary and cautious...
what to do, dhoti shiveri folks.

Kanson · Post by **Kanson** » 08 Apr 2014 09:15

Of 40 odd replies here, only me had the vision to give 300+ for BJP. Let's see.

Dilbu · Post by **Dilbu** » 08 Apr 2014 09:23

I have bought a new dhoti and will be shivering frequently. Modi will fail onlee.

Klaus · Post by **Klaus** » 08 Apr 2014 09:34

<200: 40%
200-220: 18%
221-240: 10%
241-260: 5%
261-280: 4%
281-300: 3%
>300: 20%

Manu · Post by **Manu** » 08 Apr 2014 10:51

Sticking my neck out. Below is for BJP ONLY:
1) Less than 200 5%
2) Between 200-220, 10%
3) Between 221-240, 15%
4) Between 241-260, 25%
5) Between 261-280, 40%
6) between 281-300, 5%
7) Greater than 300, 0%

NDA (BJP++) will have 300 + Seats

merlin · Post by **merlin** » 08 Apr 2014 11:15

1) Less than 200 - 5%
2) Between 200-220 - 10%
3) Between 221-240 - 20%
4) Between 241-260 - 30%
5) Between 261-280 - 20%
6) between 281-300 - 10%
7) or greater 300 - 5%

Lilo · Post by **Lilo** » 08 Apr 2014 11:19

For NDA

1) Less than 200 - 5%
2) Between 200-220, - 10%
3) Between 221-240, - 10%
4) Between 241-260, - 15%
5) Between 261-280, - 20%
6) between 281-300 - 25%
7) or greater 300 - 15%

My reason for skew of the graph to right: Momentum effect of the campaign.

rohitv · Post by **rohitv** » 08 Apr 2014 12:09

1) Less than 200 - 10%
2) Between 200-220, - 15%
3) Between 221-240, - 15%
4) Between 241-260, - 15%
5) Between 261-280, - 30%
6) between 281-300 - 10%
7) or greater 300 - 5%
[10%, 15%, 15%, 15%, 30%, 10%, 5%]

member_20317 · Post by **member_20317** » 08 Apr 2014 12:23

Musharraf basis for Pre-Poll NDA:

1) Less than 200-------01%
2) Between 200-220---20%
3) Between 221-240---50%
4) Between 241-260---25%
5) Between 261-280---02%
6) between 281-300---01%
7) or greater 300------00%

panduranghari · Post by **panduranghari** » 08 Apr 2014 12:30

90% between 250-280 for NDA
10% less than 250 or more than 280.

vijaykarthik · Post by **vijaykarthik** » 08 Apr 2014 14:24

Let me do it GJP style... if I am also allowed to make an educated guess here.

1) Less than 200 -- 0%
2) Between 200-220, -- 5%
3) Between 221-240, -- 36%
4) Between 241-260, -- 34%
5) Between 261-280, -- 20%
6) between 281-300 -- 5%
7) or greater 300 -- 0%

http://en.wikipedia.org/wiki/Opinion_po ... nion_polls

Sample size seems to be a worry. Is it that the freaks used just 1k-64k people over all the states? [NDTV Hansa seems to have a higher decentish size and might be more indicative?]

harbans · Post by **harbans** » 08 Apr 2014 14:47

No option for 350 plus?

saadhak · Post by **saadhak** » 08 Apr 2014 14:54

For NDA:
1) Less than 200 -- 5%
2) Between 200-220, -- 10%
3) Between 221-240, -- 10%
4) Between 241-260, -- 15%
5) Between 261-280, -- 20%
6) between 281-300 -- 30%
7) or greater 300 -- 10%

kapilrdave · Post by **kapilrdave** » 08 Apr 2014 15:10

1) Less than 200, 0%
2) Between 200-220, 0%
3) Between 221-240, 10%
4) Between 241-260, 10%
5) Between 261-280, 10%
6) between 281-300, 25%
7) or greater than 300, 45%

I believe in momentum not in prediction. I've seen phenomenal support for NM across the country, be it south or north. My parents did pilgrimage in south in last year. Almost everyone asked them about Modi. That was June last year. Lot of water has passed after it and the tsuNaMo has only increased. The wave will sweep every corner of the country.

merlin · Post by **merlin** » 08 Apr 2014 16:26

kapilrdave wrote:1) Less than 200, 0%
2) Between 200-220, 0%
3) Between 221-240, 10%
4) Between 241-260, 10%
5) Between 261-280, 10%
6) between 281-300, 25%
7) or greater than 300, 45%

I believe in momentum not in prediction. I've seen phenomenal support for NM across the country, be it south or north. My parents did pilgrimage in south in last year. Almost everyone asked them about Modi. That was June last year. Lot of water has passed after it and the tsuNaMo has only increased. The wave will sweep every corner of the country.

Only if people come out and vote. IMO, folks should do the probability percentages only after polls are over and the voter turnout percentage is known. If higher than historical percentage for that state only then momentum theory holds good. I'm still a little skeptical if voter turnout will be huge.

Denis · Post by **Denis** » 08 Apr 2014 16:40

228 is my estimate

member_28502 · Post by **member_28502** » 08 Apr 2014 18:20

I voted number 5
I think they will end up 261 plus for sure

Subject to not sabotaging the noting machines and or machinery

Supratik · Post by **Supratik** » 08 Apr 2014 19:22

1) Less than 200, 5%
2) Between 200-220, 10%
3) Between 221-240, 10%
4) Between 241-260, 20%
5) Between 261-280, 30%
6) between 281-300, 20%
7) or greater than 300, 5%

saip · Post by **saip** » 08 Apr 2014 19:48

10%,20%,50%,10%,10%,0%,0%

Kati · Post by **Kati** » 08 Apr 2014 21:24

Wokkay, here is my probability distribution (earlier, at the beginning of the thread, I voted for only one option
- that is the mode of my probability distribution, which is 241-260 range.)

1) Less than 200 5%
2) Between 200-220, 10%
3) Between 221-240, 25%
4) Between 241-260, 30%
5) Between 261-280, 20%
6) between 281-300, 10%
7) Greater than 300, 0%

Saral · Post by **Saral** » 08 Apr 2014 21:59

Everyone: Please give the 7 percentages for NDA (not BJP alone). So if you just give a lame-ass one estimate, please provided the full-monty of 7 numbers that add to 100% onlee. A little painful, but we need some discipline. I see that we have a TFTA GJP wala on-board.

Harbans: The > 300 covers every scenario up to 543 and beyond. Pls provide 7 percentages for NDA.
Denis: Saar, 7 estimates please. We request your estimate for each range and they add to 100%.
Nijalingappa: Ditto. NDA estimates for all 7 ranges.
panduranghari: Pls give 7 ranges estimates as defined by poll question. If you did that on page 1, then ignore my comment.
Dilbu: Then you need to give the jinx prediction onlee. 7 numbers please.
Kanson: 7 numbers please unless you are saying all ranges below 300 for NDA are all zero onlee. Super Jingo.
Devesh-ji: need NDA numbers, the probability estimate for each range please.

In case it isn't clear, the point of this exercise is to mine BR-Jingo judgment algorithms and patent them worldwide. All contributors can expect a cut in the royalties. Each number you give is the probability (0 to 100) of NDA landing in that range. If you are a super-jingo, then you might give the last range 100% and all the rest as 0%. But all the Good Judgment project data suggests that this is very risky and extreme. So please take this exercise seriously (Dilbu excepted. Jinx incantations allowed) and try to provide estimates that you believe will predict the future. Results will be reported each Friday. You can make a new set of 7 estimates every Monday to Thursday (or even after the Friday results are posted.. Till we know the actual results, the Friday will just summarize how BR sees the future).

Bonus: Please make the judgments for UPA also that is in a separate poll thread. I know it is unpleasant but someone has to do it.

RamaY · Post by **RamaY** » 08 Apr 2014 22:08

Saralji/Ramanaji,

Do we need another poll to collect data for 3rd+4th fronts? It will be fun and also give you the error margin.

Saral · Post by **Saral** » 08 Apr 2014 22:21

RamaY: Let us walk before we can run. We dont want our judges to have premature burnout before reaching escape velocity. We will update results this Friday and if enthu exists, then we can add all this other fronts about whom my knowledge is approaching zero. You can suggest the ranges for them on the GJP thread in GD and we can consider from Week 2.

vijaykarthik · Post by **vijaykarthik** » 08 Apr 2014 22:23

@saral - a few recommendations.

math is difficult when we have 7/9 ranges. make it lesser next time around [GJP has a max of 5 in the ranges]

Are you planning on a brier calculation? Ordered bins are calculated differently.

RamaY · Post by **RamaY** » 08 Apr 2014 22:27

Saral wrote:RamaY: Let us walk before we can run. We dont want our judges to have premature burnout before reaching escape velocity. We will update results this Friday and if enthu exists, then we can add all this other fronts about whom my knowledge is approaching zero. You can suggest the ranges for them on the GJP thread in GD and we can consider from Week 2.

Nope. will go with the flow

Saral · Post by **Saral** » 08 Apr 2014 22:34

vijay: Till such time we have the actual results. we can provide mean, median, mode, entropy (variance like figure) for each rater and also because our raters are small in number we can also provide similarity matrix between raters to see people who think alike. The equivalent of Brier score can be discussed on GJD thread. I don't have any clear insights except that the overall winner should be judged based on all weeks estimates (not just the last one) and the score should be based on absolute mean difference and variance/entropy (minimize both). Note that I learnt of GJP very recently and dont have time this week to read the literature.

I suppose, for entertainment purposes, we can provide a pseudo score every week that assumes that the group aggregate is a proxy for reality.

chaanakya · Post by **chaanakya** » 08 Apr 2014 22:38

Well If it is NDA then it is 300+ (100%)

Despite all permutations and combinations of caste factors notwithstanding and ignoring traditional pollsters' twisted gyrations on this factor. Results will be a big surprise for everyone.

ramana · Post by **ramana** » 08 Apr 2014 22:39

From the poll titles one can see its to get the GJP thinking going in the BRF. I think two trials will give us enough experience.
Also we chose the more complex technique of ordered bins.

Saral, Thanks. Once we get the hang of it we need team to do the math.

Vijay, Please help out if you see any mis-steps.

I already see why we need to limit it to team! It gets too big soon.

Saral and Vijay, My goal is not to pick a winner but to get a prediction right! But I guess we need those stats for the participants to get feedback on how to open their minds which is what we want.

Thanks, ramana

Wiki on Brier sccore:

http://en.wikipedia.org/wiki/Brier_score

Link to original paper by Brier in the refs.

http://docs.lib.noaa.gov/rescue/mwr/078 ... 1-0001.pdf

So Brier score is to rate the participants and not the forecast.
Except for my ego I don't care how accurate I am as long as the process gives me better forecasts from the BRF knowledge pool. Makes sense?

Is it possible to apply Brier score to the combined forecast and thus give a quantitative assessment of the forecast?

Denis · Post by **Denis** » 08 Apr 2014 22:54

OK Saral,

Here are my estimates as ordered:
1) Less than 200 0%
2) Between 200-220, 10%
3) Between 221-240, 60%
4) Between 241-260, 15%
5) Between 261-280, 15%
6) between 281-300, 0%
7) Greater than 300, 0%

Saral · Post by **Saral** » 08 Apr 2014 22:55

Ramana Saar: With all the testosterone on this forum, we need a winner; ultimately the whole process is social and people need to be engaged. I believe the answer is Yes to your last question.

vijaykarthik · Post by **vijaykarthik** » 08 Apr 2014 22:56

FWIW, Brier is done on all individual scores to see how far out they were wrt the actual happening of the event. Say the event that occurs in our eg is 240-260 and someone predicted 100% for that bin, then brier score is 0. [Perfectly calibrated].

Once we get the briers for all the people, we can structure them and make a quantitative assessment based on all individual scores. A brier on a brier doesn't make any sense, AFAIK. [Alternately, we can technically take a combined forecast of all people and make a brier out of it. however, it will have lots of outliers and lumpy probabilities which make the data very noisy and hence will not help, IMO]

ramana · Post by **ramana** » 08 Apr 2014 22:57

OK. But archan will treat them same as everyone!

Prem · Post by **Prem** » 09 Apr 2014 01:05

NDA will get ,IMHO, somewhere 309-322 seats.

ramana · Post by **ramana** » 09 Apr 2014 02:32

Jhujar, How about you give the probabilities to that 7 buckets just like others did?

Kati · Post by **Kati** » 09 Apr 2014 04:47

Saral-ji,
so far we have a sample size over 100. That's fine and good enough. Each one BRFite is giving an e
stimated probability for each interval. The sample may be a little biased since BRF is pro-NDA. That's fine.
Calculate an overall MEAN = [mid point of each subinterval * average relative frequency (%)].
I bet the actual figure will be surprisingly close. We can calculate the standard error (SE) also.
So, a reliable figure: MEAN +/- SE.
To get a better figures, I suggest a finer subinterval system (a total of ten). I finer discretization will be
better.

Shaktimaan · Post by **Shaktimaan** » 09 Apr 2014 09:05

Dear Rakshaks, long time lurker and first time poster.

Here is my probability distribution :

1) Less than 200 : 0%
2) Between 200-220 : 40%
3) Between 221-240 : 40%
4) Between 241-260 : 10%
5) Between 261-280 : 10%
6) between 281-300 : 0%
7) or greater than 300 : 0%

From the general mahoul, I think that the NDA will end up at ~220 seats.

I have to frequently travel to Tier A and B cities of India for my job. From talking to the awaam, I get the sense that the NDA has grown strong in places where it was already present, but the places where it was traditionally weak (TN, Kerala WB etc), it's still not very strong on the ground. The all-important UP is still a mystery to me.

Let's hope that I'm wrong in my pessimism.

Rudradev · Post by **Rudradev** » 09 Apr 2014 10:51

Post EVM magic numbers?

1) Less than 200. 20%
2) 200-220. 55%
3) 221-240. 20-25%
4) 241-260. 0-5%
rest 0%

Mark my words, the EVM rigging taking place right now will never, ever allow Modi led NDA to come within even 10 seats of absolute majority. Most likely is 200-230 to give SuSu and Laash the best crack at sabotaging Modi's claim and rallying maximum BJP MPs behind their push for a "secular, inclusive" alliance. That will fall apart within 18 months of chaos under a sustained barrage of (P)ukraine style, Ford-funded, AAP-driven public disorder, over a soul crushing backdrop of continuing economic insecurity and paid media fearmongering. Then back to regularly scheduled dynastic rule under Pappu & proxies.

Then again I could never be happier to get proved wrong about anything

vijaykarthik · Post by **vijaykarthik** » 09 Apr 2014 11:14

On a lighter note:

Which is the most popular cartoon character in the Indian news channels?

Answer

Arvind Kejriwal [Damn him. He can be made a permanent presence in the CNN,Pogo channels. Why come to the news channels and irritate 120 crore people?]