2014 USD/mBTC Price Prediction Contest

[sgbett]

forecasting 8 months out in month one feels a bit like a lottery imho and is 8x more work! my preference would be 1 month at a time, fwiw.

given my recent dream (yes sorry, i know!) of $64 price I would prefer if there were bands for

2.1 125.89
2.0 100
1.9 79.43
1.8 63.10
1.7 50.12
1.5 31.62
1.4 less

as they seem a bit top heavy at the moment (i.e. 100,000 is around 200x current price, whereas 158 is only around 1/3 current price) this covers going back all the way to retest the ~$32 ATH

I don't think its likely we will go that low but when you are dealing with assigning the chance of 1 in a million....

[rpietila]

forecasting 8 months out in month one feels a bit like a lottery imho and is 8x more work! my preference would be 1 month at a time, fwiw.

given my recent dream (yes sorry, i know!) of $64 price I would prefer if there were bands for

2.1 125.89
2.0 100
1.9 79.43
1.8 63.10
1.7 50.12
1.5 31.62
1.4 less

as they seem a bit top heavy at the moment (i.e. 100,000 is around 200x current price, whereas 158 is only around 1/3 current price) this covers going back all the way to retest the ~$32 ATH

I don't think its likely we will go that low but when you are dealing with assigning the chance of 1 in a million....

Thanks for the idea.

Yes I know I am in a bind, because the original idea of having 491 (and later, 5000) slots, was perceived too difficult to handle. Also the administrative cost would have been harder.

The linear 100-199, 200-299 bands become ridiculous after a few thousand$. And since the last year was 80x price appreciation, we will have to prepare for *minimum* $64,000 price by the end of 2014.

If the slots are in log scale, the number of them soon becomes large, and still the resolution in the middle of the scale is not too great (0.1 logpoints == 26%). I would perhaps go for 0.05 point resolution and 10...100,000 USD/BTC range, resulting in 80 slots. For me, the thinking is anyway a bigger exercise than filling in the numbers. And I would be producing a ready Excel sheet anyway.

As for the length of a single forecast, one month would be cool to keep the interest high if this was just a game. But my original idea (not yet presented in the forum) was indeed to have longer term also. But that would have been a separate competition. And we are already treading on the fine line of having enough interest on this one, and to have the right kind of interest that can think in the right terms and produce data which has predictive value.

Any ideas - muchos appreciados!

[sgbett]

id go for the 10-100000 on 0.05 log increments. i'd not realised that superseded the [P/Slot * width] system but i see now what you have done.

I see the benefit in collecting longer term forecasts to get an aggregate market outlook, so my complaint about having to do an 8 month forecast is really only minor i guess. I'll still enter regardless

[rpietila]

id go for the 10-100000 on 0.05 log increments. i'd not realised that superseded the [P/Slot * width] system but i see now what you have done.

I see the benefit in collecting longer term forecasts to get an aggregate market outlook, so my complaint about having to do an 8 month forecast is really only minor i guess. I'll still enter regardless

I think it'll be more interesting to make several months into the future all at one go. I (or somebody) will provide the historical price spreads over all the different timescales. The exponential trendline(s) will act as likely lower and upper bounds with high confidence.

[wachtwoord]

When a consensus system comes out please update it as I'm kind of losing track (sorry).

[rpietila]

Yeah, everybody will get an advance notice of 3 days.

OP updated to show the current, perhaps final, rules!!

[sgbett]

Looks good. Take my money!!!

[rpietila]

Looks good. Take my money!!!

I am currently making the excel with all the player aids, showing the points for each slot (should it win) automatically etc. Perhaps I get it ready even today! Mind sending your email address as a PM so you could betatest it?

[rpietila]

Excel is ready. Anyone who wants can betatest it, provided that he gives the email address and promises to do the testing immediately

[CoolStoryBro]

If your point total goes to -10 or worse, you are disqualified from the game immediately and your possible next round prediction is forfeited. This rule is to discourage trolling, gaming of the system and unresponsible forecasting. Getting -10 means that your estimated probability for a certain event which happened was 1/10 of the average. Certainly you were not being serious, were you?!

bad rule.

[rpietila]

If your point total goes to -10 or worse, you are disqualified from the game immediately and your possible next round prediction is forfeited. This rule is to discourage trolling, gaming of the system and unresponsible forecasting. Getting -10 means that your estimated probability for a certain event which happened was 1/10 of the average. Certainly you were not being serious, were you?!

bad rule.

I have experimented with this, and the reason for such a rule is to preserve all players in the game (by threatening to kick them out for non-performance).

What is wrong with it? If you want to max out your risk and leave some slots "unattended", the previous rules would anyway have doled you several minus-points, making it difficult to get back to the green. And to make it, you would have needed to gamble with the following months and make forecasts not based on what you think is probable but based on how you get back to the game. This is "gaming the system".

With a large number of other players (so that your guess does not too much move the average):

If there is only 2 choices of slots, BULL and BEAR, and others assign 80% to BULL. If you assign:

- 80% to BULL, you get 0 should BULL realize and 0 should BEAR realize
- 60% to BULL, you get -0.33 should BULL realize and +1 should BEAR realize
- 40% to BULL, you get -1 should BULL realize and +2 should BEAR realize
- 20% to BULL, you get -3 should BULL realize and +3 should BEAR realize
- 10% to BULL, you get -7 should BULL realize and +3.5 should BEAR realize
- 5% to BULL, you get -15 should BULL realize and +3.75 should BEAR realize <-- risky, because leads to elimination if BULL
- 1% to BULL, you get -79 should BULL realize and +3.95 should BEAR realize <-- risky, because leads to elimination if BULL


You can summon quite much volatility to your guess, but not arbitrarily much. The example shows that there is no merit in going gung-ho, but it disturbs the other players.

[sgbett]

If your point total goes to -10 or worse, you are disqualified from the game immediately and your possible next round prediction is forfeited. This rule is to discourage trolling, gaming of the system and unresponsible forecasting. Getting -10 means that your estimated probability for a certain event which happened was 1/10 of the average. Certainly you were not being serious, were you?!

bad rule.

I have experimented with this, and the reason for such a rule is to preserve all players in the game (by threatening to kick them out for non-performance).

What is wrong with it? If you want to max out your risk and leave some slots "unattended", the previous rules would anyway have doled you several minus-points, making it difficult to get back to the green. And to make it, you would have needed to gamble with the following months and make forecasts not based on what you think is probable but based on how you get back to the game. This is "gaming the system".

With a large number of other players (so that your guess does not too much move the average):

If there is only 2 choices of slots, BULL and BEAR, and others assign 80% to BULL. If you assign:

- 80% to BULL, you get 0 should BULL realize and 0 should BEAR realize
- 60% to BULL, you get -0.33 should BULL realize and +1 should BEAR realize
- 40% to BULL, you get -1 should BULL realize and +2 should BEAR realize
- 20% to BULL, you get -3 should BULL realize and +3 should BEAR realize
- 10% to BULL, you get -7 should BULL realize and +3.5 should BEAR realize
- 5% to BULL, you get -15 should BULL realize and +3.75 should BEAR realize <-- risky, because leads to elimination if BULL
- 1% to BULL, you get -79 should BULL realize and +3.95 should BEAR realize <-- risky, because leads to elimination if BULL


You can summon quite much volatility to your guess, but not arbitrarily much. The example shows that there is no merit in going gung-ho, but it disturbs the other players.

in principle I agree with this but I think the cutoff of -10 is pretty tight. If you believe in volatility (you should! ) and pitch your guesses over a wide spread, but then everyone else just does a normal distribution centred round the you could lose a lot of points on what would be a legitimate entry (i.e. not attempting to game the system).

I understand this rule is focused on increasing the quality of the aggregate consensus, but it might be forcing people to guess much more conservatively, instead of actually putting down what they think for fear of getting disqualifed. I think that would be just as much of a compromise to the quality of the predictive power.

[rpietila]

in principle I agree with this but I think the cutoff of -10 is pretty tight. If you believe in volatility (you should! ) and pitch your guesses over a wide spread, but then everyone else just does a normal distribution centred round the you could lose a lot of points on what would be a legitimate entry (i.e. not attempting to game the system).

I understand this rule is focused on increasing the quality of the aggregate consensus, but it might be forcing people to guess much more conservatively, instead of actually putting down what they think for fear of getting disqualifed. I think that would be just as much of a compromise to the quality of the predictive power.

I just changed it to -20.

The example above should indicate that you don't really get that much more points when you bet a certain range with as much probability as you can muster: there is a law of sharply declining marginal returns in the lucky prediction and the law of exponentially-growing punishment in the unlucky prediction.

I know that the rule alone (with or without cutoff grades) does make people conform to conformism, but this is a serious prediction contest and not a lottery.

[sgbett]

in principle I agree with this but I think the cutoff of -10 is pretty tight. If you believe in volatility (you should! ) and pitch your guesses over a wide spread, but then everyone else just does a normal distribution centred round the you could lose a lot of points on what would be a legitimate entry (i.e. not attempting to game the system).

I understand this rule is focused on increasing the quality of the aggregate consensus, but it might be forcing people to guess much more conservatively, instead of actually putting down what they think for fear of getting disqualifed. I think that would be just as much of a compromise to the quality of the predictive power.

I just changed it to -20.

The example above should indicate that you don't really get that much more points when you bet a certain range with as much probability as you can muster: there is a law of sharply declining marginal returns in the lucky prediction and the law of exponentially-growing punishment in the unlucky prediction.

I know that the rule alone (with or without cutoff grades) does make people conform to conformism, but this is a serious prediction contest and not a lottery.

-20 is better, i was going to suggest something like that, but instead I thought I'd start the negotiation at total removal hehe

[sgbett]

Also, I'm ready when you are. My probabilities are calculed! My bitcoin urges to reach you!

Did you really knock it down to 0.05? bloody cheapskates with the amount of time it takes to enter complaining about a mere $50 of BTC makes me think these people need to get proper jobs! hehe

[rpietila]

Also, I'm ready when you are. My probabilities are calculed! My bitcoin urges to reach you!

Did you really knock it down to 0.05? bloody cheapskates with the amount of time it takes to enter complaining about a mere $50 of BTC makes me think these people need to get proper jobs! hehe

I made it such that you can enter with 5 different scenarios for 0.25, if you want to maximize your chances and blow your brain in outsmarting your opponents! 

[sgbett]

It would be nice to have some opponents! 

[rpietila]

There is still a week to go, and 2200 readers of the thread. Not all are willing to show that they are entering. Also they can enter at any further point of time. It is hardly likely that you will be alone for the whole duration of the competition, and should that happen, it only guarantees you the prize pot (jackpot is always at least 1,000mBTC, otherwise see the prize pot distribution in the OP).

If you seriously think that there will be nobody else in the first round, perhaps making two entries to play against each other would be the way to go: If the price goes up, then "sgbett_bull" has a huge advantage because it is the only one getting pluspoints for the 8 month-predictions. If the price is sluggish, then "sgbett_bear" is gaining ground. But if one of them gets knocked out, the other makes only 0 points for those months where there was no other entries.... 

Perhaps you see now why the -20 rule is important 

[durrrr]

i believe that 1mbtc belongs right around 1$ yes it is undervalued at the moment but theres so many things going against bitcoin that as soon as some more good things happen we will see 1$/ mbtc again

[rpietila]

FINAL BETA RULES UPDATED IN THE OP

THE PREDICTION SHEET IS NOW DOWNLOADABLE

COMPETITION STARTS TOMORROW

DEADLINE FOR ENTRY IS 2014-5-5 23:59 GMT